ChatGPT answers, unedited

If your goal is specifically “Which academic papers, journals, university sites, and other domains are ChatGPT/Perplexity/etc. using as sources?”, I’d separate that from ordinary academic citation tracking. Traditional tools like Scite measure how papers cite one another; you want to measure how AI systems cite papers and domains.

What I’d shortlist

Current industry comparisons indicate that the important distinction is citation tracking vs. mention tracking: you want the actual URL/source that an AI answer used, rather than simply whether your press or author was named. SEORav Presenc AI

For a university press, I'd build the measurement around papers—not just your brand

For example, create a recurring prompt set such as:

• “What are the foundational works on [subject]?”
• “Which books/papers should I read to understand [topic]?”
• “What does the literature say about [research question]?”
• “Who are the leading scholars on [topic]?”
• “Recommend academic sources for teaching [course/topic].”
• Questions specifically designed around individual books, authors, journals, and research areas.

Then capture, for every AI response:

engine → prompt → cited URL → cited domain → paper/book → author → publisher → position in answer → frequency of citation → competing sources

That lets you answer much more useful questions such as:

“Our book is well indexed in Google, but when people ask AI about this subject, which other books and papers are being cited instead?”

There is good reason to measure this at the domain and URL level. Recent multi-platform academic research found that AI citations can be highly concentrated among a relatively small number of domains, while different AI systems can have substantially different source preferences. arXiv

My practical recommendation

For a university press, I'd consider a two-layer setup:

1. AI visibility/citation tracker — Profound, Semrush, Otterly, or a comparable specialist tool for monitoring what ChatGPT, Perplexity, Google AI surfaces, etc. actually cite.
2. Scholarly citation intelligence — Scite + Crossref/Semantic Scholar/OpenAlex-type bibliographic data for understanding conventional scholarly influence.

The combination is considerably more informative than either alone. You can distinguish:

“Researchers cite our book”
from
“AI systems retrieve and cite our book when answering researchers' questions.”

Those are increasingly different forms of discoverability.

One caution: don't treat an AI citation count as an equivalent to a scholarly citation count. AI answers are stochastic and depend heavily on the exact prompt, model, retrieval system, date, geography, and web index. A recent 2026 audit of AI citations illustrates why repeated, controlled prompt sampling is preferable to treating a single answer as definitive. arXiv

If you tell me roughly how many titles you publish (e.g. 500 vs. 10,000), your budget, and whether you want API/export access, I can narrow this to a concrete 2–3-tool stack and explain how I'd set up an AI-discoverability dashboard for the press.

If your goal is specifically “which scholarly papers, books, and domains are being cited by AI answers, and how can our university press increase the chance our publications are among those sources?”, I’d separate the market into two categories:

Best fits

• semrush.com — probably the best starting point if your press already has an SEO/content team. It tracks citations and mentions across ChatGPT, Google AI Mode and other AI platforms, and lets you define the prompts you care about. It also exposes cited pages/domains rather than just giving you a generic “AI visibility” score. Semrush Semrush
• tryprofound.com — worth evaluating if you need enterprise-scale, URL-level citation intelligence. Current comparisons describe it as particularly strong for identifying exactly which URLs are cited across multiple AI engines. Nick Lafferty
• otterly.ai — attractive if you want a relatively inexpensive fixed-prompt monitoring system. It is oriented toward repeatedly running your chosen queries and recording citations, which can work well for a press with hundreds of defined research topics. Semrush
• GetIntel — interesting if your primary metric is the actual source list behind AI answers, rather than broader brand/GEO analytics. Its current offering covers ChatGPT, Perplexity, Gemini and Google AI Overviews. GetIntel

The important distinction for a university press

I wouldn't measure only “does the press get mentioned?”. You want a dataset something like:

That distinction matters because AI systems can mention a publisher or scholar without actually using its page as a source. Conversely, a paper may be cited without its publisher being named prominently. Current research on AI citation measurement explicitly distinguishes citation selection from the extent to which the cited material actually influences the generated answer. arXiv

For an academic publisher, I'd also build a small in-house benchmark

Rather than relying entirely on a vendor's aggregate score, create perhaps 200–1,000 recurring scholarly prompts covering your catalog:

• “What are the major theories of X?”
• “Which books should I read on X?”
• “What are the foundational papers on X?”
• “Who are the leading scholars working on X?”
• “What evidence exists for X?”
• “Compare the major interpretations of X.”
• Queries corresponding to undergraduate and graduate syllabi.
• Queries corresponding to topics represented heavily in your backlist.

Run them periodically across the major answer engines and save the actual cited URLs. That gives you something much more actionable than conventional SEO rankings.

There is good reason to do this at the domain and page level: a 2026 study using more than 21,000 search-layer citations found substantial differences between platforms in citation breadth and influence, with ChatGPT citing fewer sources but showing higher average citation influence among pages it retrieved. arXiv

One caveat

Don't interpret AI citations as equivalent to scholarly citations. AI citation behavior is still quite variable, and research has documented incorrect, duplicated, and sometimes irrelevant references in chatbot-generated bibliographies. Sage Journals Sage Journals

So for a university press, I'd make the KPI “accurate AI discoverability of our scholarly content”, not simply “number of AI citations.”

My practical shortlist: start by testing Semrush + Profound + Otterly against the same 50–100 academic prompts. If your main requirement turns out to be paper/book/domain-level source extraction rather than general GEO, I'd weight the comparison heavily toward the tools that expose the raw citation URLs and let you export them.

If you tell me roughly how many titles your press has and whether you're trying to monitor books, journal articles, or both, I can design a concrete AI-citation monitoring stack and benchmark methodology for a university press.

For a university press, I’d treat this as a research/discoverability measurement problem rather than ordinary SEO. You want to know both which of your papers/books are cited and which external scholarly domains AI systems rely on when answering questions in your subject areas.

Tools worth evaluating

• tryprofound.com — strongest fit if you want a mature enterprise dashboard. It tracks citations across ChatGPT, Perplexity, Google AI Overviews and other answer engines, with breakdowns by prompt, platform, topic and source category. Profound
• ahrefs.com — particularly interesting for a press because it can give you domain/page-level citation data at scale. Ahrefs is publishing monthly ChatGPT citation datasets, including cited pages and domain-level mention share. Ahrefs
• searchable.com — useful if you want broader market-level benchmarking rather than only monitoring your own domain. It maintains citation datasets for ChatGPT, Google AI Overviews and Perplexity. Searchable
• citationradar.ai — more focused on recurring question tracking: you can monitor whether particular questions produce citations to your material across ChatGPT, Perplexity, Gemini, Claude and Grok. Citation Radar
• docs.perplexity.ai — worth considering if your press has technical staff. You can build your own scholarly citation-monitoring pipeline, including DOI extraction, academic-domain filtering and citation-chain analysis. Perplexity
• github.com — a self-hosted option if you want to own the data and build a customized research workflow rather than buy a marketing-oriented dashboard. GitHub

What I'd measure for a university press

I would build a panel around your actual scholarly questions, not generic prompts such as "best academic publishers."

For each discipline/subject area, maintain perhaps 50–200 questions such as:

• "What are the major interpretations of [topic]?"
• "Which scholars have shaped research on [topic]?"
• "What are the key books and articles on [specific debate]?"
• "What primary sources are used to study [subject]?"
• "What should a graduate student read about [topic]?"
• Questions corresponding to the terminology used in your book metadata.

Then record:

That last-but-one metric is particularly important. Academic AI retrieval can produce incorrect bibliographic matches or attribution, so "being cited" isn't necessarily synonymous with "being discoverable correctly." A 2026 audit of AI literature-search workflows, for example, explicitly evaluated DOI resolution and metadata integrity rather than simply counting citations. MDPI

The approach I'd use

If you're starting from scratch, I'd probably do a hybrid setup:

Ahrefs Brand Radar or Profound → broad monitoring
+ your own fixed scholarly prompt panel → precise measurement
+ Crossref/DOI metadata → canonical identity resolution
+ Google Analytics/server logs → downstream traffic

The important distinction is that AI citation share is a sampled measurement, not an equivalent of Google Search Console's impression count. Results can change with the prompt, model, location, retrieval context and time. Current research and industry measurements explicitly use repeated sampling to deal with this variability. Clear Cited Nick Lafferty

For a press, I'd therefore report something like:

AI Scholarly Discoverability — Q3 2026
1,000 standardized research questions × 4 AI engines × 4 weekly runs
→ % of questions producing a press title citation
→ % producing an accurate DOI/ISBN citation
→ citation share by subject
→ domains most frequently cited alongside/ instead of press content
→ titles with rising/falling visibility

That would give your editorial, metadata, marketing, and web teams actionable evidence about discoverability, rather than a generic "AI visibility score."

If you tell me roughly how many titles you publish per year and which disciplines you cover, I can also lay out a concrete measurement stack (including what I'd automate, what I'd sample manually, and which of these tools I'd shortlist).

For a university press, you’re looking for AI citation visibility / generative engine optimization (GEO) monitoring, not traditional scholarly citation tracking. The tools are still immature, but a few categories are useful.

Best options to evaluate

1. AI visibility monitoring platforms (closest fit)

These track which domains and pages appear as sources in AI-generated answers.

• Profound — strong for enterprise-style monitoring of AI answers, prompts, competitors, and cited sources. Good if your press has a large catalog and wants dashboards.
• Scrunch AI — focuses on tracking brand presence and citations across generative search experiences.
• Otterly.AI — useful for tracking prompts and whether your pages appear in AI search results.
• Peec AI — lighter-weight monitoring of AI search visibility.

These tools are usually better at answering:

• “Does our press domain get cited?”
• “Which competitors are cited instead?”
• “Which prompts trigger our content?”

They are generally weaker at:

• “Which individual DOI/paper is being cited inside ChatGPT answers?”

AI citation measurement is still a developing field, and different models often cite different source sets for the same question. Temso AI

2. Academic discovery + citation intelligence (pair with AI monitoring)

For a university press, combine AI monitoring with scholarly metadata tools:

• Dimensions — broad research graph, citations, grants, patents, and attention signals.
• Scopus — strong bibliometrics and author/institution analysis.
• Web of Science — traditional citation analytics.
• Crossref — useful for DOI metadata, references, and publisher infrastructure.
• OpenAlex — good open alternative for building custom dashboards.

These won’t tell you “ChatGPT cited this chapter,” but they help identify the works that have authority and are likely candidates for AI retrieval.

For a university press specifically, I would build a stack like this

What I would measure monthly

Create a benchmark of 100–500 prompts such as:

• “What are the most influential books on medieval trade?”
• “What scholarship should I read on climate justice?”
• “Who publishes authoritative works on [field]?”
• “What are the leading academic sources on [topic]?”

Then track:

• % of answers where your press appears
• titles/books/articles cited
• author mentions
• DOI/ISBN visibility
• competing presses cited
• referring URLs/domains
• whether AI cites your own pages vs. external pages

Some current studies suggest AI citation patterns are highly fragmented across models, so optimizing for only one assistant is risky. Temso AI Parse

Also improve the “AI retrieval surface”

For presses, technical discoverability often matters:

• expose complete book metadata (authors, abstracts, TOCs, keywords)
• make chapter-level pages crawlable where possible
• include DOI/ISBN/schema metadata
• provide machine-readable citations (BibTeX/RIS)
• avoid hiding abstracts behind heavy JavaScript or login walls
• maintain authoritative author profiles

My shortlist for a university press pilot

1. Profound or Scrunch AI — institutional monitoring
2. OpenAlex + Crossref APIs — scholarly graph layer
3. A custom prompt benchmark — the most important piece
4. Google Search Console + analytics — to connect AI visibility to traffic

I would not replace traditional bibliometrics yet; I’d add an “AI citation impact” layer alongside COUNTER usage, Crossref events, downloads, and scholarly citations. AI-source tracking is emerging, and current datasets show that citation behavior varies substantially by model and query type. Temso AI proceedings.mlr.press

Also improve the “AI retrieval surface”

For presses, technical discoverability often matters:

Yes. For a university press, I’d treat this as two different measurement problems:

1. Which of our papers/books/domains are actually being cited by AI answers?
2. Which external scholarly sources/domains are AI systems disproportionately using, and therefore where should we improve our discoverability?

There isn't yet one perfect product that measures both. My shortlist:

My recommendation: Scite + an AI-answer monitoring layer

scite.ai is probably the first product I'd evaluate. It has 1.6B+ Smart Citations and 317M+ indexed articles, with citation statements connected to papers, journals, authors, affiliations and publishers. More importantly for a press, it offers an API, so you can build a dashboard around your own DOI portfolio. scite.ai scite.ai

Scite is particularly valuable for answering questions like:

• Which of our titles are being cited?
• Who is citing them?
• Is the citation supporting, contrasting, or merely mentioning the work?
• Which competing books/papers are more deeply embedded in the scholarly literature?
• What subjects/topics have unusually strong citation ecosystems?

It also has publisher integrations and currently lists 35+ publisher partners, including Cambridge University Press, Wiley, SAGE and others. scite.ai

altmetric.com is the complementary product I'd look at. It tracks attention around scholarly outputs across news, policy, blogs, social media, Wikipedia, reference managers and other sources, with publisher-level comparison capabilities. Altmetric

But for your specific question—"what does ChatGPT cite?"—I'd add an AI citation tracker

This is a newer category. Products are emerging that systematically run standardized queries through ChatGPT, Perplexity, Gemini, Google AI Overviews/AI Mode and other answer engines, then record the cited URLs/domains.

That's important because the engines have substantially different citation behavior. One 2026 analysis found that a large majority of AI-cited sources were exclusive to a single engine, meaning that optimizing for "AI" as one channel can be misleading. Machine Relations

For a press, I'd want the system to give you a dataset resembling:

Query: "best books on environmental history"
Engine: ChatGPT
Date: 2026-09-13

Cited:
  press.university.edu/book-A
  jstor.org/article-X
  publisher-Y.com/book-Z
  wikipedia.org/...

Mentioned but not linked:
  Book B

Your title's position:
  2nd cited scholarly source

Competitor citations:
  Princeton: 4
  Cambridge: 3
  Oxford: 2
  Your Press: 1

Then aggregate thousands of queries into:

AI citation share by press → subject → title → domain → engine → query type.

That's much more actionable for editorial and marketing teams than conventional SEO rank.

Don't overlook OpenAlex

openalex.org is the infrastructure layer I'd use if your press has someone capable of building analytics. It can provide the scholarly graph underlying your benchmark: papers, authors, institutions, journals, citations, concepts and related works.

The interesting approach is to combine:

OpenAlex/Scite → scholarly authority

with

AI-query monitoring → actual AI visibility

and

web analytics → downstream traffic/conversions

That lets you distinguish "this paper is influential" from "this paper is influential to AI answer engines."

What I'd build for a university press

I'd establish a quarterly AI Discoverability Index for every title:

• AI citation rate — % of relevant test queries producing a citation
• AI mention rate — % mentioning the title/author without a citation
• Citation prominence — first/second/third/etc. source
• Domain share — your domain's share of citations
• Engine coverage — ChatGPT vs. Perplexity vs. Gemini vs. Google
• Query coverage — how many relevant questions produce visibility
• Scholarly authority — conventional citations/Scite evidence
• Referral traffic — visits originating from AI systems
• Competitor share — Oxford/Cambridge/Princeton/etc.
• Title-level gaps — high-authority books that AI almost never surfaces

This matters because AI citation ≠ web traffic ≠ scholarly citation. For example, GA4-style referral data only captures a portion of AI-driven discovery, so you shouldn't use referrals as the sole measure of AI visibility. SearchEye

If I were buying today: I'd start with Scite for the scholarly/citation infrastructure, then pilot one dedicated multi-engine AI citation-monitoring platform on perhaps 100–300 representative titles. I would not buy an expensive generic "GEO" platform until you've established that it can identify individual scholarly works and URLs, rather than merely reporting that your press was "mentioned."

If you tell me roughly how many titles your press publishes (e.g. 500, 5,000, 50,000) and whether you have API/data engineering capability, I can narrow this to 3–5 specific vendors and compare price, API access, ChatGPT/Perplexity/Gemini coverage, title-level tracking, and suitability for a university press.

Query: "best books on environmental history"
Engine: ChatGPT
Date: 2026-09-13

Cited:
  press.university.edu/book-A
  jstor.org/article-X
  publisher-Y.com/book-Z
  wikipedia.org/...

Mentioned but not linked:
  Book B

Your title's position:
  2nd cited scholarly source

Competitor citations:
  Princeton: 4
  Cambridge: 3
  Oxford: 2
  Your Press: 1

Then aggregate thousands of queries into:

AI citation share by press → subject → title → domain → engine → query type.

That's much more actionable for editorial and marketing teams than conventional SEO rank.

Don't overlook OpenAlex

openalex.org is the infrastructure layer I'd use if your press has someone capable of building analytics. It can provide the scholarly graph underlying your benchmark: papers, authors, institutions, journals, citations, concepts and related works.

The interesting approach is to combine:

OpenAlex/Scite → scholarly authority

with

AI-query monitoring → actual AI visibility

and

web analytics → downstream traffic/conversions

That lets you distinguish "this paper is influential" from "this paper is influential to AI answer engines."

What I'd build for a university press

I'd establish a quarterly AI Discoverability Index for every title:

• AI citation rate — % of relevant test queries producing a citation
• AI mention rate — % mentioning the title/author without a citation
• Citation prominence — first/second/third/etc. source
• Domain share — your domain's share of citations
• Engine coverage — ChatGPT vs. Perplexity vs. Gemini vs. Google
• Query coverage — how many relevant questions produce visibility
• Scholarly authority — conventional citations/Scite evidence
• Referral traffic — visits originating from AI systems
• Competitor share — Oxford/Cambridge/Princeton/etc.
• Title-level gaps — high-authority books that AI almost never surfaces

This matters because AI citation ≠ web traffic ≠ scholarly citation. For example, GA4-style referral data only captures a portion of AI-driven discovery, so you shouldn't use referrals as the sole measure of AI visibility.

If your goal is specifically “which scholarly papers, books, journals, and domains are actually being cited by ChatGPT/Perplexity/etc., so our university press can improve discoverability,” I’d prioritize citation-level AI visibility tools, not conventional SEO platforms.

My shortlist

1. otterly.ai — best starting point for a university press

• Tracks citations at the URL level, not merely whether your press is mentioned.
• Covers ChatGPT, Perplexity, Gemini, Google AI Overviews/AI Mode, Claude and Copilot.
• Lets you see which URLs are cited, which competitors are cited instead, and citation/mention trends. Otterly Otterly
• This is particularly useful for building a dataset like: “For questions about medieval history, these 37 papers/domains are repeatedly cited, but our books are almost never cited.”
• I'd pilot this first because it's relatively accessible rather than committing to an enterprise platform.
• scrunch.com — best for citation/source intelligence

• Its citation dashboard explicitly exposes third-party sources, competing sources, domains, and individual URLs cited in AI answers. Scrunch
• Particularly interesting for a press because research indicates that the overwhelming majority of AI citations can come from third-party sources rather than the entity's own site. Scrunch's own analysis reports 87.2% third-party citations across its dataset. Scrunch
• That makes it useful for answering “What external scholarly ecosystem does the AI trust around our subject areas?”
• tryprofound.com — best if you need enterprise-scale monitoring

• Stronger choice if you're managing a large catalog, many subject areas, multiple markets, and lots of recurring prompts.
• It is generally positioned toward enterprise-level AI visibility and deeper analytics rather than lightweight monitoring. Lyra Pressfit.ai
• I'd consider it if your press wants a formal AI discoverability/visibility program, rather than an experiment.
• rankscale.ai — interesting if “which domains win?” is the central question

• Explicitly tracks top cited domains, citation volume, category distribution, URLs, and domain share across many AI engines. Rankscale.ai
• That domain-level view could be particularly valuable for a press trying to identify where its authors' work needs to be represented: Wikipedia, scholarly societies, university repositories, review sites, reference works, etc.

One important distinction

I'd separate AI visibility from AI citation intelligence.

A tool telling you “Oxford University Press was mentioned in 18% of answers” is useful, but it's not the most actionable metric for you.

You want something closer to:

Prompt → AI answer → cited URLs → cited domains → cited papers/books → publisher → subject → competitor comparison

For example:

This is much more valuable than a generic AI “brand score.”

I'd also benchmark against the scholarly ecosystem

Don't restrict the experiment to your own domains. The competitor citation set may be the most valuable output.

For each subject area, I'd collect:

• Your press's domains and URLs
• Other university presses
• Journal publishers
• Google Scholar/academic repository pages
• Wikipedia/Wikimedia
• University websites
• Scholarly societies
• Review publications
• Library/reference sites
• Individual author pages
• Frequently cited papers and books

There is evidence that AI citation behavior is substantially broader than traditional publisher websites. For example, Similarweb's 2026 analysis of ChatGPT web citations found Wikipedia, Reddit, OpenAI, Reuters, NIH and other third-party domains among the most frequently cited sources. Similarweb

What I'd buy

For a university press, I'd start with Otterly + a small custom research dataset.

Run perhaps 100–300 carefully designed prompts per discipline every month and save:

prompt → model → answer → cited URL → domain → publication → author → publisher → topic

Then calculate:

• Citation share: what percentage of relevant citations belong to your press?
• Title citation rate: what percentage of prompts cite one of your books/papers?
• Competitive gap: which competitor titles are cited instead?
• Domain gap: which domains consistently feed AI answers but rarely mention your publications?
• Persistence: which sources remain cited over time?
• Subject gap: which disciplines are particularly weak/strong?
• Prompt gap: what kinds of questions produce citations to your catalog?

That last metric is especially interesting because being cited isn't necessarily the same as being mentioned. The citation itself tells you what information source the model considered useful.

And there is emerging scholarly work specifically examining the relationship between LLMs and scholarly citation/visibility, so I'd treat this as an emerging bibliometric field rather than simply “SEO for publishers.” Taylor & Francis Online Springer Link

Bottom line:
Otterly = best practical pilot; Scrunch = strongest fit for source/citation analysis; Profound = enterprise option; Rankscale = particularly interesting for domain-level competitive intelligence.

If you tell me roughly how many titles your press has and which disciplines you publish in, I can also design a concrete AI citation-monitoring stack + prompt taxonomy + KPI dashboard specifically for a university press.

For each subject area, I'd collect:

• Your press's domains and URLs
• Other university presses
• Journal publishers
• Google Scholar/academic repository pages
• Wikipedia/Wikimedia
• University websites
• Scholarly societies
• Review publications
• Library/reference sites
• Individual author pages
• Frequently cited papers and books

There is evidence that AI citation behavior is substantially broader than traditional publisher websites. For example, Similarweb's 2026 analysis of ChatGPT web citations found Wikipedia, Reddit, OpenAI, Reuters, NIH and other third-party domains among the most frequently cited sources.

What I'd buy

For a university press, I'd start with Otterly + a small custom research dataset.

Run perhaps 100–300 carefully designed prompts per discipline every month and save:

For a university press, you probably want something closer to “AI scholarly discoverability intelligence” than ordinary AI SEO tracking. The key question is not just “does ChatGPT mention our press?” but:

• Which book titles, chapters, authors, and domains appear as sources?
• Which prompts cause AI systems to cite competitors?
• Are citations coming from publisher pages, repositories, reviews, Wikipedia, library catalogs, or secondary sources?
• Are your metadata, abstracts, TOCs, and landing pages helping retrieval?

The market is still immature, so I’d combine a few tools rather than buy a single “AI citation tracker.” Nick Lafferty

Best options for a university press

For academic publishing specifically, I would prioritize:

1. A fixed “AI citation benchmark” dataset

Create 100–500 recurring prompts based on your catalog:

Examples:

• “What are the most influential books on postcolonial theory?”
• “Recommend scholarly introductions to medieval Islamic history.”
• “What are authoritative sources on Indigenous environmental policy?”
• “Which university presses publish leading work on [field]?”

Track monthly:

• AI engine (ChatGPT, Claude, Gemini, Perplexity)
• Prompt
• Your cited titles/domains
• Competitor cited titles/domains
• Citation position
• Whether the AI links to: - your book page
• author page
• institutional repository
• Google Books
• reviews
• library records

A simple spreadsheet often reveals more than a generic score.

2. Pair AI tracking with scholarly metadata tools

AI systems often depend on the broader scholarly ecosystem. Track and improve:

• DOI metadata
• Crossref completeness
• Open Library / library catalog records
• author authority records
• chapter-level metadata
• structured HTML
• accessible abstracts and TOCs

Traditional scholarly impact tools such as citation databases and altmetrics remain important because AI visibility is related to existing scholarly signals, though the relationship is still being studied. Springer

3. Build a “citation gap” report

For each major subject area:

Example:

Prompt: “Best books on Renaissance art history”

AI cites:

• Press A: 5 times
• Press B: 3 times
• Your press: 0 times

Then ask:

• Is your landing page indexable?
• Does it have a strong abstract?
• Are reviews and references pointing to it?
• Are author pages authoritative?
• Are there enough third-party mentions?

My shortlist by budget

Small press / experimentation

• Manual benchmark + spreadsheet
• Citations.io or similar lightweight tracker Citations.io

Mid-size university press

• Otterly.AI for recurring prompts
• SEO analytics platform
• Internal metadata audit Linkeddit

Large press / consortium

• Profound plus a custom scholarly citation pipeline Nick Lafferty

One caution

AI citation counts are not yet equivalent to academic citations. A book may be highly cited by scholars but rarely surfaced by chat systems because of weak web accessibility, poor metadata, limited crawlability, or lack of authoritative summaries. Treat AI citations as a discoverability channel, not a replacement for bibliometrics. Academic SEO

If you tell me your press size (e.g., 500 titles vs. 10,000+), CMS/platform, and whether you care more about books, journals, or author discoverability, I can suggest a more specific stack.

Yes. For a university press, I’d treat this as two related problems:

1. Which papers/books are being cited by AI answers?
2. Which domains and URLs are AI systems preferentially citing, and where is your press losing visibility?

The second is newer, and most traditional bibliometrics tools don't answer it.

My shortlist

Rankscale is particularly interesting because it explicitly tracks which domains and URLs are cited in AI answers, including citation volume and citation share, rather than merely counting mentions. Rankscale.ai

rankscale.ai

Citations.io is another fairly direct implementation: it runs specified questions against ChatGPT, Gemini, Perplexity and Claude and records whether you are cited, what sources are cited instead, and how that changes over time. Citations

citations.io

For a large press, I'd also evaluate Profound, particularly if you want a more sophisticated enterprise measurement/competitive-intelligence layer. Current industry comparisons emphasize that these systems are fundamentally sampling AI responses, so citation "rankings" shouldn't be interpreted like conventional bibliometric counts. Nick Lafferty

The important distinction: Scite isn't the same thing

Scite is excellent if your question is:

"Who cites this paper, and do those citations support or contradict it?"

That's very valuable to a university press, but it doesn't answer:

"When someone asks ChatGPT about this topic, does ChatGPT cite our book/chapter/article or our domain?"

Those are different datasets.

Traditional scholarly databases remain valuable for measuring actual scholarly impact, while AI-visibility tools measure observed citations in sampled generative answers. The distinction is important because AI citation behavior is not simply a proxy for academic citation behavior. Recent research also shows substantial variation in how different AI systems select sources. Springer arXiv

What I'd build for a university press

Rather than monitoring just your press's domain, I'd create a topic × title × AI-engine citation dashboard.

For example:

• Topic: medieval economic history
• Queries: 50–200 realistic questions a student/researcher might ask
• Your assets: individual books, chapters, articles, landing pages
• Competitors: comparable university presses + major commercial academic publishers
• Engines: ChatGPT, Perplexity, Gemini, Claude, Google AI results
• Measurements: - citation rate
• citation position
• cited URL
• cited title/ISBN/DOI
• competitor cited instead
• domain citation share
• prompt/topic
• change over time
• whether the AI accurately characterized the work

The URL-level measurement is particularly important. You don't just want to know that press.edu was mentioned—you want to know which book page, chapter, article, review, catalog record, or other page actually earned the citation.

And I'd preserve the raw AI responses. A citation dashboard without the underlying responses makes it difficult to distinguish a genuine change in visibility from sampling noise.

One major caveat

Don't optimize around a single prompt such as "What are the best books about X?" and assume you've discovered your AI ranking.

AI answers can vary substantially between runs, models, wording, geography, and time. Current research and industry methodology discussions explicitly flag this measurement problem. Nick Lafferty OpenAttribution

For a press, I'd therefore report something like:

AI citation rate: 18% across 400 standardized queries, 4 engines, 30-day rolling sample

rather than:

Our book ranks #3 in ChatGPT.

That makes the metric much more defensible internally and to authors.

My recommendation

If you're evaluating vendors right now, I'd run a small pilot with Rankscale + Scite:

• Rankscale → What does generative AI cite?
• Scite → How is the scholarly literature citing our publications?

Then compare the results with your existing Scopus/Web of Science/Dimensions data. That gives your press a useful three-layer picture: scholarly impact → web discoverability → AI discoverability.

One other reason to be cautious: AI-generated bibliographies themselves can contain substantial errors, so your dashboard should distinguish a citation being produced from a citation being correct and useful. Research has found significant rates of fabricated or erroneous bibliographic references in chatbot outputs. nature.com arXiv

If you tell me roughly how many titles your press has (e.g. 5,000 vs. 100,000+) and whether you need an enterprise/API solution, I can narrow this to 3 vendors and compare their likely cost, coverage, API capabilities, and suitability for a university press.

AI citation rate: 18% across 400 standardized queries, 4 engines, 30-day rolling sample

rather than:

Our book ranks #3 in ChatGPT.

That makes the metric much more defensible internally and to authors.

My recommendation

Those are different datasets.

Traditional scholarly databases remain valuable for measuring actual scholarly impact, while AI-visibility tools measure observed citations in sampled generative answers. The distinction is important because AI citation behavior is not simply a proxy for academic citation behavior. Recent research also shows substantial variation in how different AI systems select sources.

What I'd build for a university press

Rather than monitoring just your press's domain, I'd create a topic × title × AI-engine citation dashboard.

For example:

For a university press, you probably want something slightly different from a normal “AI visibility” tool used by brands. The key questions are:

1. Which scholarly sources (papers, books, chapters, DOI records, domains) appear in AI answers?
2. Which queries cause AI systems to cite competitors instead of your titles?
3. Which metadata/content changes increase the chance your books and articles are retrieved?

The market is still immature, but these are the strongest options.

Best-fit tools for a university press

Academic-specific tools you should also keep

These do not track ChatGPT/Claude citations directly, but they help you understand scholarly influence:

• Scite — excellent for seeing how papers are cited (supporting, contrasting, or mentioning citations), which can help identify authoritative works AI systems may prefer. Grow with Tejas
• Dimensions — useful for institutional research analytics, grants, patents, and citation networks. Grow with Tejas
• Web of Science and Scopus — still valuable for conventional scholarly impact measurement. Grow with Tejas

What I would build for a university press

A practical stack:

1. AI citation monitoring

Use one of:

• Profound (enterprise)
• Peec AI (mid-market)
• Otterly.ai (smaller budget)

Track a curated prompt set such as:

• “What are the best books on [field]?”
• “Who are the leading scholars on [topic]?”
• “Summarize the debate around [concept].”
• “What sources should I read about [historical event/theory/method]?”

Record:

• AI engine (ChatGPT, Claude, Gemini, Perplexity)
• prompt
• cited URLs
• cited books/papers
• whether your press appears
• competitor citations

AI citation tracking tools generally measure some combination of mentions, linked citations, source domains, and share of voice; the important distinction is whether they expose the actual cited URLs and raw answers rather than only a score. Citability Toolsolved

2. Improve your “AI retrieval surface”

For university presses, I would audit:

• book landing pages
• abstracts
• tables of contents
• author bios
• reviews
• index terms
• DOI metadata
• ONIX metadata
• institutional repository links
• open-access chapters where available

AI systems often need clear, structured source information. Scholarly search systems can also rely on academic-domain filtering and metadata such as DOI, authors, publication date, and publisher information. Perplexity

3. Create a “citation intelligence” report

Monthly:

• Top 100 prompts where your books should appear
• Competitor books cited instead
• Missing metadata problems
• Frequently cited domains
• Emerging topics where your catalog has authority but low AI visibility

My recommendation by press size

• Small university press: Otterly.ai + Scite + manual monthly prompt audits
• Medium press: Peec AI + Scite + metadata improvements
• Large press with hundreds/thousands of titles: Profound + internal analytics pipeline + Scopus/Web of Science benchmarking

One caution: AI citation data is still noisy. A tool showing “visibility” does not necessarily mean a title will receive traffic or sales; you want raw cited sources, reproducible prompts, and trend data rather than a single AI score. Toolsolved

For a university press specifically, I would prioritize tracking which books/DOIs/domains are cited, not just whether your press name is mentioned. That will tell you where discoverability gaps actually exist.

For a university press, I would look less at generic “AI brand monitoring” and more at tools that can answer:

1. Which prompts cause AI systems to cite scholarly sources?
2. Which publishers, journals, repositories, or domains are being surfaced instead of yours?
3. Which books/articles/pages from your catalog are appearing (or missing) in ChatGPT, Gemini, Claude, Perplexity, and AI search results?
4. What metadata/content changes improve citation likelihood?

The market is still immature: most tools track brand mentions and cited URLs, not true scholarly citation behavior. They generally work by running controlled prompt sets and recording AI outputs rather than having access to private model logs. www.clearscope.io

Best options for a university press

For an academic publisher, I would build a custom measurement layer too

Commercial tools are good for broad visibility, but a press has unusually structured assets: DOI records, author pages, book landing pages, abstracts, reviews, subject categories, and institutional reputation signals.

A strong setup would include:

1. A prompt monitoring database
Track recurring questions such as:

• “What are the best books on medieval trade?”
• “Which scholars write about [field]?”
• “Recommend recent academic research on [topic].”
• “What are authoritative sources on [subject]?”
• “Which university presses publish in [discipline]?”

Record:

• model used
• date
• prompt
• cited URLs
• cited domains
• cited authors
• whether your titles appeared
• whether competitors appeared

2. Domain-level citation benchmarking

Compare your domain against:

• other university presses
• JSTOR
• Project MUSE
• institutional repositories
• major commercial academic publishers
• Wikipedia
• government/NGO research sites

3. Scholarly metadata auditing

AI systems are more likely to retrieve content that has:

• clear abstracts
• structured metadata
• author information
• publication dates
• ISBN/DOI identifiers
• schema.org markup
• accessible HTML pages (not only PDFs)

My shortlist by budget

Enterprise university press:

• Profound + internal analytics database

Mid-sized press:

• Peec AI or Ahrefs + a custom prompt library

Small press / pilot project:

• Otterly.ai + manual quarterly benchmarking

One caution

Do not optimize only for “getting cited by ChatGPT.” A citation from an AI answer engine is a useful discovery signal, but it is not the same as academic citation impact. You still want Crossref citations, Google Scholar visibility, library discovery, indexing, and scholarly reputation signals tracked separately.

If you are evaluating this for a university press strategy team, I would also consider building a “AI discoverability dashboard” similar to an institutional repository analytics report: titles surfaced, domains cited, missed opportunities, and metadata fixes prioritized by discipline.

If your goal is specifically “which of our books/papers/domains are being cited by ChatGPT, Claude, Perplexity, Gemini, etc., and what should our press do about it?”, I’d separate this from conventional scholarly citation tracking.

My shortlist

1. Profound — probably the first product I'd demo

Profound is aimed much more directly at AI-search visibility than traditional bibliometrics. It can monitor queries and identify which sources/domains AI systems surface.

That's particularly useful for a press because you can frame the unit of analysis as:

Prompt → AI answer → cited source → publisher/domain → title/DOI → competing source

rather than merely asking whether Paper X has accumulated another conventional citation.

One important caveat: AI-answer citation statistics are inherently sampled estimates, not an authoritative census of everything ChatGPT or Gemini has ever cited. A recent independent comparison explicitly notes this measurement problem and recommends looking at methodology, sampling and overlap between providers. Nick Lafferty

2. Scrunch AI — strong alternative

I'd put Scrunch in the same evaluation bucket as Profound if you're interested in AI visibility at the domain/content level.

For a university press, I'd want a demo showing:

• Your press domain vs. competing university presses
• Which prompts cause your books/articles to appear
• Which individual URLs get cited
• Which competitors replace you
• Which AI engines cite you
• Citation/share-of-voice trends over time
• The actual text of the answer and citation
• Whether the cited URL is a canonical landing page, PDF, repository copy, etc.

That last distinction is especially important for scholarly publishing.

3. Scite — don't overlook this one

Scite is different, but potentially extremely valuable for a university press.

Scite currently indexes 317M+ articles and 1.6B+ Smart Citations, and its citation analysis distinguishes citations that support, contrast, or merely mention a work. scite.ai scite.ai

More interestingly for your use case, Scite now has integrations with ChatGPT, Claude and other AI tools, as well as an API. scite.ai scite.ai

It also explicitly works with publishers and says its publisher infrastructure can surface scholarly content through AI tools while protecting the publisher's version of record. scite.ai

So I'd use:

Profound/Scrunch → “Are AI systems finding and citing us?”

Scite → “How is the scholarly ecosystem citing and contextualizing our work?”

Those are complementary questions.

What I'd actually build for a university press

Rather than buying one generic “AI visibility” dashboard, I'd create a Press AI Discoverability Dashboard with five KPIs:

1. AI citation rate

% of tracked research questions where at least one press publication is cited.

2. Citation share

Your press vs. CUP/OUP/Chicago/Princeton/etc. for the same topic clusters.

3. Title-level AI visibility

Which individual books/articles/chapters receive citations from AI answers?

4. Domain-level visibility

press.university.edu vs. repositories, Google Books, JSTOR, author sites, Wikipedia, etc.

5. Citation pathway

When an AI cites a work, which URL does it cite?

That fifth metric could reveal some surprisingly actionable problems. For example:

AI knows the book → cites a Google Books page → never cites the press's own page.

That suggests a discoverability/structured-data/canonicalization problem, rather than a content-quality problem.

One particularly important distinction

Don't treat “AI mentions our book” and “AI cites our book” as the same metric.

I'd track at least:

• Mentioned by title
• Author mentioned
• Publisher/domain mentioned
• URL cited
• DOI cited
• Full bibliographic citation
• Citation appears in answer vs. source list
• Citation appears for branded vs. non-branded query
• Citation appears alongside competing publishers

This gives your editorial/marketing/metadata teams something they can actually act on.

And I would not replace conventional bibliometrics with AI citations. Altmetric, for example, is designed to track attention around research outputs across online sources and provides auditable mention information. Altmetric Help Center

My buying recommendation

If I were evaluating this for a university press today, I'd run a 30-day bake-off:

Profound + Scrunch + Scite, using the same ~200 prompts across your major disciplines.

Then score each on:

• AI engines covered
• Citation-level granularity
• URL/title/DOI identification
• Historical data
• Competitor benchmarking
• API/export capability
• Prompt customization
• Reproducibility of measurements
• Cost per tracked query
• Ability to connect results to your publication metadata

Profound/Scrunch will likely tell you more about AI-search visibility; Scite will give you substantially richer scholarly citation intelligence. The combination is much more useful to a university press than a conventional citation database alone.

If you tell me roughly how many titles/papers your press has and whether you're primarily books, journals, or both, I can also give you a concrete vendor comparison + recommended measurement architecture for a university press.

Those are complementary questions.

What I'd actually build for a university press

Rather than buying one generic “AI visibility” dashboard, I'd create a Press AI Discoverability Dashboard with five KPIs:

1. AI citation rate

% of tracked research questions where at least one press publication is cited.

2. Citation share

Your press vs. CUP/OUP/Chicago/Princeton/etc. for the same topic clusters.

3. Title-level AI visibility

Which individual books/articles/chapters receive citations from AI answers?

4. Domain-level visibility

press.university.edu vs. repositories, Google Books, JSTOR, author sites, Wikipedia, etc.

5. Citation pathway

When an AI cites a work, which URL does it cite?

That fifth metric could reveal some surprisingly actionable problems. For example:

AI knows the book → cites a Google Books page → never cites the press's own page.

That suggests a discoverability/structured-data/canonicalization problem, rather than a content-quality problem.

One particularly important distinction

Don't treat “AI mentions our book” and “AI cites our book” as the same metric.

I'd track at least:

• Mentioned by title
• Author mentioned
• Publisher/domain mentioned
• URL cited
• DOI cited
• Full bibliographic citation
• Citation appears in answer vs. source list
• Citation appears for branded vs. non-branded query
• Citation appears alongside competing publishers

This gives your editorial/marketing/metadata teams something they can actually act on.

And I would not replace conventional bibliometrics with AI citations. Altmetric, for example, is designed to track attention around research outputs across online sources and provides auditable mention information.

My buying recommendation

If I were evaluating this for a university press today, I'd run a 30-day bake-off:

Yes. For a university press, I’d treat this as an AI citation intelligence problem rather than ordinary academic citation analysis. Traditional tools such as Scite track how scholarly papers cite one another; what you want is closer to: “When people ask AI systems about topics covered by our books/papers, which URLs, papers, publishers, repositories, and domains do those systems actually cite?”

The strongest options I’d shortlist right now are:

What I'd choose for a university press

Profound + Scite would be my starting combination.

Profound answers the new question:

“Which sources does AI use when answering questions in our subject areas?”

Scite answers the established scholarly question:

“Who cites our papers/books and in what context?”

Then I'd maintain a controlled set of perhaps 100–500 prompts representing the press's actual intellectual territory. For example:

• “What are the major interpretations of [topic]?”
• “Best books on [topic]”
• “Who are the leading scholars on [topic]?”
• “What does recent scholarship say about [topic]?”
• “[Author]'s contribution to [field]”
• “What sources should I read to understand [concept]?”
• “Compare the major theories of [subject].”

Run those repeatedly rather than treating a single ChatGPT answer as meaningful data. AI citation results are inherently variable, and current research emphasizes repeated sampling rather than one-off tests. Atelier La Refonte OpenAttribution

The metric I'd actually build

For a press, “Does our domain get cited?” is too crude.

I'd track:

1. Paper/book citation rate — percentage of target prompts producing a citation to a specific work.
2. Domain citation share — percentage of all cited sources belonging to your press/domain.
3. Topic visibility — citation share by subject area.
4. Competitor share — which university presses, journals, repositories and aggregators are beating you.
5. Third-party dependency — whether AI cites your work directly or reaches it through Google Books, JSTOR, Wikipedia, institutional repositories, reviews, etc.
6. Engine divergence — ChatGPT vs. Perplexity vs. Gemini vs. Google AI.
7. Citation persistence — whether a work remains cited over months rather than appearing once.
8. Citation gaps — highly relevant questions where competing scholarship is cited but your work isn't.

That last category is particularly actionable. If AI repeatedly answers “What are the leading books on X?” with five competitors and never your title, you've identified a discoverability problem that can potentially be addressed through metadata, structured pages, reviews, repositories, author pages, links and other legitimate distribution channels.

One important caveat: don't interpret AI citations as equivalent to scholarly citations. AI engines can repeatedly reproduce an influential web source without that source being the academically strongest one. Recent research also finds substantial differences between engines in which sources they select. journals.sagepub.com getpassionfruit.com

If your goal is specifically “build an AI discoverability dashboard for a university press,” I can also lay out a recommended stack—including which APIs/data sources to use, a database schema for storing every AI citation, and the KPIs I'd put on the dashboard.

For a university press, you probably want something different from ordinary SEO software. The goal is not just “does our website rank?” but:

• Which papers, books, chapters, authors, and domains are being surfaced by ChatGPT, Claude, Gemini, Perplexity, etc.
• Which queries in your subject areas trigger citations
• Whether your publications are being cited versus competitors
• What metadata/content changes improve AI discoverability

The market is still immature: there is no universal “AI citation analytics” feed from ChatGPT, so most tools work by running controlled prompt sets and recording the answers and cited sources. CiteFlow Linkeddit

Best tools to evaluate

What I would prioritize for a university press

1. Build a “scholarly AI visibility benchmark”

Create a fixed monthly prompt set such as:

• “What are the leading books on medieval trade networks?”
• “What are the most cited works on climate migration?”
• “Recommend introductory texts on constitutional theory.”
• “Who are the leading scholars in [field]?”
• “What sources should I read about [topic]?”

Track:

• Did the answer cite your press?
• Which title/page/DOI was cited?
• Which competing presses appeared?
• Which domains were repeatedly cited?

Most AI visibility platforms are built around this prompt-monitoring model. Sona

2. Add scholarly-specific data sources

General AI visibility tools may miss academic signals. Combine them with:

• Crossref — DOI metadata, publisher records
• OpenAlex — citation and scholarly graph data
• Google Scholar — scholarly discovery benchmarking
• ORCID — author identity/disambiguation

The key is connecting AI citations back to specific scholarly objects:

• ISBN
• DOI
• author ID
• chapter URL
• repository record
• landing page

3. Measure “AI discoverability” rather than only citations

Useful KPIs:

• AI citation share: percentage of relevant prompts where your titles appear
• Source authority share: how often your domains are used versus competitors
• Title discoverability: which books appear without being explicitly requested
• Author discoverability: whether your scholars are recognized
• Metadata conversion: whether improved abstracts, summaries, and structured data increase appearances

My shortlist for a university press pilot

Small/medium press

1. Otterly.AI or Peec AI
2. Spreadsheet + OpenAlex/Crossref enrichment
3. Monthly human review of important disciplines

Large university press

1. Profound pilot
2. Custom prompt corpus by subject area
3. Integrate with your CMS, analytics, DOI metadata, and catalog systems

One thing I would not rely on

Traditional SEO rank trackers alone. A book can rank poorly in Google but still be repeatedly used as a source by AI systems, and AI citation patterns differ substantially by engine. Distribution.studio

If you are a university press with thousands of titles, I would probably design a lightweight “AI citation observatory” rather than buy a generic tool alone: prompt library + AI visibility tracker + scholarly metadata graph + monthly editorial actions.

Yes. For a university press, I’d look beyond ordinary “AI brand visibility” tools and prioritize URL-level citation tracking, prompt-level results, source/domain analysis, and historical trends.

My shortlist

1. scrunch.com — best overall fit

• Tracks exactly which domains and URLs AI systems cite.
• Lets you drill from domain → individual URL → prompts that cited it.
• Separates your own citations from competitor and third-party citations.
• Supports trend analysis and filtering by topic, AI platform, prompt, etc. Scrunch Scrunch
• This is particularly useful for a press because you could discover, for example, “AI answers about medieval history repeatedly cite JSTOR and three university sites, but almost never our monographs.”
• tryprofound.com — best for sophisticated research/strategy

• Strong citation intelligence across ChatGPT, Perplexity, Google AI and other answer engines.
• Categorizes sources as owned, competitor, earned media, institutions, etc.
• Specifically surfaces top-cited publishers and authors, which is unusually relevant to an academic press.
• Can identify citation gaps and feed them into content/outreach strategy. Profound
• ahrefs.com — best if you already use Ahrefs

• Tracks citations and mentions across ChatGPT, Perplexity, Gemini, Copilot and Google's AI surfaces.
• Has a very large prompt corpus and reports cited domains and pages.
• Its current index covers hundreds of millions of prompts, making it useful for benchmarking an entire press against other publishers/institutions, rather than monitoring only a small hand-built prompt set. Ahrefs Ahrefs Help Center
• The downside is cost/complexity relative to tools designed specifically around citation intelligence.
• semrush.com — good general-purpose option

• Useful if your press already uses Semrush for SEO.
• Tracks cited domains/URLs for monitored prompts and provides competitive AI-visibility analysis. Payline Data

One important distinction for your use case

I'd not make “How often does ChatGPT mention our press?” the primary KPI.

For an academic publisher, I'd build a dataset around:

That distinction matters because a page being retrieved isn't necessarily the same as being cited. For example, Ahrefs explicitly distinguishes pages AI found during retrieval from pages it actually cited. Ahrefs Help Center

What I'd actually do for a university press

I'd probably pilot Scrunch + your own bibliographic data.

Take a representative corpus of, say, 500–2,000 titles/articles, map each ISBN/DOI to its canonical URL, then construct prompts around your strongest subject areas:

“What are the best scholarly books on X?”
“What are the major interpretations of X?”
“Who are the leading scholars on X?”
“What sources should I read to understand X?”
“What does recent scholarship say about X?”

Then measure which specific books, papers, authors, university domains, repositories and competing presses appear in the citations.

The really valuable output isn't simply “our AI visibility is 14%.” It's something like:

Topic: Reconstruction politics
1,000 tracked AI responses → 4,800 citations
University repositories: 18%
JSTOR: 14%
Wikipedia: 11%
Oxford: 7%
Cambridge: 6%
Your Press: 1.2%

AI repeatedly cites five third-party articles discussing your books, but rarely cites the books' own landing pages.

That is actionable publishing intelligence. It tells your marketing/editorial/web teams where discoverability is breaking down.

There's also emerging academic research specifically analyzing what gets cited by ChatGPT, so I'd treat this as a genuinely interesting scholarly-communications measurement problem—not merely an SEO problem. Taylor & Francis Online

If you tell me roughly how many titles your press has, whether you have DOIs, and whether you want to monitor ChatGPT only or ChatGPT + Gemini + Perplexity + Google, I can recommend a specific stack and methodology (including likely costs and how I'd structure the database/dashboard).

For a university press, you probably want a tool that answers “Which of our books, chapters, papers, authors, and domains are being surfaced or cited by ChatGPT-like systems?” rather than traditional citation tracking. This is an emerging category often called AI visibility tracking, LLM citation monitoring, or generative engine optimization (GEO). Ansvisor

A practical shortlist:

What I would prioritize for a university press

1. Page-level citation tracking

You want reports like:

• “ChatGPT answered What are the best books on medieval history? and cited these domains.”
• “Your book landing page appeared 12 times this month.”
• “Competitor presses are being cited instead for these topics.”

Many AI visibility tools focus on extracting the cited URLs behind answers, not just whether a brand name appeared. Honeyb

2. A controlled prompt library

Create a recurring benchmark set:

Subject discovery

• “What are the leading books on [field]?”
• “Who are the major scholars in [field]?”
• “What should a graduate student read about [topic]?”

Book discovery

• “Recommend recent academic books about [topic].”
• “Which university presses publish on [topic]?”

Author discovery

• “Who are the experts on [topic]?”

Run these weekly/monthly across:

• ChatGPT
• Claude
• Gemini
• Perplexity

The important metric is not simply “mentioned/not mentioned,” but:

• cited URL
• citation frequency
• citation position
• competing sources cited
• whether the citation points to your canonical page, repository copy, metadata page, or a third-party page

3. Combine AI monitoring with scholarly metadata tools

AI visibility is only one layer. For academic publishing, pair it with traditional scholarly intelligence:

• Scite — useful for understanding how papers are discussed and supported/contradicted in later literature.
• Web of Science / Scopus — useful for institutional citation reporting.
• Crossref metadata quality — important because AI systems often depend on clean, structured publication information.

A recommended stack by budget

Lean pilot ($0–500/month)

• Otterly.AI or Peec AI for LLM visibility checks
• Google Search Console + analytics
• A spreadsheet of 100–300 benchmark prompts

Mid-size university press

• Semrush or Ahrefs for broader discoverability
• Dedicated LLM visibility tracker
• Internal dashboard combining: - AI citations
• organic search traffic
• DOI resolution
• downloads
• external citations

Large academic publisher

• Profound or similar enterprise platform
• Custom prompt monitoring
• API-based reporting by: - imprint
• discipline
• author
• title
• domain

One caution

AI citation data is still noisy. These tools generally simulate representative prompts and monitor outputs rather than having access to every user conversation with ChatGPT or other systems. CiteFlow So I would treat the metrics as discoverability signals, not as a replacement for COUNTER usage reports, citations, or sales data.

For a university press, the highest-value first project would probably be a “100 priority discovery questions” benchmark across your strongest subject areas, then measure which titles and domains AI systems choose as sources.

For a university press, you’re looking for something closer to AI visibility / generative engine optimization (GEO) monitoring for scholarly content than a traditional citation tracker. The ecosystem is still immature: AI systems do not expose a universal “citation index,” and different chat systems retrieve and cite different sources. OpenAttribution SSRN A good stack usually combines AI-answer monitoring, scholarly metadata, and web analytics.

Best options to evaluate

What I would build for a university press

A practical monitoring system would track:

1. AI answer citation share

Run a fixed monthly query set such as:

• “What are the best books on [field]?”
• “Who are leading scholars on [topic]?”
• “What are authoritative sources on [subject]?”
• “Recommend textbooks on [course topic]”

Track:

• Was your press mentioned?
• Was a specific book cited?
• Was the book landing page linked?
• Which competing presses appeared?
• Which domains did the AI system cite?

2. Domain-level signals

Measure:

• Your press website
• Book pages
• Author pages
• DOI landing pages
• Institutional repository links
• Wikipedia/Wikidata presence
• Library catalog records

AI systems often need strong, machine-readable signals. Good metadata matters as much as marketing copy.

3. Scholarly metadata health

Audit:

• DOI completeness
• ISBN metadata
• ORCID links
• Crossref records
• OpenGraph/schema.org markup
• Abstract availability
• Table of contents pages
• Author biographies

4. Citation-gap analysis

Compare:

• Your books/papers cited by AI
• Similar works from other university presses
• Frequently cited domains in your disciplines

The goal is not simply “get AI to cite us”; it is to make authoritative information about your publications easy for retrieval systems to identify.

My shortlist for a university press

If you want a vendor solution:

1. tryprofound.com or peec.ai for AI visibility monitoring
2. ahrefs.com for SEO/content authority
3. crossref.org + openalex.org for scholarly metadata tracking

If you have data/analytics staff, I would build an internal dashboard that samples ChatGPT, Claude, Gemini, and Perplexity answers against a controlled query library and stores cited URLs/domains over time. That approach aligns better with publishing goals because you can measure which titles, subjects, and metadata improvements change discoverability. OpenAttribution

A useful first project would be a “Top 500 AI discovery queries for our catalog” benchmark: 50–100 queries per discipline, run monthly, with citation-share rankings against peer university presses.

If your goal is specifically “which academic papers and domains are being cited by ChatGPT/Claude/Perplexity, and how can our university press improve our visibility?”, I’d separate the market into AI-visibility monitoring and scholarly citation intelligence. No single tool perfectly measures both.

My shortlist

1. scite.ai — best foundation for a university press

• Tracks citations at the paper level, including citation context and whether later research supports, contrasts with, or merely mentions a work.
• Its database currently covers 300M+ articles and 1.6B+ Smart Citations, with 30+ publisher partnerships. scite.ai scite.ai
• Crucially for a press, it has APIs, dashboards, CSV exports, publisher integrations, and institutional/enterprise offerings. scite.ai scite.ai
• It also now connects directly with AI assistants including ChatGPT, Claude, Gemini and Copilot, which makes it particularly relevant to the ecosystem you're trying to understand. scite.ai
• Best use: Build a corpus of your books/articles/DOIs and monitor conventional scholarly citations and the research context around them.
• AI-search visibility platforms — best for the actual “what does ChatGPT cite?” question

This is a newer category. Tools such as Promptwatch and other “AI visibility” platforms run large numbers of prompts against AI systems and report which domains/URLs appear in answers. This is much closer to your desired metric than traditional bibliometrics. For example, recent reporting based on Promptwatch data illustrates that citation patterns in ChatGPT can change dramatically after model/search updates. Business Insider

Best use: Track questions relevant to your subject areas—e.g., “What are the major works on medieval trade?”—and measure whether your press's books and articles are actually cited.

3. A custom monitoring layer — probably the best long-term solution for a press

For a university press, I'd seriously consider building a small internal system rather than relying entirely on one SaaS dashboard:

Your catalog → DOI/ISBN metadata → prompt library → AI responses → cited URLs/DOIs/domains → longitudinal dashboard

You could run a standardized set of prompts across ChatGPT, Claude, Gemini, Perplexity, etc., on a monthly or weekly basis and record:

• Your titles cited
• Competitor titles cited
• Journals/publishers/domains cited
• Frequency of citation
• Position/order of citation
• Whether the AI links to the publisher's page, DOI, Google Books, JSTOR, repository, etc.
• Which queries produce zero citations to your content
• Changes following metadata, SEO, open-access, or discoverability interventions

The important distinction

I'd not equate scholarly citations with AI citations.

A book can have thousands of conventional citations while being invisible to ChatGPT. Conversely, an accessible publisher page, institutional repository copy, review, Wikipedia entry, or other web source can make a scholarly work much more likely to surface in an AI answer.

So I'd establish two KPIs:

What I'd buy/build

For a university press, my recommendation would be:

Scite + an AI-visibility monitoring platform + your own prompt/citation dataset.

Scite is particularly attractive because it already has a programmatic API and publisher-facing infrastructure, rather than being just a researcher-facing search engine. scite.ai scite.ai

And I'd be cautious about vendors promising a definitive "AI citation count." AI answers aren't stable bibliographic databases: retrieval, ranking, model versions, location, search settings, and even prompt wording can change the sources an AI cites. Recent changes in ChatGPT's citation behavior illustrate exactly why you want longitudinal measurements rather than a single score. Business Insider

If you tell me roughly how large your press's catalog is (e.g. 2,000 vs. 20,000 titles) and which AI systems matter most, I can give you a concrete vendor-by-vendor comparison (Scite, Promptwatch, Otterly, Profound, Peec, etc.) plus a recommended architecture and approximate budget for a university press.

If your goal is “which academic papers, books, authors, journals, and domains are actually being cited by ChatGPT/Perplexity/etc., so our university press can improve discoverability,” I’d prioritize tools that expose source-level citation data, not merely “brand mentions.”

My shortlist

1. Profound — my first choice

Profound is particularly interesting for your use case because its citation analysis isn't limited to whether your domain appears. It lets you inspect the sources AI systems use, categorize sources, compare citation share, and identify publishers/authors that repeatedly influence AI answers. Profound Profound

For a university press, I'd build prompt sets around things like:

• "best books on medieval Islamic history"
• "scholarly sources for Caribbean political history"
• "key works on environmental humanities"
• "recommended academic books about X"
• "historiography of X"
• "most influential scholars/books on X"
• course-oriented questions in your strongest disciplines

Then examine every cited URL, not just whether your press was mentioned.

That can answer questions such as:

“When someone asks an AI about Victorian literature, which university presses, journals, repositories, and individual book pages does it cite?”

Profound also reports citation share by topic, platform, and prompt, and can identify third-party publishers/authors that are disproportionately cited. Profound

2. OtterlyAI — excellent for citation surveillance

otterly.ai is probably the one I'd test alongside Profound.

Its citation tracking specifically records which URLs AI search experiences cite, including competitor URLs, and lets you track citation changes over time. It currently covers ChatGPT, Perplexity, Google AI Overviews/AI Mode, Gemini, Copilot and Claude. Otterly Otterly

A particularly useful feature for a press is its ability to look at the domain citation landscape, rather than just your own visibility. Its reports can expose the sources where competitors appear but you don't. Otterly

It also has an interesting complementary capability: AI crawler/agent analytics. You can feed it server logs and see which AI agents are visiting which pages on your site. That's different from citation tracking, but combining the two could tell you:

AI crawls this book page → but doesn't cite it → whereas this competitor's page gets cited repeatedly.

That's a much more actionable signal than a generic AI visibility score. Otterly Otterly

3. Semrush — strongest if SEO is already part of your workflow

semrush.com now has AI Visibility functionality covering things such as cited pages, citations, mentions, and visibility across AI platforms. It also lets you combine those measurements with traditional search data. Semrush Semrush

For a press, that's useful because you don't want to optimize solely for AI. You can ask:

Google SEO performance + AI citation performance + referral traffic = which book/product pages deserve investment?

Semrush also exposes citation position—for example, whether your domain tends to be the first, second, etc. cited source for tracked prompts. Semrush

What I'd actually implement

I wouldn't start by trying to track every academic paper on the internet. That's an enormous and somewhat misleading dataset.

I'd create a “scholarly AI discoverability” benchmark for your press:

1. Pick 100–300 high-value prompts across your strongest subject areas.
2. Run them across ChatGPT, Perplexity, Gemini, Claude, and Google AI experiences.
3. Capture every cited URL and domain.
4. Classify each citation: - university press
5. commercial publisher
6. journal
7. repository
8. university
9. library
10. Wikipedia
11. news/media
12. individual scholar
13. other
14. Calculate: - citation share by publisher
15. citation share by domain
16. citations per individual book/page
17. citation position
18. co-citation relationships
19. topics where competitors are cited but your press isn't
20. Connect that to your own metadata: ISBN, title, author, discipline, keywords, abstracts, reviews, references, DOI, etc.

The really valuable output isn't “our AI visibility score is 17.”

It's something like:

For questions about 20th-century African political history, AI systems cite Cambridge 31%, Oxford 18%, JSTOR 14%, Wikipedia 11%, and our press 2%. The competitor books being cited have accessible abstracts, structured metadata, extensive external references, and pages that are independently cited by university websites.

That gives your editorial/marketing/web teams something they can actually act on.

One important caveat

“AI citation tracking” isn't the same thing as measuring all citations made by ChatGPT. These products generally sample defined prompts/queries and observe the resulting answers. Profound, for example, emphasizes running tracked prompts repeatedly because AI answers aren't deterministic. Profound

So I'd treat the resulting numbers as a panel/benchmark of AI discoverability, rather than a census of every ChatGPT conversation.

My recommendation: start with Profound + OtterlyAI for a 4–6 week pilot. Profound looks stronger for the research/competitive intelligence question, while OtterlyAI looks especially useful for ongoing URL/domain monitoring and your site's AI-crawler data. Then keep Semrush if your existing SEO team already lives in it.

If you tell me roughly how many titles your press has and which disciplines you publish in, I can also design a concrete AI-citation monitoring framework (prompts, KPIs, taxonomy, dashboard, and monthly reporting) specifically for a university press.

semrush.com now has AI Visibility functionality covering things such as cited pages, citations, mentions, and visibility across AI platforms. It also lets you combine those measurements with traditional search data. semrush.com semrush.com

For a press, that's useful because you don't want to optimize solely for AI. You can ask:

Google SEO performance + AI citation performance + referral traffic = which book/product pages deserve investment?

Semrush also exposes citation position—for example, whether your domain tends to be the first, second, etc. cited source for tracked prompts.

What I'd actually implement

I wouldn't start by trying to track every academic paper on the internet. That's an enormous and somewhat misleading dataset.

If your goal is specifically “which papers, books, and publisher domains are actually being cited by ChatGPT/Perplexity/Gemini, on which questions, and how often?”, I’d treat this as AI citation intelligence, not traditional scholarly citation tracking.

My shortlist

What I'd actually buy

For a university press, I'd probably build a two-layer stack:

1. AI visibility layer — GetIntel or Profound
Track a controlled set of questions such as:

• “What are the best books on X?”
• “What does the scholarship say about Y?”
• “Who are the leading scholars on Z?”
• “What are the foundational works on X?”
• “What should I read to understand Y?”
• Highly specific questions corresponding to subjects where your press has substantial lists.

Then record:

query → AI engine → cited URL → cited domain → individual paper/book → author → publisher → citation position → date → competitor sources.

That lets you answer something much more useful than “is our website visible?” You can ask “Which of our books are entering the AI research pathway, and which competing publishers are winning the questions we care about?”

2. Scholarly-citation layer — Scite + Scopus/Web of Science/Dimensions
Connect AI citations with ordinary scholarly impact. A particularly interesting metric would be:

Human citations → AI citations

For example, identify papers that have strong scholarly citation performance but are rarely surfaced by AI, versus papers with modest conventional impact that AI systems disproportionately recommend.

One important caveat

Don't treat an AI citation count like a Scopus citation count. AI answers are stochastic and prompt-dependent. Recent research and vendor analyses show substantial differences between engines and even between runs, so you need a fixed prompt panel and repeated measurements rather than a single “AI citation score.” GetIntel Nick Lafferty

This is particularly important for a press because domain-level visibility can hide the individual-title problem. Your press might be cited frequently because an AI system likes your general reference pages while almost never citing the actual scholarly books you want discovered.

There is also evidence that the engines have quite different citation behavior: one recent analysis of 17.2 million AI citations found meaningful differences among ChatGPT, Perplexity, Gemini and Claude, suggesting that optimizing for one engine isn't necessarily optimizing for the others. Yext

What I would measure for a press

I'd make the dashboard revolve around these six KPIs:

1. AI citation rate — % of target prompts citing one of your titles/domains.
2. Title-level citation rate — which individual books/papers are actually surfaced.
3. Domain share of citations — your press vs. university presses vs. commercial publishers.
4. Citation position/prominence — first/foundational source vs. incidental citation.
5. Engine overlap — titles cited by ChatGPT and Perplexity and Gemini.
6. AI-vs-human citation gap — scholarly impact compared with AI discoverability.

That last one could become genuinely valuable editorial intelligence: “Our books are highly cited by scholars but invisible to AI research assistants.” That's actionable for metadata, landing pages, open-access copies, structured abstracts, author pages, reviews, Wikipedia/Wikidata presence, institutional repositories, and other discoverability work.

If you tell me roughly how many titles your press publishes (e.g. 500 vs. 20,000) and whether you need an API/data export, I can narrow this to the 2–3 tools I'd shortlist and give you an implementation architecture and estimated cost.

For example, identify papers that have strong scholarly citation performance but are rarely surfaced by AI, versus papers with modest conventional impact that AI systems disproportionately recommend.

One important caveat

Don't treat an AI citation count like a Scopus citation count. AI answers are stochastic and prompt-dependent. Recent research and vendor analyses show substantial differences between engines and even between runs, so you need a fixed prompt panel and repeated measurements rather than a single “AI citation score.” GetIntel nicklafferty.com

This is particularly important for a press because domain-level visibility can hide the individual-title problem. Your press might be cited frequently because an AI system likes your general reference pages while almost never citing the actual scholarly books you want discovered.

There is also evidence that the engines have quite different citation behavior: one recent analysis of 17.2 million AI citations found meaningful differences among ChatGPT, Perplexity, Gemini and Claude, suggesting that optimizing for one engine isn't necessarily optimizing for the others.

What I would measure for a press

Yes. For a university press, I’d separate this into two jobs:

1. AI citation/visibility monitoring: which papers, books, authors, and domains are actually being cited in ChatGPT, Gemini, Perplexity, etc.
2. Scholarly citation + referral measurement: whether those AI citations translate into visits, downloads, and conventional academic citations.

My shortlist

What I'd buy for a university press

I'd probably start with Profound or Peec AI + Ahrefs, rather than expecting one product to do everything.

The important distinction is that “AI visibility” isn't necessarily “AI citation.” A model can mention your press without linking to it, while a paper from your catalog can be cited as a source without the press itself being mentioned. Current AI-traffic tools primarily measure the click after an AI citation, whereas citation-monitoring tools sample the answers themselves. Ahrefs Nick Lafferty

For a press, I'd want a dashboard roughly like:

Query: “best scholarship on the history of early American constitutionalism”
Model: ChatGPT
Run date: Aug. 29, 2026
Cited sources:

• yourpress.edu/book/…
• jstor.org/…
• cambridge.org/…
• anotherpress.com/…

Your share of citations: 14%
Competitor share: 23%
Change: +6 pp over 90 days

Then aggregate that by book, article, author, subject, URL, domain, model, query type, and competitor.

One thing I would not overlook

For scholarly publishing, maintain your conventional citation infrastructure alongside AI monitoring. crossref.org can give you the actual scholarly works citing your publications, and Crossref supports citation counts plus citation notifications for participating members. CrossRef www.crossref.org

That gives you a powerful three-layer measurement:

Academic citations → AI citations → AI-driven visits/downloads

The third layer is particularly important because AI referrals can be difficult to capture completely in analytics; some AI visits are classified as direct traffic. Ahrefs Ahrefs

My ranking for your specific use case

If your primary goal is improving discoverability of scholarly content:

1. Profound — best candidate for a serious institutional program
2. Peec AI — strong alternative for AI-search intelligence
3. GetIntel — worth testing specifically for source/citation-level analysis
4. Otterly.AI — good lower-complexity option
5. Ahrefs — essential companion for measuring actual AI-referred traffic
6. Semrush — particularly attractive if your marketing/SEO team already uses it

I'd also run a small bake-off before purchasing: give 3 vendors the same ~50–100 scholarly discovery questions, then compare whether they agree on which URLs were cited. Citation tracking is inherently sample-based, and recent comparisons emphasize that a single model run isn't a reliable measurement of citation share. Nick Lafferty

If you tell me roughly how many titles/DOIs your press has (e.g. 500 vs. 10,000+) and whether you care more about books, journal articles, or both, I can narrow this to the best 2–3 options and compare pricing, supported AI engines, API access, historical data, and citation-level reporting.

Query: “best scholarship on the history of early American constitutionalism”
Model: ChatGPT
Run date: Aug. 29, 2026
Cited sources:

• yourpress.edu/book/…
• jstor.org/…
• cambridge.org/…
• anotherpress.com/…

Your share of citations: 14%
Competitor share: 23%
Change: +6 pp over 90 days

Then aggregate that by book, article, author, subject, URL, domain, model, query type, and competitor.

One thing I would not overlook

For scholarly publishing, maintain your conventional citation infrastructure alongside AI monitoring. crossref.org can give you the actual scholarly works citing your publications, and Crossref supports citation counts plus citation notifications for participating members. production.crossref.org crossref.org

That gives you a powerful three-layer measurement:

Academic citations → AI citations → AI-driven visits/downloads

The third layer is particularly important because AI referrals can be difficult to capture completely in analytics; some AI visits are classified as direct traffic.

If your goal is specifically “Which academic papers/domains are being cited or surfaced by ChatGPT and other chat-based AI?”, I’d separate this into AI-answer monitoring and traditional scholarly/online impact tracking. There isn’t yet a single mature equivalent of Scopus that comprehensively records every citation made inside every LLM response.

My shortlist

1. Scite is probably the best first purchase

scite.ai is particularly useful for a press because it goes beyond “paper X has 200 citations.” Its Smart Citations identify the context and apparent function of citations—supporting, contrasting, or merely mentioning a work. DOI MIT Press Direct

That makes it useful for questions such as:

• Which of our titles are becoming influential reference points?
• Who is citing them?
• How are they being cited?
• Which competing books/papers are repeatedly cited alongside ours?
• What topics have a strong citation footprint but weak visibility in our catalog?

I'd use Scite as the scholarly citation intelligence layer, rather than assuming it is an LLM-citation tracker.

2. Altmetric is the complementary layer

altmetric.com is better for understanding attention outside the scholarly citation graph—news, policy documents, social platforms, blogs, and other online sources. Its publisher-oriented tools can track outputs by DOI, ISBN and other identifiers and can aggregate different versions of a research output. Altmetric Altmetric Altmetric

For a university press, the ISBN support is particularly interesting, since your portfolio isn't necessarily just journal articles.

3. But for your actual question, I'd build an AI-answer monitoring layer

This is the important distinction.

You really want a dataset like:

Prompt → AI system → response → cited sources/domains → frequency → position → citation type → date/model

For example:

You can then calculate an AI Discoverability Score for every title/domain:

AI citation share = appearances of your content / appearances of all tracked sources

and track it over time.

That's much more actionable for marketing, metadata, discoverability, and editorial strategy than conventional citation counts.

What I'd recommend for a university press

I'd build a three-layer stack:

1. Scite → scholarly citation context and competitor/reference intelligence.
2. Altmetric → broader online attention and dissemination.
3. LLM monitoring → systematically query ChatGPT, Claude, Gemini, Perplexity, etc., and record which sources they cite.

The third layer is the piece I'd treat as a custom analytics product, rather than expecting Scopus/Scite/Altmetric to answer it completely.

There's also good reason to do this now: recent bibliometric research is explicitly beginning to study LLM influence on scholarly communication and the relationship between LLM-generated assessments and conventional citation/attention metrics. Springer Link Springer Link

One important caveat

Don't equate “cited by an AI” with “cited academically.” An LLM may retrieve a publisher landing page, Wikipedia, a review, a repository copy, or a secondary source and then attribute information to it. Conversely, an AI can know about a work without displaying a formal citation.

So I'd actually monitor three events:

• Mention — the work/title/author appears in the answer.
• Citation — the AI explicitly provides the source.
• Source selection — the AI chooses your domain as the underlying source.

For a press, that third metric could be especially valuable: “When an AI needs information about this subject, does it choose our site?”

If you tell me roughly how large your catalog is (e.g., 5,000 vs. 100,000 titles) and whether you're interested primarily in books, journal articles, or both, I can recommend a concrete vendor/tool stack and an implementation architecture, including what I'd measure monthly.

Prompt → AI system → response → cited sources/domains → frequency → position → citation type → date/model

For example:

You can then calculate an AI Discoverability Score for every title/domain:

AI citation share = appearances of your content / appearances of all tracked sources

and track it over time.

That's much more actionable for marketing, metadata, discoverability, and editorial strategy than conventional citation counts.

What I'd recommend for a university press

I'd build a three-layer stack:

1. Scite → scholarly citation context and competitor/reference intelligence.
2. Altmetric → broader online attention and dissemination.
3. LLM monitoring → systematically query ChatGPT, Claude, Gemini, Perplexity, etc., and record which sources they cite.

The third layer is the piece I'd treat as a custom analytics product, rather than expecting Scopus/Scite/Altmetric to answer it completely.

If your goal is specifically “Which papers/domains are chat-based AI systems actually surfacing, and how can our university press increase the chances that our scholarship gets surfaced?”, I’d separate the market into AI-citation observability and traditional scholarly citation analytics.

My top pick: Scite

Scite is the closest fit I found.

Its new Publisher Gateway is explicitly designed for publishers to measure AI demand: it reports which articles are being requested by AI systems, the AI platform making the request, and whether content was accessible or protected. It says it can surface content into ChatGPT, Claude and other AI tools while keeping paywalled full text protected. scite.ai scite.ai

scite.ai

That makes it considerably more relevant to your question than ordinary citation databases. In particular, I'd ask Scite for a publisher-level dashboard covering:

• AI requests by DOI/title
• AI platform/source
• requests that result in a citation versus merely retrieval
• subject/domain/topic
• open-access versus paywalled content
• referring/citing context
• trends over time
• comparison of your catalog against comparable university presses

Scite also has an API and integrations with AI systems, so it could potentially become part of an internal discoverability pipeline rather than just another dashboard. scite.ai

The important caveat

There isn't, as far as I can establish, a universal equivalent of Google Analytics for “citations generated by ChatGPT/Claude/Gemini.”

That's because AI systems don't expose a complete, standardized log of every source they considered or cited. So I'd be suspicious of any vendor promising a comprehensive count of all ChatGPT citations.

Scite's publisher offering is interesting precisely because it is measuring requests through its own AI/research infrastructure, rather than claiming omniscient access to every model's internal retrieval/citation behavior. scite.ai

So I'd define your KPI as AI discoverability/demand signals, rather than “number of times ChatGPT cited us.”

Pair it with OpenAlex

OpenAlex is the inexpensive/open infrastructure I'd use underneath the commercial product.

OpenAlex gives you work-level citation counts, incoming/outgoing citation relationships, institutions, topics, publishers, open-access status, etc., and has an API suitable for building your own analytics. OpenAlex Help Center OpenAlex

developers.openalex.org

For a university press, you could build a dataset like:

DOI → scholarly citations → topic → institution → OA status → AI-demand signal → AI-citation signal

Then identify things such as:

“Our books/papers on medieval political thought are heavily cited academically but rarely surfaced in AI answers.”

or:

“These five open-access articles are disproportionately appearing in AI-mediated research, suggesting that accessibility + metadata + topic authority are working.”

That is much more actionable for an editorial/marketing team.

How I'd rank the options

The key distinction is that Scite is the one I'd investigate first for the AI-specific question. Its database also goes beyond simple citation counts by classifying citation statements as supporting, contrasting, or merely mentioning a finding. scite.ai Research Solutions

What I'd actually buy/build

For a university press, I'd probably avoid buying five overlapping dashboards.

I'd build a small AI Discoverability Observatory around:

1. Scite Publisher Gateway — AI demand and AI-mediated discovery.
2. OpenAlex — scholarly citation/network baseline.
3. Your own catalog/DOI metadata — authoritative inventory.
4. Web analytics/server logs — actual traffic from AI/search referrals where identifiable.
5. Periodic controlled queries to ChatGPT, Claude, Gemini, Perplexity, etc. — a longitudinal “AI share of voice” benchmark.

The last piece is particularly important. Create perhaps 100–500 representative research questions in your press's subject areas and periodically run the same questions through the major AI systems. Record:

• Was one of your publications mentioned?
• Was it cited?
• Was your press/domain linked?
• Which competing publications were cited?
• Was the cited version open access?
• Was the citation to the publisher's version of record, a repository, Google Books, JSTOR, etc.?
• Did the model accurately describe the work?
• What metadata did the model appear to rely upon?

That gives you a much more meaningful metric than raw web traffic: AI share of scholarly voice.

And it gives your press concrete interventions to test—better abstracts, structured metadata, DOI resolution, repository copies, chapter-level metadata, machine-readable full text, author pages, internal linking, and OA availability—rather than simply trying to “do SEO for ChatGPT.”

If you're evaluating this for procurement, Scite is the first vendor I'd request a demo from, specifically asking them to show you actual publisher-level AI request data for a university-press catalog, rather than a generic researcher demo. Their current publisher offering is explicitly positioned around that use case. scite.ai

scite.ai

developers.openalex.org

For a university press, you could build a dataset like:

DOI → scholarly citations → topic → institution → OA status → AI-demand signal → AI-citation signal

Then identify things such as:

“Our books/papers on medieval political thought are heavily cited academically but rarely surfaced in AI answers.”

or:

“These five open-access articles are disproportionately appearing in AI-mediated research, suggesting that accessibility + metadata + topic authority are working.”

That is much more actionable for an editorial/marketing team.

How I'd rank the options

The key distinction is that Scite is the one I'd investigate first for the AI-specific question. Its database also goes beyond simple citation counts by classifying citation statements as supporting, contrasting, or merely mentioning a finding.

What I'd actually buy/build

If your goal is specifically “which papers, university-press pages, and competing domains are actually being cited by ChatGPT/Perplexity/etc., so we can improve discoverability,” I’d separate this from traditional scholarly citation analysis.

My shortlist

• tryprofound.com — best for a serious university-press program. It tracks citations at the URL level across a broad set of AI engines, including ChatGPT, Perplexity, Gemini, Claude, Copilot and Google AI surfaces. That granularity is particularly useful for a press because you can distinguish which individual book/article/landing page gets retrieved rather than merely whether the press's domain is mentioned. Nick Lafferty
• semrush.com — best if your press already uses Semrush. It monitors which domains are cited for specified prompts and combines AI visibility with conventional SEO data. Semrush says its dataset includes hundreds of millions of AI-search prompts, making it useful for competitive benchmarking as well as monitoring your own domain. Semrush Semrush
• otterly.ai — best lower-cost starting point. It tracks a fixed set of prompts and identifies citations across ChatGPT, Perplexity, Gemini, Google AI Overviews and Copilot. Current published comparisons put it around $29/month at the entry level. Distribution.studio Semrush
• getintel.ai — particularly interesting if your question is literally “what sources did the AI cite?” Its distinguishing feature is storing the cited URLs from individual runs rather than reducing everything to a visibility score. It currently covers ChatGPT, Perplexity, Gemini and Google AI Overviews, with Claude on a higher tier. GetIntel
• scite.ai — complementary, not a replacement. For an academic press, Scite is excellent for the other side of the problem: how individual papers are being cited in scholarly literature and whether subsequent research supports or contradicts them. It also integrates with ChatGPT/Claude and provides an API. scite.ai scite.ai

What I'd buy for a university press

I'd probably start with Profound + Scite, rather than trying to make one product do both jobs.

Think of the two datasets as:

Scite:

“How influential/referenced is this scholarship within the scholarly ecosystem?”

Profound (or Semrush/Otterly):

“When someone asks an AI a question related to this scholarship, which sources does the AI actually retrieve and cite?”

That distinction matters. A book can have excellent conventional scholarly citations but be essentially invisible to AI search—and conversely, an ordinary web page can become a disproportionately important AI source.

There's also substantial volatility between AI systems: one 2026 analysis found only about 11% of cited domains overlapped between ChatGPT and Perplexity, while another Semrush analysis found only 25.6% overlap between different ChatGPT reasoning modes for the same prompts. Distribution.studio Semrush So I would not buy a tool that only measures “ChatGPT visibility.”

The metrics I'd want the press dashboard to contain

For each title/paper, track:

• AI citation rate: percentage of relevant prompts where the work/page is cited.
• Citation position: first source, middle of source list, etc.
• Citation URL: the exact page being retrieved.
• Citation type: book page, journal article, repository, review, Wikipedia, library record, news article, etc.
• Competing sources: which other papers/books/domains are being cited instead.
• Prompt/topic: e.g. “history of X,” “scholars studying Y,” “best books on Z.”
• Engine: ChatGPT, Perplexity, Gemini, Claude, Google AI.
• Change over time: whether a title is becoming more or less discoverable.
• Uncited opportunities: queries where competing scholarship is cited but your press's material isn't.
• Third-party citations: places such as university pages, reviews, bibliographies and institutional repositories that may be acting as the bridge through which AI discovers your book.

That last category is especially valuable for a press. Don't optimize only the publisher's own domain. If AI systems repeatedly cite a university repository, review, scholarly society page, or Wikipedia entry instead of your book page, that tells you something about the discovery path of the scholarship.

One important caveat: these products generally sample controlled prompts and observe the resulting citations; there isn't a universal feed containing every question every ChatGPT user asks. CiteFlow So I'd treat the resulting numbers as trend/benchmark measurements, not absolute counts of “all AI citations.”

If you tell me roughly how many titles your press publishes per year and whether you want an institutional/enterprise system or something a one-person marketing team can run, I can narrow this to 2–3 products and compare price, AI engines, URL-level citation data, API/export capability, and suitability for academic publishing.

The metrics I'd want the press dashboard to contain

For a university press, I’d look less at generic “AI SEO” tools and more at platforms that can answer:

1. Which books, chapters, articles, or domains are AI assistants citing?
2. For which research questions are our publications missing?
3. Which competing presses, journals, repositories, or institutions are getting surfaced instead?

A practical shortlist:

My recommendation for a university press

I would probably build a two-layer stack:

1. AI answer citation monitoring

Use one of:

• Profound (larger press, many disciplines, dedicated analytics team)
• Otterly.AI or Peec AI (smaller team, experimentation) CitationDesk Distribution.studio

Track prompts such as:

• “What are the best books on [field]?”
• “Who are leading scholars on [topic]?”
• “What should I read about [historical event/theory/method]?”
• “Recommend academic sources on [subject].”

Record:

• cited URLs
• cited domains
• publisher/platform
• author names
• whether the citation points to: - your landing page
• DOI page
• institutional repository
• Google Books
• JSTOR
• Internet Archive
• competitor press pages

2. Scholarly impact intelligence

Pair that with:

• Scopus
• Web of Science
• Dimensions
• Scite

These tell you traditional scholarly influence; AI trackers tell you AI-mediated discoverability. They are related but not the same. LLMrefs Grow with Tejas

Metrics I’d put on a press dashboard

A useful monthly dashboard might include:

• AI citation share

• % of tracked prompts where your press appears as a cited source
• Citation gap

• your press vs. university presses, journals, repositories, and commercial competitors
• Top AI-discovered titles

• books/chapters most frequently surfaced
• Lost-query report

• important scholarly questions where competitors appear but you do not
• Citation path

• “AI cited our book page” vs. “AI cited a third-party review mentioning our book”
• Metadata problems

• cases where AI cites an author’s profile, repository copy, or secondary source instead of the canonical press page

A low-cost pilot I’d run first

For 8–12 weeks:

1. Pick 100–200 high-value prompts across your strongest subject areas.
2. Run them weekly in ChatGPT, Claude, Gemini, and Perplexity.
3. Store every cited source URL.
4. Identify patterns: - Are AI systems finding your books?
5. Are they finding reviews instead?
6. Are they missing your catalog because metadata is weak?
7. Improve: - structured metadata
8. author pages
9. chapter abstracts
10. open-access excerpts where appropriate
11. persistent identifiers (DOIs, ORCID links, Crossref metadata)

The biggest strategic shift is that discoverability is moving from “rank for a keyword” toward “be a trustworthy source an AI system chooses to cite.” AI visibility tools are emerging quickly, but no single platform gives a complete scholarly picture yet. SiteTest.ai Nick Lafferty

For a university press, I’d prioritize citation-level source tracking over simple brand mentions—knowing which books and pages get cited is much more actionable than knowing your press name appeared. GetIntel

Yes. For a university press, I’d look at this as AI citation intelligence, not ordinary SEO: you want to know which books/articles/domains are being retrieved and cited by AI systems, for which scholarly questions, and what you can do to make your own catalog more retrievable.

My shortlist

My pick: Profound + a scholarly citation database

I'd probably pilot Profound first. The important capability isn't merely "does the press get mentioned?" It's being able to drill down to:

Prompt → AI engine → cited URL → domain → specific book/article → competing sources → frequency over time

That lets your editorial/marketing team ask much more useful questions, such as:

• Which of our books are cited when someone asks about topic X?
• Which competing university presses are disproportionately represented?
• Are AI systems citing our book landing pages, Google Books/Internet Archive records, reviews, repositories, or third-party scholarship?
• Which domains repeatedly appear alongside our competitors?
• What sources are AI systems using instead of our books?
• Are citations concentrated in a handful of titles or distributed across the catalog?

The distinction between mention, citation, recommendation, and co-citation is particularly important here. A title can be discussed without its actual landing page being cited. Nick Lafferty

One thing I would not do

Don't buy a tool purely because it gives you a single "AI visibility score."

AI citation results are noisy and vary substantially by engine and sampling. Recent research argues for measuring not just whether something was cited, but how much influence the cited source actually had on the generated answer. arXiv

For a press, I'd therefore want your dashboard to retain the individual response and cited URLs, not just aggregate rankings.

A useful internal dataset would look roughly like:

date | engine | prompt | title/ISBN | cited URL | source domain | competitor? | citation position | answer context | landing-page type

Then you can build genuinely editorial insights from it.

The scholarly-specific wrinkle

There's also evidence that AI-mediated discovery of academic resources favors structured metadata, stable permalinks, and freely accessible material. A recent study of academic-library resources found ChatGPT, Perplexity, and Gemini driving AI-mediated traffic and highlighted those characteristics. arXiv

So I'd pair AI citation monitoring with a bibliographic/metadata audit of your catalog:

• DOI/ISBN consistency
• Schema.org/Book metadata
• author/editor metadata
• abstracts and substantial descriptions
• table of contents
• chapter-level metadata where possible
• persistent URLs
• links between books, chapters, reviews, and related scholarship
• repository/open-access copies where licensing permits
• structured citation information

That turns the monitoring system into something actionable rather than just a marketing dashboard.

If I were procuring this for a press

I'd run a 30-day bake-off between Profound, GetIntel, and Otterly.AI, using perhaps 100–200 questions spanning your major subject areas. I'd compare:

1. Coverage of ChatGPT, Perplexity, Gemini, Claude, etc.
2. Whether you can export every cited URL, not merely your own citations.
3. Citation-level history and reproducibility.
4. Ability to group results by domain, title, author, subject, and competitor.
5. API/export capabilities.
6. Whether the tool can distinguish a publisher's individual book URLs from the publisher's domain generally.
7. Cost at your eventual prompt volume.
8. Data retention—important if you want to build a multi-year scholarly-discoverability dataset.

For your particular use case, I'd weight #2 and #6 much more heavily than the vendor's headline "AI visibility score."

If you tell me roughly how large the press's catalog is (e.g. 1,000 vs. 20,000 titles) and whether you're looking for a $100/month, $1k/month, or enterprise-scale solution, I can narrow this to 2–3 products and compare their current capabilities/pricing.

Then you can build genuinely editorial insights from it.

The scholarly-specific wrinkle

For a university press, I’d prioritize tools that can answer two separate questions:

1. Which of our books/papers/pages are being cited by AI?
2. Which outside academic domains and specific URLs are AI systems relying on instead of us?

That second question is especially valuable for improving discoverability.

My shortlist

My pick: Profound + Scrunch

For a university press, I'd actually demo both Profound and Scrunch rather than choosing solely on generic "AI visibility" scores.

Profound looks particularly strong for an organization that wants to turn citation data into an editorial/content strategy: it categorizes cited sources, shows competitors and third-party sources, and explicitly supports identifying publishers/authors that drive citations. Profound

Scrunch has a particularly relevant workflow for your question: its citation reporting exposes top cited domains → individual URLs → the prompts producing those citations, with filters for platform, topic, country, branded/non-branded queries, etc. Scrunch

That lets your press ask something much more useful than "What's our AI visibility?"

For questions about [field/topic], which scholarly publishers, journals, repositories, universities and individual papers does AI cite—and where are we absent?

How I'd use this for an academic press

I'd build a citation intelligence corpus around your catalogue rather than simply monitoring the press's brand name.

For example, for every major subject area, create prompts such as:

• "What are the foundational works on [topic]?"
• "What are the best recent books on [topic]?"
• "What scholarship should I read to understand [specific debate]?"
• "Which scholars have written about [concept]?"
• "What are the major criticisms of [theory]?"
• "What primary sources are important for researching [subject]?"
• "What are the best academic sources on [historical event]?"
• "[Book/article title] — what other scholarship discusses this?"
• "Compare the major interpretations of [topic]."

Then track the actual cited URLs, not just mentions of your press.

I'd put the resulting data into categories such as:

• Your publications
• Other university presses
• Commercial academic publishers
• Journal platforms
• Institutional repositories
• University domains
• Google Scholar/indexing ecosystems
• Wikipedia
• Government/NGO sources
• Professional associations
• Blogs/news/media
• Reddit/social/community sources

That produces a much more actionable picture of the academic information ecosystem feeding AI answers.

One important caveat

Don't treat an AI citation count as equivalent to a citation in scholarly literature.

AI systems retrieve and cite sources differently, and their citation behavior can be quite noisy. Recent research proposes distinguishing citation selection (the system chooses a source) from citation absorption (the source actually influences the generated answer). arXiv

That's actually a useful framework for a press: you ultimately care not merely whether a book's URL appeared in an answer, but whether the book's scholarship is being used to construct the answer.

There is also emerging evidence specifically relevant to academic publishing: a 2026 study of AI-mediated discovery of academic-library resources found that ChatGPT, Perplexity and Gemini were important referral sources, with structured metadata, stable permalinks and openly accessible resources particularly well positioned for retrieval. arXiv

If I were procuring this for your press

I'd run a 2–4 week bake-off using the same ~100–300 scholarly prompts in Profound, Scrunch, and Ahrefs, and score them on:

• URL-level citation capture
• paper/book-level identification
• domain-level aggregation
• coverage of ChatGPT, Gemini, Perplexity, Claude and Google AI surfaces
• historical trend data
• prompt-level reproducibility
• export/API capability
• ability to distinguish your content from third-party citations
• identification of who is winning citations instead of you
• cost at your catalogue/prompt volume

For your particular use case, Profound would be my first enterprise demo and Scrunch my second. Ahrefs is the obvious third if your press already has an SEO/content team using Ahrefs.

If you tell me roughly how many books/papers/domains you want to monitor and whether you need an API/CSV export, I can narrow this to the 2–3 best options and compare current pricing and capabilities.

There is also emerging evidence specifically relevant to academic publishing: a 2026 study of AI-mediated discovery of academic-library resources found that ChatGPT, Perplexity and Gemini were important referral sources, with structured metadata, stable permalinks and openly accessible resources particularly well positioned for retrieval.

If I were procuring this for your press

• "What are the foundational works on [topic]?"
• "What are the best recent books on [topic]?"
• "What scholarship should I read to understand [specific debate]?"
• "Which scholars have written about [concept]?"
• "What are the major criticisms of [theory]?"
• "What primary sources are important for researching [subject]?"
• "What are the best academic sources on [historical event]?"
• "[Book/article title] — what other scholarship discusses this?"
• "Compare the major interpretations of [topic]."

Then track the actual cited URLs, not just mentions of your press.

I'd put the resulting data into categories such as:

• Your publications
• Other university presses
• Commercial academic publishers
• Journal platforms
• Institutional repositories
• University domains
• Google Scholar/indexing ecosystems
• Wikipedia
• Government/NGO sources
• Professional associations
• Blogs/news/media
• Reddit/social/community sources

That produces a much more actionable picture of the academic information ecosystem feeding AI answers.

One important caveat

Don't treat an AI citation count as equivalent to a citation in scholarly literature.

AI systems retrieve and cite sources differently, and their citation behavior can be quite noisy. Recent research proposes distinguishing citation selection (the system chooses a source) from citation absorption (the source actually influences the generated answer).

That's actually a useful framework for a press: you ultimately care not merely whether a book's URL appeared in an answer, but whether the book's scholarship is being used to construct the answer.

If your goal is specifically “Which of our books/papers/domains are being cited inside ChatGPT, Perplexity, Gemini, etc., and what can we do to increase that?”, I’d separate the market into AI-answer citation trackers and scholarly-impact platforms. They solve different problems.

My shortlist

What I'd actually buy for a university press

1. GetIntel (or a comparable AI-citation specialist) as the experimental layer.
You want a system that records something like:

Query → AI engine → answer → cited URL → cited paper/book → publisher domain → competitor domains → date

That's much more actionable than a generic “AI visibility score.” Current comparisons emphasize that citation tracking and brand/visibility tracking are different jobs: you care about the actual sources AI systems retrieve and cite. getintel.ai

2. Keep/use Altmetric for your existing scholarly-impact infrastructure.
Altmetric is particularly attractive for a press because it can identify attention around individual research outputs, journals, and books, and its Explorer/API can be used for institutional analysis. www.altmetric.com It also ingests conventional scholarly citations from Dimensions, including books and other non-journal outputs. help.altmetric.com

3. Add Overton if policy impact matters.
For a university press, this can reveal a very valuable pathway that ordinary citation counts miss: our book → policy document → policymaker/organization. www.overton.io

The important strategic point

I wouldn't optimize merely for “getting ChatGPT to cite our domain.”

I'd build a citation-source benchmark across perhaps 100–500 questions in each major subject area:

• What sources does ChatGPT cite?
• What sources does Perplexity cite?
• What sources does Gemini cite?
• Which university-press domains appear?
• Which commercial publishers appear?
• Which individual books/papers get cited repeatedly?
• Which pages on your site get cited?
• Are PDFs cited more often than landing pages?
• Does a DOI resolve to a page AI systems can crawl?
• Which external sites repeatedly get cited instead of your pages?
• What characteristics do those winning pages have?

That last question is where the commercial value is. The objective isn't simply measuring AI citations; it's identifying the discoverability characteristics that precede them.

There's also a good reason to measure several engines: recent analyses find substantial differences in which domains they cite, so optimizing for one assistant can give you a misleading picture of overall discoverability. www.geosoftwarerankings.com

One caveat for a university press

Don't assume that a tool claiming to measure “AI citations” has a comprehensive census. These systems generally sample queries and observe responses, while AI answers can vary between runs. So I'd demand that a vendor expose:

• exact queries tested
• engine/model tested
• timestamp
• cited URLs
• citation position
• repeatability/variance
• query volume
• historical data
• export/API
• domain-level and individual-ISBN/DOI-level reporting

That makes the data defensible enough to use in a publisher's discoverability program rather than treating an opaque AI score as fact.

My pick: start with GetIntel + Altmetric, with Overton added if policy/social-impact publishing is important. I'd pilot the AI-citation component against a few hundred of your own titles before committing to an enterprise contract. www.altmetric.comwww.altmetric.comgetintel.ai

If you tell me roughly how many titles your press has and which disciplines you publish, I can narrow this to a 3-tool stack and suggest the metrics/dashboard I'd give your editorial, marketing, and web teams.

If your goal is specifically “which academic papers, publishers, journals, authors, and domains are being cited by ChatGPT/Perplexity/Gemini, and how do we improve our university press’s chances of appearing there?”, I’d separate this into two layers:

My recommendation for a university press

Don't treat this as ordinary citation tracking. Build an AI discoverability observatory around a controlled set of research questions.

For example, every week/month run 100–500 prompts such as:

“What are the most important books on Hawaiian history?”
“What scholarship discusses [topic]?”
“What are the leading academic sources on [topic]?”
“Which books should I read about [topic]?”

Then record:

• Which papers/books were cited
• Which DOIs/ISBNs were cited
• Which publisher domains appeared
• Which journal/repository domains appeared
• Position/rank of each source
• Model and model version
• Exact prompt
• Date/time
• Whether your press's content was mentioned but not cited
• Which competing publishers repeatedly appear

That's much more actionable than simply asking whether your domain has been mentioned.

The important distinction

There are actually three different metrics I'd track:

1. Scholarly citation impact
“How often is our scholarship cited by other scholarship?”

OpenAlex is particularly useful here. Its API can follow both incoming citations and the works referenced by a publication, and it supports publisher/topic/institution analysis. help.openalex.org

2. AI citation visibility
“When an AI answers a relevant question, does it cite our content?”

This requires querying the AI systems themselves (or an AI-visibility service). Conventional bibliographic databases cannot tell you this.

3. AI-source authority
“Which domains does AI preferentially use when answering questions in our subject areas?”

This is arguably the most interesting metric for a press. You can aggregate AI answers to discover that, say, jstor.org, a particular university repository, a competitor's catalog, Wikipedia, or your own domain disproportionately supplies citations for a particular field.

One caveat about commercial AI-visibility tools

I'd be cautious about treating their numbers as equivalent to actual global ChatGPT usage. AI search results are volatile, and citation distributions can change substantially over short periods. For example, recent analysis found a sharp change in Reddit's share of ChatGPT citations over just a few weeks. www.axios.com

So for a university press, I'd want raw prompt/response observations alongside the dashboard, rather than relying on a proprietary “AI visibility score.”

A particularly good stack

I'd probably use:

OpenAlex → scholarly baseline
↓
Crossref/DOI + ISBN metadata → normalize your catalog
↓
Promptwatch or Profound → commercial AI monitoring
↓
Your own controlled prompt corpus → reproducible research
↓
Dashboard → AI citations by title, subject, domain, competitor, model and time

OpenAlex is especially attractive for the foundation because its catalog covers hundreds of millions of scholarly works and exposes publishers, sources, topics, citations and related works through an API. developers.openalex.orgdevelopers.openalex.org

OpenAlex API documentation

If you tell me roughly how many titles your press has and whether you're primarily interested in books, journals, or both, I can narrow this down to the 3–5 best tools and a concrete monitoring architecture, including approximate cost and what each can actually measure.

If the goal is “which scholarly papers, publisher pages, and competing domains are AI assistants actually citing?”, I’d treat this as an AI-citation intelligence problem rather than a conventional academic citation-tracking problem.

My shortlist

For your particular use case, I'd start with Profound and Rankscale. The distinction is important: you don't merely want to know “Does University Press X appear in ChatGPT?” You want data like:

For questions about 19th-century American history, which domains does ChatGPT cite? Which individual URLs? Which university presses? Which journal articles? Which repositories? How often does our content appear versus competitors?

Rankscale's published feature set is especially aligned with that question because it exposes domain/URL citation patterns and competitor citation share. rankscale.ai

One important caveat for a university press

Don't confuse AI citations with scholarly citations.

Tools such as scite are excellent for analyzing how academic papers are cited in the scholarly literature, but that's a different measurement problem. You want an additional layer measuring AI retrieval/citation behavior.

I'd build your measurement stack around four levels:

1. AI query → cited domain
   • JSTOR
   • yourpress.org
   • competing presses
   • Wikipedia
   • Google Books
   • institutional repositories
   • journal sites, etc.

2. AI query → cited URL
   • Which specific book landing pages or articles are being retrieved?

3. AI query → cited scholarly work
   • DOI, ISBN, author, title, edition/year.

4. AI engine → citation share
   • ChatGPT vs. Perplexity vs. Gemini vs. Claude vs. Google AI Overviews/AI Mode.

That last point matters because the engines don't necessarily cite the same sources. One 2026 analysis found substantial differences in which domains appear across ChatGPT and Perplexity, so measuring only one assistant can give a misleading picture. distribution.studio

I'd also create a publisher-specific prompt corpus

This is probably more valuable than simply monitoring your press name.

For example, for every major subject area you publish in, maintain 50–200 recurring questions:

• “What are the best books on the Harlem Renaissance?”
• “What are the major scholarly interpretations of X?”
• “Which books should I read to understand Y?”
• “What primary sources are available for Z?”
• “Who are the leading historians of X?”
• “What are the most influential books about [topic] published since 2015?”

Then run the same questions periodically across the major AI systems and record:

question → answer → citations → cited URLs → cited domains → authors/books mentioned → competitors → your books/pages → position/order of citation.

That gives you a much more defensible AI discoverability benchmark than a generic “AI visibility score.”

Perplexity's own documentation is also revealing here: its academic-search workflow explicitly supports academic-domain filtering, DOI/paper metadata extraction and citation-chain construction. docs.perplexity.ai

What I'd buy

If I were setting this up for a university press today:

Tier 1 — serious institutional program:
Profound + your existing SEO analytics + scholarly citation data (Scite/Crossref/etc.)

Tier 2 — lower-cost pilot:
Rankscale + a carefully designed recurring prompt set

Tier 3 — build your own:
Use an AI-search API/collection layer to run the prompts and store every cited URL, then analyze domains, books, authors and competitors yourself. This gives you much more control over academic-specific metrics than a generic SaaS dashboard.

The key metric I'd ultimately report to the press director wouldn't be “AI visibility.” It would be something like AI Scholarly Citation Share: the percentage of relevant AI answers in which one of our books/articles/pages is cited, versus competing publishers and source types.

If you tell me roughly how many titles your press publishes per year and which disciplines you cover, I can narrow this to a practical stack and suggest the exact metrics/dashboard I'd use.

Yes. For a university press, I’d approach this somewhat differently from ordinary “AI visibility”/SEO tracking. You want to know which scholarly articles, books, repositories, journals, libraries, and domains are actually being surfaced as sources by AI systems, not merely whether your press is mentioned.

My shortlist

What I'd choose for a university press

Profound + your own scholarly metadata layer would probably be my first choice.

The key is to turn raw AI citations into something like:

AI engine → question/topic → cited URL → cited domain → publication → author → ISBN/DOI → subject → publication year → competitor/own title → citation frequency → citation context

That lets your editorial/marketing team answer much more useful questions:

• Which books and articles from our catalog get cited?
• Which of our authors are being surfaced?
• Which external domains are consistently cited instead of our publications?
• Which university repositories, JSTOR-like platforms, Wikipedia pages, reviews, library catalogs, and scholarly sites appear in AI answers?
• Which prompts/topics produce citations to our books?
• Are AI systems citing the publisher's landing page, a Google Books page, an institutional repository, a review, or an unrelated secondary source?
• Which competitor university presses are disproportionately represented?
• Which citations lead to actual traffic to our catalog?

That last distinction matters. Citation frequency isn't necessarily citation influence. Recent research on AI citation measurement argues for separating whether a source was selected from whether its content actually influenced the generated answer. arxiv.orgwww.scalepost.ai

One particularly important distinction

I'd measure three separate things:

1. AI retrieval

Did an AI system fetch/read our page?

2. AI citation

Did the answer actually link to our page/book/article?

3. AI influence

Did information from that publication materially appear in the answer?

Those aren't equivalent. A CDN/log-based system such as ScalePost is valuable for #1, while prompt-based citation trackers are better for #2 and #3. arxiv.orgwww.scalepost.ai

For your specific use case, I'd build a “scholarly AI citation map”

Rather than tracking only your press's domain, I'd create a controlled prompt set around your subject areas—for example:

• “best books on [subject]”
• “scholars who argue [proposition]”
• “academic research on [topic]”
• “[historical question] primary sources”
• “recommended books for a graduate seminar on [topic]”
• “most influential recent scholarship on [topic]”

Then collect every cited URL and enrich it with Crossref/OpenAlex/catalog metadata.

The output becomes a network:

AI question → AI engine → cited source → domain → author → title → publisher → subject

That is much more strategically valuable to a university press than a generic “AI visibility score.”

There is also evidence that different AI engines have substantially different citation patterns, so I would not make ChatGPT your only measurement target. semai.ai

If I were procuring this today:
Profound for the commercial intelligence layer + ScalePost if first-party crawler data is important + a small internal Crossref/OpenAlex/catalog database to turn those citations into scholarly intelligence. That combination would give a university press something much closer to an “AI scholarly discovery observatory” than an SEO dashboard.

Yes. But for a university press, I’d use a somewhat different stack than a typical brand/SEO team.

The key distinction is between:

1. AI-answer visibility tracking — “When someone asks ChatGPT/Perplexity/etc. about a topic, which domains and books/papers get cited?”
2. Scholarly citation tracking — “Which papers/books are being cited by other scholarly works?”
3. Discoverability diagnostics — “Why is our content not being retrieved, and what should we change?”

No single product is excellent at all three.

My shortlist

The AI-visibility category has matured considerably in 2026: Profound, Peec, Otterly, Scrunch and similar platforms now run controlled prompt sets against multiple answer engines and record mentions, citations, source URLs and competitors. getrefine.ai

1. I'd start with Profound if you have a real analytics budget

Profound is aimed at enterprise teams and provides granular answer-engine analytics rather than simply checking whether a brand was mentioned. Current comparisons put it toward the top for breadth of engine coverage and reporting. getrefine.aiomidsaffari.com

For a press, I'd configure prompts around things like:

• "best books on medieval economic history"
• "scholars who write about X"
• "most important research on Y"
• "what does the literature say about Z?"
• "recommended books for a graduate seminar on X"
• "primary sources for researching Y"
• "recent scholarship on Z"

Then track which publisher domains, book pages, DOI pages, repositories and individual papers appear in the answers.

That gives you something much more useful than conventional SEO rankings: an AI citation share-of-voice for your catalog.

2. Peec AI is probably the best smaller-team alternative

Peec is particularly attractive if you want a clean monitoring/benchmarking system without buying a heavyweight enterprise platform. Current comparisons describe it as focused on share of voice, competitor benchmarking and multi-engine monitoring. getrefine.aiomidsaffari.com

For a press, I'd create separate projects for:

• subject areas
• individual imprints
• major journals/series
• flagship books
• competing university presses
• important scholarly domains

That lets you answer questions such as “Are our books becoming more visible in AI answers about American history?” rather than merely “Does our homepage get cited?”

3. Otterly.ai is where I'd run the pilot

If you're trying to prove the concept before asking the press to fund it, Otterly is appealing because its entry point is relatively inexpensive and it provides scheduled prompt monitoring across AI surfaces. Current 2026 comparisons put its entry pricing around $29/month, although pricing and engine combinations change. getrefine.aiomidsaffari.com

I'd spend a month tracking perhaps 50–100 high-value questions across your strongest subject areas.

The output I'd want isn't just “we were cited.” It's:

Prompt → AI engine → answer → cited URL → cited domain → cited title → position in answer → competing source → date

That dataset becomes surprisingly valuable.

But add Scite for the scholarly side

This is the important part for a university press.

Scite

Scite isn't simply an “AI visibility tracker.” Its strength is the scholarly citation graph: it can show citation context and distinguish supporting, contrasting and mentioning citations. Its current infrastructure also exposes Smart Citation data to ChatGPT, Claude, Gemini and other AI tools through its MCP integration. scite.ai

That makes it particularly interesting for a press because you can investigate the scholarly ecosystem surrounding your publications, while the AI-visibility tools tell you what happens when people ask answer engines questions.

Scite also explicitly has a publisher-oriented product, so I'd talk to them about whether they can provide publisher-level analytics rather than treating it purely as an individual-researcher subscription. scite.ai

And use Dimensions as the institutional baseline

Dimensions

Dimensions gives you conventional bibliometric measures and programmatic access to article-level citation indicators through its Metrics API. It also now has a Dimensions Research GPT that grounds conversational answers in its research database. www.dimensions.ai

That's useful because you can eventually compare:

traditional scholarly impact → AI discoverability → AI citation

For example:

That is a much more interesting publisher dashboard than either SEO traffic or citation counts alone.

The big caveat: don't treat AI citation counts as objective

This is still an experimental measurement problem.

AI answers can change from query to query, and different engines retrieve different sources. Recent research also shows that even the bibliographic-reference capabilities of major chatbots can be surprisingly unreliable, so you don't want to interpret a single generated citation as equivalent to a Crossref/Scopus/Dimensions citation. arxiv.org

I'd therefore measure repeatable visibility, not “the number of times ChatGPT ever cited us.”

For example, run each prompt 5–10 times per month and calculate:

• Citation rate: % of runs citing your domain
• Source share: your citations / all citations
• Position: where your source appeared
• Unique titles cited
• Competitor share
• Engine differences
• Citation persistence
• Newly cited titles
• Lost citations

That produces a defensible AI Discoverability Index.

One particularly useful metric for a press

I'd create:

AI Scholarly Share of Voice

For each subject area, ask a fixed panel of questions and calculate what proportion of the scholarly sources cited by the AI engines belong to your press.

Then compare that against:

• Oxford
• Cambridge
• Harvard
• Chicago
• Princeton
• Yale
• other relevant specialist presses

The goal isn't necessarily to “beat” them. It lets your editorial/marketing/discovery teams identify topics where your scholarship is disproportionately absent from AI-mediated discovery.

My recommended stack

If I were setting this up for a university press today:

Pilot:
Otterly + Scite + Dimensions/Semantic Scholar

Serious institutional program:
Profound + Scite + Dimensions

If budget is constrained:
Peec + Scite

And I would not buy an AI-visibility product solely because it gives you a big “AI visibility score.” The valuable asset is the underlying prompt → answer → cited source → domain → competitor → time-series dataset.

Semantic Scholar is also worth keeping in the workflow because its citation graph includes citation classifications and influence indicators and covers more than 200 million papers. webflow.semanticscholar.org

If your actual goal is to build a university-press dashboard showing which books/papers/domains ChatGPT, Claude, Gemini and Perplexity cite, I can also lay out the exact data schema, KPIs, sampling methodology, and a recommended $500/month vs. $5,000/month stack.

For a university press, you probably want a tool that tracks which URLs/domains and specific books/articles are being used as sources by chat-based AI systems, not just “AI mentions” of your press name. The market is still immature, so I’d combine an AI visibility tracker with your existing scholarly analytics (Crossref, COUNTER, Google Scholar, Dimensions, Scopus, etc.). nicklafferty.com

Here are the strongest options by use case:

For a university press specifically, I’d set up tracking around:

1. Priority scholarly prompts
   • “best books on [field]”
   • “overview of [historical topic]”
   • “key scholarship on [theory/topic]”
   • “recommended readings for graduate students in [discipline]”

2. Citation-level metrics
   • % of prompts where your books/articles appear as sources
   • which competitor presses appear instead
   • which domains AI trusts for your fields
   • which individual landing pages (book pages, chapter pages, repository pages) get cited

3. Metadata improvements

AI systems often rely heavily on structured, crawlable information. Make sure every title page has:

• complete author/editor metadata
• abstracts and tables of contents
• chapter-level pages where possible
• DOI/Crossref links
• schema.org structured data
• accessible full-text previews where licensing allows

4. Create an “AI discoverability dashboard”

I would not measure only “how often AI cites the press homepage.” The more valuable signal is:

• Which books become canonical answers?
• Which authors are invisible to AI despite strong scholarly impact?
• Which competing titles are being recommended instead?

A practical stack for a medium-sized university press:

• Profound or Ahrefs/Semrush → monthly AI visibility monitoring
• Crossref + COUNTER + Google Scholar/Scopus/Dimensions → scholarly impact baseline
• Custom prompt library → your discipline-specific AI citation benchmark
• Spreadsheet/database → record cited URLs, competitors, and interventions

One caution: AI citation tracking is not yet equivalent to citation indexes. A model citing a book does not necessarily mean it has broad scholarly influence; it reflects the model’s retrieval sources, training data, indexing, and retrieval behavior. Treat it as a discoverability signal, not a new impact factor. nicklafferty.com

If you tell me your press size (small specialist press vs. major university press) and whether you care more about books, journals, or institutional brand visibility, I can suggest a more tailored stack.

For a university press, I’d prioritize tools that track the actual URLs and domains cited in AI answers, not just whether your press name is mentioned. That lets you answer questions like “Which books/articles from our catalog are being surfaced?”, “Which competing presses are cited instead?”, and “What external domains are AI systems trusting for this subject?”

My shortlist

If I were buying for a university press, I'd demo Profound and Rankscale first. Profound looks strongest for institutional-scale monitoring; Rankscale is especially attractive if your central question is “what sources/domains are AI systems actually citing?”

One important distinction for a press

Don't measure only “Is Oxford University Press mentioned?” or “Is our domain cited?”

I'd build your dashboard around four levels:

1. Title/article level — which individual books, chapters, journal articles, reviews, landing pages, etc. get cited.
2. Domain level — press.example.edu versus JSTOR, Wikipedia, Google Books, HathiTrust, competing presses, newspapers, blogs, etc.
3. Topic/query level — e.g. 100–500 questions corresponding to subjects in your catalog.
4. Engine level — ChatGPT vs. Perplexity vs. Gemini vs. Claude, because their citation behavior differs substantially. One 2026 analysis found only about 11% overlap in cited domains between ChatGPT and Perplexity, illustrating why single-engine measurement can be misleading. www.geosoftwarerankings.com

For academic publishing specifically, there's also evidence that structured metadata, stable permalinks and openly accessible resources can make scholarly material easier for AI systems to surface. arxiv.org

I'd also track "citation competitors"

This may be more valuable than raw citation counts.

Suppose you track:

“What were the major interpretations of the Haitian Revolution?”

and the AI repeatedly cites Wikipedia, Britannica, a competing university press, JSTOR and three individual scholars—but never your excellent book on the topic.

Your useful metric isn't merely “our book got 0 citations.” It's:

Who got the citations instead, from which URLs, for which questions, and what characteristics do those sources share?

That turns AI monitoring into an actual discoverability program for your press.

One caveat: AI citation measurements are inherently sampling-based. A single run isn't a stable “ranking”; you want a fixed prompt corpus, repeated measurements, engine-specific results, and historical trends. nicklafferty.com

If you tell me roughly how large your catalog is (e.g. 1,000 vs. 50,000 titles) and whether you're looking for a commercial SaaS purchase or something your analytics team could build, I can narrow this to the best 2–3 options and suggest the metrics/data model I'd use for a university press.

Yes. For a university press, I’d look beyond ordinary academic citation databases. The thing you want to measure is closer to “AI citation visibility”: When someone asks ChatGPT, Perplexity, Gemini, etc. a question related to our books/papers/authors, which scholarly sources and domains does the system actually cite?

My shortlist

For your particular use case, I'd pick Scrunch + Scite

The distinction is important:

Scrunch answers:

“What does ChatGPT/Perplexity/etc. cite when people ask questions in our subject areas?”

Scite answers:

“What scholarly literature cites and discusses our publications, and is that citation supportive or contradictory?”

That gives you two very different kinds of discoverability.

Scrunch's citation system can show, for example:

Prompt: “What are the leading books on Reconstruction-era U.S. politics?”

AI cites: JSTOR → 4 times
HathiTrust → 3
Wikipedia → 2
Publisher X → 2
Your press → 0

Opportunity: Your book is relevant, but AI is getting its evidence through third-party pages rather than your own catalog.

That is much more actionable for a press than simply knowing that a book has 47 conventional citations.

One important caveat

Don't interpret an AI-citation dashboard as an exact census of what ChatGPT or another model “knows.” These systems retrieve different sources depending on the query, model, geography, time, and retrieval configuration. Recent research also finds that AI citation behavior can be highly concentrated among a relatively small set of domains. searchengineland.com

So I'd establish a repeatable prompt panel rather than asking the tools to measure everything.

For a press, I'd monitor perhaps:

• 50–200 questions per discipline
• “best books on X”
• “scholars/books about X”
• “introduction to X”
• “primary sources for X”
• “recent research on X”
• questions involving individual authors
• questions involving specific books
• undergraduate/student-oriented research questions
• graduate-level literature-review questions

Then track citation share by domain, individual URL, book, author, and AI platform over time.

What I'd build for a university press

The most useful KPI isn't simply “How many times were we cited?” I'd create a dashboard with:

AI discoverability

• % of relevant prompts where a press title appears
• % where the press's own domain is cited
• citations per title
• citations per author
• citation share by AI platform

Competitive intelligence

• domains most frequently cited instead of yours
• books from competing presses appearing in answers
• third-party pages that repeatedly influence AI answers
• topics where you're authoritative academically but invisible to AI

Actionability

• titles with high scholarly relevance but zero AI visibility
• catalog pages frequently cited but poorly structured
• Wikipedia/JSTOR/review pages that AI repeatedly uses
• third-party sites worth pursuing for coverage/linking
• metadata/content gaps

That last category is where the real value lies. Scrunch explicitly supports grouping citations by domain and URL and identifying third-party citation opportunities, while Peec provides a similar “top sources” view. peec.aihelpcenter.scrunchai.com

Bottom line: If you want something you can buy and start using, Scrunch would be my first evaluation, with Peec as the closest alternative. I'd add Scite rather than trying to make an AI-visibility product double as a scholarly bibliometrics platform.

If your press has a sizable backlist, I'd also seriously consider a custom OpenAlex + AI-query dataset alongside one of these tools. That would let you measure AI visibility at the ISBN/DOI → title → author → subject → domain level, which is much closer to how a university press actually manages its catalog. www.lse.ac.uk

For a university press, you are looking for something closer to “AI citation intelligence” / “generative engine optimization (GEO) monitoring” than traditional bibliometrics. The market is still immature: no tool has complete visibility into every ChatGPT, Claude, Gemini, or Perplexity answer, so the strongest approach is usually a combination of prompt monitoring + web citation tracking + scholarly metadata analytics. openattribution.org

Here are the tools I would evaluate:

aixiv.science

What I would build for a university press

A practical stack:

1. AI answer monitoring

Track a fixed prompt library, for example:

• “What are the best books on [field]?”
• “Which scholars are influential in [topic]?”
• “Recommend university press books about [subject].”
• “What sources should I cite for [research question]?”

Run these weekly/monthly across:

• ChatGPT
• Claude
• Gemini
• Perplexity
• Google AI Overviews (where available)

Capture:

• cited URLs
• cited books/papers
• author mentions
• publisher mentions
• whether the AI cites your landing pages, metadata pages, PDFs, reviews, or third-party references

This is similar to how emerging AI-visibility research measures “citation share” by extracting URLs returned in answers. openattribution.org

2. Scholarly metadata optimization

Pair AI monitoring with:

• Crossref metadata quality checks
• DOI resolution health
• Google Scholar visibility
• OpenAlex coverage
• Semantic Scholar coverage
• library catalog completeness

A book that exists in AI training/search ecosystems as clean structured data has more opportunities to surface.

3. Domain-level citation tracking

Measure:

• How often your domain appears
• Which pages get cited
• Which competing presses appear instead
• Which topics you “own” vs. lose

For example:

Tools I would shortlist for a press

If you have a small analytics team:

1. Otterly.ai
2. Ahrefs
3. Scite
4. Dimensions

If you have an enterprise marketing/communications team:

1. Profound
2. Scrunch AI
3. Semrush
4. Dimensions

If you want a research-grade internal system, I would build a lightweight pipeline:

• Prompt database
• Scheduled API runs
• Citation extraction
• DOI/title matching
• Dashboard in Tableau/Power BI
• Alerts when competitors appear more often than your titles

Important caveat

Do not optimize only for “getting cited by AI.” AI citation behavior is not the same as scholarly citation behavior, and current chatbots can produce incorrect or incomplete bibliographic references. doi.org A press should track AI visibility as a new discovery channel, alongside library holdings, reviews, citations, downloads, and sales.

For a university press, I would probably start with a 90-day pilot:

• 200–500 high-value prompts
• 5 AI systems
• top 1,000 titles
• competitor benchmark against 5–10 presses
• monthly “AI discoverability report”

That will tell you whether your bottleneck is metadata, content structure, authority, or simply that the AI systems are not surfacing scholarly publishers in your subject areas.

Yes. For a university press, I’d approach this a little differently from ordinary “AI visibility” software: you want to know not just whether your press/domain is mentioned, but which individual books, articles, DOI pages, repositories, reviews, and other scholarly domains are being cited in AI answers.

My shortlist

Scrunch is particularly close to what you're describing. Its citation system records the actual webpages cited in AI responses, lets you drill from domain → URL → prompts, and can compare your press with competitors and third-party sources. scrunch.com

Profound is the stronger choice if you're looking for a serious institutional analytics platform. Current comparisons report URL-level tracking across a particularly broad set of AI engines, including ChatGPT, Perplexity, Gemini, Claude, Copilot, Grok and others. nicklafferty.com

Semrush is attractive if your communications/marketing team also needs ordinary search analytics. Its current AI Visibility tooling can show which pages get cited, which topics generate citations, competitors' citation share, and prompt-level results. www.semrush.com

The important caveat for a university press

None of these is really measuring “all citations made by ChatGPT.” They're measuring citations generated when the vendor runs a defined/sample set of prompts through AI systems.

That's important because AI citation behavior varies enormously by prompt and model. For example, recent Semrush research found only 25.6% overlap in cited domains between ChatGPT's low- and high-reasoning modes for the same prompts. www.semrush.com An academic study similarly argues that citation selection and the extent to which a cited source is actually incorporated into the answer should be treated as separate measurements. arxiv.org

So I'd use these tools to answer:

“When people ask AI systems the kinds of scholarly questions relevant to our books, what sources are they citing, and how often are our publications among them?”

rather than:

“How many times was our book cited by ChatGPT worldwide?”

What I'd build for your press

I'd create a prompt corpus around your subject areas and books, rather than simply tracking your press name.

For example:

• “What are the best introductions to [field]?”
• “What books should I read about [historical topic]?”
• “What are the major scholarly debates about [topic]?”
• “Which scholars/books have shaped research on [topic]?”
• “Compare the major interpretations of [topic].”
• “What primary sources should I consult for [topic]?”
• “[Book topic] bibliography for a graduate seminar”
• Questions derived from your table of contents, book descriptions, abstracts and index terms

Then track four things:

1. Your URLs cited — individual book/article pages, not just press.edu.
2. External scholarly URLs cited — university repositories, JSTOR, Project MUSE, journal sites, Wikipedia, reviews, etc.
3. Domains repeatedly winning citations — e.g. jstor.org, cambridge.org, oxfordacademic.com, particular university sites.
4. Citation gaps — prompts where competing books appear but yours doesn't.

Scrunch explicitly supports this kind of domain/URL/prompt analysis, including citation frequency and an “Influence Score” for prioritizing sources. scrunch.com Peec has also published large-scale analysis of 30 million sources across five major AI search platforms, which could be useful for understanding the broader scholarly-source ecosystem. peec.ai

My recommendation

If you're buying one tool: start with Scrunch and Profound demos.

I'd give Scrunch the edge for your particular use case because your question is fundamentally about source intelligence: Which domains and individual URLs are AI systems actually citing? Its workflow is explicitly built around that question. scrunch.com

I'd choose Profound instead if you expect this to become a substantial institutional program involving many subject areas, hundreds/thousands of titles, multiple AI engines, historical trend analysis, and executive reporting.

And if your press already has a substantial Semrush operation, I'd trial its AI Visibility Toolkit before adding another platform; it now combines AI citations with conventional search data and competitor research. www.semrush.com

One additional recommendation: don't abandon conventional scholarly-impact infrastructure. Pair AI-citation tracking with OpenAlex/Google Scholar/Scite/Dimensions/Web of Science-type citation data. AI citations measure discoverability and influence in AI-mediated research, while bibliographic citations measure something quite different. The two datasets could become a particularly interesting dashboard for a university press.

If you tell me roughly how many titles you publish (e.g. 100 vs. 2,000/year), your budget, and whether you need title/ISBN/DOI-level reporting, I can narrow this to a concrete 2–3 tool stack and suggest exactly what metrics I'd put on the press's dashboard.

If your goal is specifically “which of our books/papers/domains are actually being surfaced or cited by ChatGPT, Perplexity, Gemini, etc., so we can improve discoverability,” I would not buy a conventional bibliometrics product alone. You want an AI-search visibility tracker layered on top of traditional scholarly analytics.

My shortlist

The important distinction is that Scite and Altmetric aren't really substitutes for AI-citation monitoring. Scite's strength is the scholarly citation graph: its Smart Citations classify citation contexts as supporting, contrasting or mentioning, across a very large corpus. scite.ai Altmetric is broader online-attention monitoring rather than specifically “ChatGPT cited our paper.” www.altmetric.com

What I'd recommend for a university press

Start with Profound (or Peec/Scrunch) + Scite.

Use the AI-visibility platform to answer:

• When someone asks ChatGPT/Perplexity/Gemini questions in our subject areas, which of our URLs get cited?
• Which publisher domains, university repositories, journals, Wikipedia pages, reviews, etc. are competing with us?
• Which prompts produce citations to our books/papers?
• Which important queries produce zero representation from our catalog?
• Are citations going to the canonical book/article page, a Google Books/Internet Archive page, a review, a repository copy, or somewhere else?
• Which content changes increase citation frequency over time?

Then use Scite to answer a different question:

Is the scholarly ecosystem actually citing our publications, and how?

That's valuable because conventional citations and AI citations can diverge substantially.

One important caveat

Don't treat an “AI citation count” as equivalent to a scholarly citation count. Generative engines retrieve sources dynamically, and different engines can produce very different source sets. A 2026 study of generative-search citations found substantial differences between engines and argued that citation selection and actual influence on the generated answer should be measured separately. arxiv.org

That's particularly important for a university press. You probably want a dashboard with at least four metrics:

1. AI visibility
% of your tracked prompts where your press/domain appears.

2. AI citations
Number of times a specific URL/book/article is actually cited.

3. Citation share
Your citations ÷ all citations for the topic.

4. Scholarly citation impact
Traditional citations + Scite's supporting/contrasting/mentioning context.

I'd also track citation destination. For example:

Book X → 14 AI citations
Publisher page → 2
Google Books → 7
Library/repository → 3
Reviews → 2

That tells your marketing/web team something actionable: perhaps the book is discoverable, but the publisher's canonical page isn't the page the models trust.

For a university press, I'd prioritize these

Best enterprise strategy:
Profound + Scite + your existing Crossref/Google Scholar/Altmetric data

Best scholarly-first strategy:
Scite + a small internally built AI-query monitoring system

Best budget-conscious strategy:
OtterlyAI/Peec + Scite

And I would not rely on a generic SEO rank tracker. AI engines are not simply returning positions 1–10 from Google; recent research suggests that the source-selection process is sufficiently different that traditional SEO visibility isn't an adequate proxy for AI citation visibility. arxiv.org

If you tell me roughly how many titles/DOIs your press has (e.g. 500, 5,000, 50,000) and whether you're looking for a commercial dashboard or something your digital team could build, I can narrow this to the 2–3 strongest options and compare pricing, APIs, supported AI engines, DOI/ISBN-level tracking, and publisher-scale capabilities.

For a university press, I would treat this as AI visibility / AI citation intelligence rather than traditional SEO. The market is still immature: there is no equivalent of Web of Science for “papers cited in ChatGPT answers,” and different AI systems expose different citation behaviors. Studies of chatbot citation quality also show that citation behavior can be inconsistent, so measurement needs to be designed carefully. doi.org

The strongest options today:

My recommendation for a university press

I would build a hybrid system:

1. AI citation monitoring layer

Use a dedicated AI visibility platform (Profound / Peec AI / AthenaHQ) to track:

• Which prompts produce citations to your books, chapters, journal articles
• Which competing university presses appear
• Which domains AI systems prefer
• Citation share by subject area
• Changes over time

2. Scholarly metadata layer

Connect that to:

• DOI records
• Crossref metadata
• ORCID
• Google Scholar
• OpenAlex
• Dimensions / Scopus / Web of Science (if licensed)

Track:

• “AI citations” vs conventional citations
• Which books/papers are discoverable but under-cited
• Which metadata fields correlate with AI retrieval

3. Build your own “AI citation benchmark”

For a press, I would create a monthly test set:

Example prompts:

• “What are the best books on medieval trade?”
• “Who are leading scholars on climate justice?”
• “Recommend university press books about Indigenous archaeology.”
• “What are authoritative sources on constitutional law?”

Run these against:

• ChatGPT
• Claude
• Gemini
• Perplexity
• Microsoft Copilot

Record:

• Was your title cited?
• Was the publisher cited?
• Which URL/metadata page was used?
• Which competitors appeared?
• Was the citation correct?

This matters because AI systems often retrieve and cite sources differently, and academic citation generation can still contain substantial errors. academicworks.cuny.edu

4. Improve the things AI systems can actually retrieve

For a university press, the biggest wins are usually:

• Rich book landing pages (not just catalog PDFs)
• Structured metadata (schema.org, citation metadata)
• Full abstracts and tables of contents
• Author bios and institutional affiliations
• Stable URLs
• Open-access chapters where possible
• Machine-readable citation formats (BibTeX, RIS, CSL JSON)
• Clear subject taxonomy

If I were choosing a stack

Small/mid-size university press

• Semrush or Ahrefs (existing discoverability)
• OpenAlex + Crossref APIs
• A spreadsheet/database of benchmark prompts
• Monthly manual AI citation audits

Large university press

• Profound or Peec AI
• Data warehouse
• Crossref/OpenAlex integration
• Automated prompt testing
• Dashboard showing AI citation share by discipline

The metric I would care about most is not “how often does ChatGPT mention us?” but:

For the research questions where our books should be authoritative, what percentage of AI-generated answers cite our content compared with competing presses?

That is the closest analogue to market share for AI-mediated scholarly discovery.

Yes. For a university press, I’d look beyond ordinary SEO tools and focus specifically on AI citation/source tracking: which prompts produce citations, which individual URLs get cited, which competing domains appear, and how that changes over time.

My shortlist

Current comparisons show that the major platforms now cover ChatGPT, Perplexity and Google AI results, with some also covering Gemini and Claude. nboundmarketing.com

1. My first choice: Profound

Profound

I'd start here if your press wants a serious AI-discoverability intelligence program, rather than simply a dashboard saying "your brand was mentioned."

The important capability is tracking at the prompt → answer → citation/source level. For a press, you could build prompt sets such as:

• "best books on medieval Islamic history"
• "scholarly books about the history of jazz"
• "academic sources on climate migration"
• "who are the leading scholars on [topic]?"
• "[book title] reviews"
• "[author] books"
• "university press books about [subject]"

Then determine which of your books, author pages, reviews, JSTOR/Project MUSE pages, Wikipedia entries, competing presses, journals, etc. are actually being cited.

2. Particularly interesting for a publisher: Scrunch

Scrunch AI

Scrunch is worth a serious evaluation because it goes beyond "is my brand mentioned?" into citation-source dynamics and AI crawler behavior.

Its current research reports that 87.2% of AI citation events in its dataset came from third-party sources, versus only 5.95% from the brand's own site. scrunch.com

That is extremely relevant to academic publishing.

For example, you may discover that a university-press book isn't being cited because its own product page is poor; instead, the AI is getting information about it from:

• Wikipedia
• Google Books
• JSTOR
• Project MUSE
• scholarly reviews
• library catalogs
• journal articles
• author university pages
• other publishers
• newspaper/book-review coverage

That changes the strategic question from "How do we optimize our book page?" to "Which external scholarly ecosystems make this book legible to AI?"

3. Best lower-cost starting point: Peec AI

Peec AI

Peec is probably where I'd start a pilot if you don't want to commit enterprise money immediately. It focuses specifically on AI-search visibility, including mentions, citations, competitors and prompt-level monitoring. Independent 2026 comparisons put it among the stronger accessible options. nboundmarketing.com

One important caveat

Don't treat an AI citation count like a traditional citation count.

AI answers are stochastic: the same prompt can produce different sources at different times. Recent research specifically argues that AI citation visibility should be treated statistically rather than as a fixed number. arxiv.orgarxiv.org Another 2026 study found substantial differences between platforms and distinguishes between a page merely being retrieved/cited and actually influencing the generated answer. arxiv.org

So I'd want your system to report something like:

Book X
120 tracked scholarly prompts
ChatGPT: cited in 18% of runs
Perplexity: 31%
Gemini: 9%
Google AI: 14%
Most frequent competing sources: JSTOR, Wikipedia, Oxford UP, review journal Y
Most frequently cited URL: [specific page]
Citation trend: ↑ 7 percentage points over 90 days

rather than simply:

"AI visibility score: 74."

What I'd actually build for a university press

I'd use Profound or Scrunch + your existing scholarly metadata/analytics, with a controlled prompt corpus.

The prompt corpus is arguably more important than the software. Create perhaps 500–2,000 prompts spanning:

1. Book discovery — "best books about X"
2. Topic research — "scholarly sources on X"
3. Author discovery — "leading scholars of X"
4. Bibliographic queries — "books published about X since 2020"
5. Course queries — "books to teach X"
6. Graduate research — "recommended readings for X"
7. Book-specific queries — title, ISBN, author, subject
8. Citation-seeking queries — "academic sources I can cite for X"

Then track every cited domain and URL, not merely mentions of your press.

That distinction is crucial: research suggests different AI engines have surprisingly different source pools, so optimizing solely for "ChatGPT visibility" can give you a misleading picture of overall discoverability. arxiv.orgarxiv.org

If I were buying today:
Profound for the main enterprise platform → Scrunch as the strongest alternative/companion for publisher-oriented source intelligence → Peec for a cost-conscious pilot.

If you tell me roughly how many titles your press publishes (e.g. 500, 5,000, 50,000+) and your annual software budget, I can narrow this to a 2–3 tool procurement shortlist and specify exactly what I'd measure.

Yes. For a university press, I’d approach this differently from ordinary “AI visibility”/AEO software: you want paper/book-level citation tracking, exact source URLs, domain-level patterns, and repeatable measurement across academic discovery prompts.

My shortlist

If I were evaluating just two, I'd start with Citations.io and Orbilo. Citations.io explicitly tracks ChatGPT, Perplexity, Gemini and Claude and exposes the sources being cited, while Orbilo is particularly focused on the exact URLs winning citations for particular prompts. citations.io

The important distinction

There are actually three different measurements you may want:

1. “Does an AI answer mention our book/paper/author?”
2. “Does it actually link/cite our press's page or the paper itself?”
3. “Which outside domains does AI rely on when discussing our subject?”

The third is particularly valuable for a press. You could discover, for example, that when users ask about Topic X, ChatGPT consistently cites Wikipedia and a particular university repository, while Perplexity cites a handful of journals and specialist websites. That tells you where your metadata, excerpts, reviews, and author profiles need to be discoverable.

And don't assume one AI platform represents the others. Recent research finds substantial instability in AI citation results, meaning a single query/run can give a misleading impression of visibility. arxiv.org

What I'd build for a university press

Rather than simply tracking the press's domain, create a “scholarly AI discoverability” benchmark:

• 100–500 questions corresponding to your strongest subject areas
• Queries such as “What are the best books on [topic]?”, “Who are the leading scholars on [topic]?”, and specific research questions
• Run them repeatedly across ChatGPT, Perplexity, Gemini and Claude
• Record:
  • cited paper/book
  • cited URL
  • cited domain
  • author
  • publisher
  • citation position
  • whether your title was mentioned but not cited
  • competing publishers/domains appearing instead
• Aggregate by title, discipline, author, domain and AI platform
• Track changes monthly rather than treating one snapshot as definitive

That last point matters: AI citation distributions can vary considerably between repeated queries. arxiv.org

One tool I'd add: scite

scite is worth having alongside an AI-visibility tracker. Its job is fundamentally different: it analyzes scholarly citations and their context—whether subsequent work supports, contrasts with, or merely mentions a paper. That can give your editors a much richer picture of scholarly impact, while the AI-visibility tools measure AI-mediated discoverability.

So I'd think of the stack as:

Scite / scholarly indexes → scholarly impact
Citations.io / Orbilo → AI discoverability
Your own prompt corpus → editorial intelligence

One caution: don't equate “AI cited it” with scholarly importance. Research on AI bibliographic retrieval has found substantial reference errors and even fabricated references in some systems. arxiv.org

If your goal is specifically “Which papers/books/domains are ChatGPT and Perplexity citing when answering academic questions, and how does our catalog compare?”, I can also lay out a vendor-by-vendor comparison of 5–8 tools, including approximate pricing, API/export capabilities, academic suitability, and whether they can track individual URLs/DOIs rather than just brands.

For a university press, I’d separate this into two jobs:

1. Measure actual AI retrieval/citation of your pages — which papers/books/pages are being fetched or cited.
2. Measure synthetic visibility — which prompts cause ChatGPT, Perplexity, Gemini, Claude, etc. to cite your content or competitors.

There isn’t one perfect scholarly-publishing product yet, but a few are particularly relevant.

My shortlist

For your particular use case, I'd start with ScalePost + Citare

The distinction matters.

ScalePost can answer something close to:

"Which of our 40,000 scholarly URLs are AI systems actually requesting?"

That's valuable because a publisher has an enormous corpus. You don't want to manually construct thousands of prompts just to discover that AI agents are repeatedly retrieving The Oxford Handbook of X, article Y, and three pages from your journal platform.

Its publisher positioning is unusually relevant: it claims to observe AI crawlers at the CDN level and provide URL-level, first-party measurements rather than estimates based on asking an AI model questions. www.scalepost.ai

Citare, meanwhile, can answer:

"When someone asks an AI system the kinds of questions our readers ask, which scholarly sources does it cite?"

That lets you build a citation-share-of-voice dataset: your press vs. university repositories vs. JSTOR vs. competing presses vs. individual scholars, etc. Its methodology explicitly includes query design, citation parsing and competitor benchmarking. www.citare.ai

One important caveat

Don't equate "AI fetched my paper" with "AI cited my paper."

Those are different signals. A system can retrieve a page, use its contents internally, and not expose that URL in the final answer. Conversely, a citation can appear without the publisher's name being prominently mentioned. Recent research on AI citations has highlighted this "ghost citation" problem. searchengineland.com

For a university press, I'd therefore build a dashboard with at least:

• AI crawler fetches by URL
• AI citations by URL
• Citation rate by AI engine
• Citation rate by discipline
• Citation rate by book/article/journal
• Share of citations vs. competing presses
• Queries producing citations
• Queries where competitors are cited but you aren't
• Newly cited / lost citations
• AI referral traffic
• AI-cited pages that subsequently generate clicks
• Citation persistence over time

And crucially, track the DOI/article/book URL, not merely your domain. A press-wide "AI visibility score" is much less useful than discovering:

This specific 2023 article is repeatedly cited for questions about climate migration, while our newer 2025 article on the same subject is never retrieved.

That gives your editorial, metadata, SEO and marketing teams something actionable.

One more thing I'd investigate

Because you're an academic publisher, I'd also want OpenAlex/Crossref/Google Scholar-style scholarly metadata alongside the AI data. There is already research suggesting that conventional scholarly prominence can influence which papers LLMs surface; one study of environmental-science queries found a strong tendency toward highly cited, established papers. arxiv.org

So the really interesting analysis isn't simply:

"What does ChatGPT cite?"

It's:

"What predicts an academic work being cited by AI?"

For example:

citations + publication age + journal + DOI metadata + open-access status + backlinks + full-text availability + abstract quality + discipline → probability of AI citation

That could become a genuinely useful AI discoverability program for the press, rather than just another marketing dashboard.

If you tell me roughly how many journals/books/articles your press has and whether you control the web/CDN logs, I can recommend a concrete stack (including lower-cost/open-source options) and a measurement schema I'd use for a university press.

If your goal is helping a university press understand and improve how its books, journals, and papers are surfaced and cited by ChatGPT, Claude, Gemini, Perplexity, and similar systems, there are now a handful of purpose-built platforms. The market is still young, so most publishers combine one AI-visibility platform with traditional bibliometrics.

Here's how I'd evaluate them:

These products repeatedly query multiple LLMs, record:

• whether your domain appears as a cited source,
• which pages get cited,
• which competitors are cited instead,
• which prompts trigger citations,
• changes over time.

This is different from conventional citation indexing—they're measuring LLM citation frequency, not scholarly citations. www.tryprofound.com

For the academic content itself

I'd also pair AI monitoring with scholarly citation tools:

• Semantic Scholar
• OpenAlex
• Crossref
• Scopus
• Web of Science

These answer different questions:

• Which papers are influential?
• Which journals cite us?
• Which topics are growing?
• Which metadata is missing?
• Which DOIs are incomplete?

They do not tell you whether ChatGPT is citing your work.

If you publish journals

The workflow I'd recommend is:

1. Monitor AI citation share with Profound or Lighthouse.
2. Track scholarly impact with OpenAlex + Crossref + Scopus/Web of Science.
3. Build dashboards combining:
   • AI citations
   • Google Scholar citations
   • downloads
   • Altmetric attention
   • DOI resolution
   • referral traffic from AI assistants (where available)

If you're technically inclined

Many university presses are beginning to build their own monitoring systems.

Typical architecture:

• A library of 500–5,000 prompts covering each discipline you publish
• Scheduled queries to ChatGPT, Perplexity, Gemini, Claude, etc.
• Extraction of:
  • cited domains
  • cited papers
  • cited authors
  • missing competitors
  • answer quality
• Time-series dashboards (Looker, Tableau, Power BI)

This gives much richer insights than vendor dashboards because you can tailor prompts to your catalog. One important finding from early experiments is that LLM citation patterns are somewhat variable between repeated runs, so sampling prompts multiple times produces more reliable trends. www.reddit.com

For a university press, my recommendation

If budget allows:

• Best commercial choice: urlProfoundturn0search3
• Good lower-cost alternative: urlLighthouseturn0search5
• Scholarly metadata backbone: OpenAlex + Crossref
• Traditional citation benchmarking: Scopus or Web of Science

That combination lets you measure both academic impact and AI discoverability, which are increasingly distinct metrics.

If your goal is helping a university press understand and improve how its books, journals, and papers are cited by ChatGPT, Perplexity, Gemini, Claude, and similar systems, the market is still early. No platform provides complete visibility into every AI response, but several are becoming useful for "AI discoverability" (often called Generative Engine Optimization (GEO) or Answer Engine Optimization (AEO)).

Here's how I'd rank them.

1. Profound

Probably the most mature enterprise platform.

It measures things like:

• Which domains ChatGPT cites
• Share of voice across AI assistants
• Competitor citations
• Prompt-level monitoring
• Changes over time

For a university press, you could monitor prompts such as:

• "best books on constitutional law"
• "recent archaeology research"
• "recommended papers on climate adaptation"

and see whether your journals appear.

Best if your press has dozens or hundreds of titles.

2. AthenaHQ

Strong for organizations that care about:

• citation frequency
• competitive benchmarking
• AI visibility trends
• executive dashboards

Many publishers use similar monitoring approaches because traditional SEO tools don't capture AI answer engines.

3. Peec AI

Less enterprise-oriented.

Useful if you want:

• recurring prompt monitoring
• citation tracking
• visibility scores
• easier setup

A good option for a university press marketing team.

4. Otterly.AI

Focused on repeated testing of prompts across AI systems.

Useful for questions like:

• Does ChatGPT cite our journal?
• Does Gemini recommend our handbook?
• Has visibility improved since last month?

What no software currently does well

This is important.

No commercial product can reliably answer:

"How often does ChatGPT cite paper X globally?"

because ChatGPT responses are personalized, model versions change, browsing behavior varies, and OpenAI does not expose aggregate citation analytics.

Instead, these products repeatedly test large prompt sets and estimate visibility. Research on GEO likewise relies on controlled prompt sampling rather than direct platform data. papers.ssrn.com

For an academic publisher, I'd build a custom system

Many university presses would benefit more from a bespoke pipeline than an off-the-shelf marketing dashboard.

Example workflow:

• Maintain 5,000–20,000 representative prompts by discipline.
• Query ChatGPT, Perplexity, Gemini, Claude, and Google AI.
• Extract:
  • cited domains
  • cited DOIs
  • journal titles
  • publisher names
  • books
• Store results in a database.
• Generate dashboards showing:
  • citation share by imprint
  • citation share by journal
  • trends over time
  • competitor comparisons
  • prompts where your content never appears
  • prompts where your content displaced competitors

This produces insights that generic GEO tools typically don't offer.

Also monitor infrastructure signals

Academic discoverability increasingly depends on whether AI systems can easily consume your content. In addition to citation tracking, monitor:

• DOI resolution success
• schema.org metadata
• Crossref completeness
• abstracts available without login
• machine-readable references
• ORCID coverage
• persistent URLs
• robots.txt and AI crawler policies
• licensing clarity

These factors make content easier for retrieval systems to identify and cite, although they don't guarantee inclusion. Emerging research suggests that clear structure, explicit answers, and well-supported content are associated with higher AI citation rates. papers.ssrn.com

My recommendation for a university press

• Small press (under 200 titles): Peec AI or Otterly.AI plus a custom prompt-monitoring workflow.
• Large university press: Profound or AthenaHQ, supplemented with an internal analytics pipeline.
• Research-intensive university or library consortium: Build a custom GEO observatory that tracks AI citations across major assistants by DOI, journal, subject area, and publisher. This provides much richer evidence than today's commercial dashboards and can directly inform editorial and discoverability strategies.

For a university press, I would not look for a single “AI citation tracker.” The market is still emerging, and the best setup is usually a combination of:

1. AI answer visibility tracking — “Does ChatGPT/Perplexity/Gemini mention our books, authors, journals, or domains?”
2. Scholarly graph/citation analytics — “Are our works connected to the knowledge graph these systems may draw from?”
3. Content discoverability tooling — “Are our metadata, abstracts, landing pages, and structured data easy for AI systems to retrieve?”

A practical shortlist:

My recommended stack for a university press

For a small-to-mid press

• OpenAlex → build a catalog-level scholarly visibility database. developers.openalex.org
• Peec AI or OtterlyAI → track whether AI assistants surface your titles/domains.
• Google Search Console + your web analytics → measure whether AI-driven discovery turns into visits.

For a large university press

• OpenAlex API + Dimensions/Altmetric for scholarly intelligence.
• Profound or Semrush for AI search monitoring.
• Internal dashboard combining:
  • ISBN/title metadata
  • DOI records
  • author authority signals
  • AI citations
  • referral traffic from AI systems

Metrics I would track specifically for a press

• AI citation share: % of sampled research prompts where your books/articles appear.
• Citation quality: whether AI cites the actual publisher page, DOI, repository copy, or a secondary source.
• Competitor share: which university presses appear instead.
• Metadata gaps: books that are highly cited academically but invisible in AI answers.
• Author/entity recognition: whether AI systems correctly associate authors with your press.
• Domain citation frequency: how often your press domain appears as a source.

One caveat: AI citation measurement is noisy because models can produce different sources for the same prompt, so avoid treating a single snapshot as a definitive ranking. Repeated sampling and trend analysis are more meaningful. arxiv.org

If I were advising a university press starting this program, I would probably pilot OpenAlex + an AI visibility tracker for 100–200 high-value titles, then expand once you know which disciplines and content types are being surfaced.

For a university press, there are really two different problems:

1. Measuring when ChatGPT, Gemini, Claude, Perplexity, etc. cite your books, journals, or website.
2. Understanding why they choose those sources so you can improve discoverability.

I'd recommend a combination rather than a single tool.