ChatGPT answers, unedited

Yes. There’s a real category of platforms built around automating the creative-testing → audience-testing → winner-scaling loop, although the degree of automation varies.

The strongest options

• marpipe.com — probably the closest match if your priority is rigorous creative experimentation at scale. It uses multivariate testing: you can define variables like hook, image/video, headline, CTA, etc., automatically generate combinations, launch them, and measure which elements drive performance. It also has statistical-confidence tooling. Marpipe Marpipe

Best for: sophisticated growth teams, ecommerce, large creative volumes.

• madgicx.com — broader Meta ad automation. It can generate/launch creative variations, analyze creative performance by audience, automate budget/stop-loss rules, and launch winning creatives against different audiences. Madgicx Academy Madgicx

Best for: wanting one system to handle testing and ongoing Meta campaign optimization.

• smartly.io — worth looking at if you're operating at enterprise scale across multiple social networks. It's more of an end-to-end creative/media platform than a pure experimentation tool.

The important distinction

There are actually two different products hiding inside your question:

1. "Which finished ad wins?"

Creative A vs. B vs. C → automatically allocate spend → identify winner.

2. "Which creative ingredients win?"

Hook × visual × headline × CTA × offer × audience → determine which components drive incremental performance.

The second is much more interesting if you're trying to build a creative experimentation engine. Marpipe is particularly strong here because it treats creative as modular variables rather than simply a pile of finished ads. Marpipe

For example, you could have:

Instead of manually creating and monitoring hundreds of combinations, the platform manages the experiment and tells you something closer to:

"Problem-based hooks outperform benefit hooks by 31% for cold audiences, but benefit hooks win for 1% lookalikes."

That's considerably more valuable than simply telling you "Ad #47 won."

If you're thinking about building something in this space

There is still an interesting gap: a genuinely autonomous experimentation layer that sits above Meta/TikTok/Google and continuously:

1. Generates creative variants.
2. Constructs controlled experiments.
3. Tests creative × audience combinations.
4. Detects statistical/causal winners.
5. Automatically kills losers.
6. Moves budget toward winners.
7. Feeds the winning attributes back into creative generation.
8. Starts the next experiment automatically.

Madgicx is moving toward this workflow, while Marpipe is particularly focused on the structured experimentation side. Madgicx Academy Marpipe

If you're asking because you're evaluating this as a product/startup opportunity, I can also map out the existing players and identify where the biggest whitespace is—especially around autonomous creative × audience experimentation.

Instead of manually creating and monitoring hundreds of combinations, the platform manages the experiment and tells you something closer to:

"Problem-based hooks outperform benefit hooks by 31% for cold audiences, but benefit hooks win for 1% lookalikes."

That's considerably more valuable than simply telling you "Ad #47 won."

If you're thinking about building something in this space

There is still an interesting gap: a genuinely autonomous experimentation layer that sits above Meta/TikTok/Google and continuously:

1. Generates creative variants.
2. Constructs controlled experiments.
3. Tests creative × audience combinations.
4. Detects statistical/causal winners.
5. Automatically kills losers.
6. Moves budget toward winners.
7. Feeds the winning attributes back into creative generation.
8. Starts the next experiment automatically.

Madgicx is moving toward this workflow, while Marpipe is particularly focused on the structured experimentation side.

Yes. There are several platforms now that get pretty close to “continuously generate → test → identify winners → scale” for social ads.

The strongest options

Madgicx is probably the closest match if you're specifically thinking about a system that can systematically test creative × audience combinations. Its Audience Studio can identify profitable audiences, create/intersect lookalikes and interests, and launch combinations; its Creative Insights analyzes formats and creative elements. academy.madgicx.com

It also explicitly supports a workflow of test creatives → identify winners → put winners against new audiences → scale the winning combinations. academy.madgicx.com

Smartly is the more enterprise-grade answer if you're running substantial spend across multiple networks. It combines creative production, campaign management, DCO, audience personalization, and automated optimization. It supports Meta, TikTok, Pinterest, Snapchat, Google and several other channels. docs.smartly.io

The interesting distinction is:

• Madgicx: “Find what works in my Meta account and automate the buying/testing.”
• Smartly: “Run a large-scale creative + media experimentation machine across channels.”
• Motion: “Tell me why my creatives are winning/losing and help my creative team iterate.”
• Meta itself: Cheapest/native infrastructure, but you still have to build much of the experimentation framework yourself.

If you're imagining a more sophisticated system

The workflow I'd want is something like:

100 creative variants × 10 audience hypotheses → controlled test → statistical/ROAS evaluation → kill losers → promote winners → generate new variants from winners → retest against new audiences → continuously learn.

That's a little different from ordinary “A/B testing.” It's closer to a creative experimentation engine or multi-armed-bandit-style optimization system.

If that's what you're actually looking for, I can also map out the 5–10 platforms that specifically do this, including which ones automate the generation, experiment design, budget allocation, and winner promotion pieces.

Yes. There are several platforms built specifically to automate paid social creative testing, audience testing, and scaling winning combinations. The best fit depends on whether you need more creative generation, media experimentation, or full campaign automation.

Here are some strong options:

A few newer AI-native tools are also targeting this exact workflow:

• Ad Flywheel — focuses on taking winning Meta creatives, generating variants, testing them, and automatically finding stronger performers. AdFlywheel
• AdFabric — combines creative generation, A/B testing, analytics, and multi-channel campaign workflows. AdFabric
• AdMove — focuses on generating many video/social creative variants for testing across Meta, TikTok, and YouTube. Admove

What a mature automated testing stack usually looks like

For a high-volume performance team, the workflow often looks like:

1. Generate variants

• 10–100 creative variations
• Different hooks, thumbnails, headlines, offers, formats
• Create experiment matrix

• Creative A × Audience 1
• Creative A × Audience 2
• Creative B × Audience 1
• etc.
• Launch automatically

• Push to Meta/TikTok/Google Ads
• Allocate test budgets
• Detect winners

• CPA/ROAS/CVR/CTR thresholds
• Statistical confidence checks
• Scale winners

• Increase spend
• Create derivative creatives
• Refresh fatigued ads

My shortlist by use case

• DTC/ecommerce spending $10k–$1M+/month: Motion + Marpipe + creative generation tools
• Agency managing many client accounts: Smartly.io or Metadata
• Meta-heavy performance marketing: AdStellar AI, Ad Flywheel, Motion
• Need AI to produce hundreds of creative variants: AdCreative.ai, AdMove, AdStyle

If you mean "hands-off AI agent that continuously creates ads, tests audiences, kills losers, and scales winners", that is still an emerging category; most platforms today automate parts of the loop rather than the entire growth engine.

Yes. There are platforms specifically built to automate creative × audience testing rather than making you manually duplicate campaigns in Ads Manager.

The strongest options I'd look at:

1. Madgicx — probably closest to what you're describing

Madgicx lets you create/duplicate creative variations, launch them directly into Meta, monitor creative-level performance, and automate rules that shut down losers and scale winners. It also has an Audience Launcher for testing different audiences and lookalikes. madgicx.com

So you could conceptually run:

10 hooks × 5 visual concepts × 3 offers × 4 audiences

…and have the system help identify which combinations are producing the best results rather than manually inspecting every ad set.

Its main limitation is that it's very Meta-centric. It does have integrations/reporting beyond Meta, but its deepest automation is around Meta advertising. academy.madgicx.com

2. Smartly — better if you're talking serious scale

Smartly is more of an enterprise advertising operating system. It combines creative generation/variation, campaign management, audience personalization, testing, and optimization across multiple social channels. www.smartly.io

Its Dynamic Creative Optimization can automatically serve different creative variations to different audience segments, while its media tooling handles campaign workflows and optimization. www.smartly.io

This is much closer to:

Generate → launch → test → identify winning creative/audience combinations → automatically rotate/scale → detect fatigue → generate new variants

rather than simply "A/B testing software."

The interesting gap

If you're asking because you're considering building something in this space, there's actually an important distinction:

Most existing platforms optimize ads/campaigns, but the ideal system would optimize the experiment itself.

For example:

Creative dimensions

• Hook
• Visual
• UGC creator
• Offer
• CTA
• Format
• Copy
• Landing page

Audience dimensions

• Broad
• Lookalike %
• Interest
• Retargeting window
• Demographic
• Geographic segment

Then an experimentation engine could automatically determine:

"Hook #7 works particularly well with UGC creator #3 among 1–3% LAL audiences, but not broad audiences. Allocate the next $2,000 there and generate 8 mutations of Hook #7."

That's more sophisticated than conventional A/B testing because you're effectively doing automated creative/audience search under a budget constraint.

If that's what you're getting at, I can also map out the current landscape of companies doing this + what each one is missing, which would be useful if you're evaluating the opportunity.

Yes. There are a few platforms in this space, but they differ quite a bit in what they automate.

Best fits

Marpipe is probably closest to what you're describing. It lets you define variables—e.g. hook × image × headline × CTA—and programmatically generate combinations, launch them as structured tests, and analyze performance down to individual creative elements. Its testing methodology is specifically designed to prevent Meta's optimization from simply funneling budget toward one variant. www.marpipe.com

For example, you could structure:

5 hooks × 4 visuals × 3 offers × 3 audiences
→ 180 combinations
→ automated deployment
→ conversion/CPA analysis
→ identify winning combinations/elements
→ feed winners into the next test

Motion is slightly different

Motion is particularly strong if your problem is "We have hundreds of ads running—tell me what's working and what we should make next." It automatically analyzes creative, tags things like hooks, messaging angles, visual formats and intended audiences, and lets you compare creative performance across Meta and TikTok. help.motionapp.com

It also explicitly analyzes which audiences are responding to which creatives, so you can uncover things like:

UGC + problem/solution hook → strongest with 25–34
Founder-led video + price messaging → strongest with retargeting
Product demo → strongest with broad prospecting

But I'd think of Motion more as creative intelligence/optimization than a fully automated experimental engine.

Smartly is the enterprise option

Smartly is interesting if you want the entire pipeline—dynamic creative generation, audience/persona-specific variants, media buying and optimization—under one platform. For example, it can dynamically combine video components to create many versions targeted to different demographics/personas. docs.smartly.io

The important distinction

If by "automate A/B testing" you mean:

Upload 20 creatives + 10 audiences → platform intelligently runs experiments → reallocates spend → identifies winners → generates the next round → repeats

…then there still isn't a universally dominant "self-driving experimentation" platform that does all of that exceptionally well across Meta/TikTok.

Most products fall into one of three buckets:

1. Experimentation: Marpipe
2. Creative analytics/intelligence: Motion
3. Creative + media automation: Smartly

And Meta itself increasingly makes traditional A/B testing less clean because its algorithms optimize delivery dynamically. Marpipe's approach of deliberately structuring tests with controlled budgets is specifically designed around that problem. www.marpipe.com

If you're asking because you're thinking about building a product in this space, there's actually an interesting gap here: an autonomous creative experimentation layer that sits on top of Meta/TikTok, continuously decides what to test next, creates the variants, launches them, evaluates statistical/causal evidence, and feeds the result back into the next generation. That is substantially more ambitious than today's creative analytics tools.

Yes. There are platforms now that get fairly close to “always-on A/B testing” for paid social, although they differ in how much of the testing loop they automate.

The strongest options

Madgicx is probably closest to what you're describing if you want the platform to actually launch and manage lots of variations. It can create multiple creative variations, launch/test audiences, and automate campaign management. Its Audience Launcher is specifically designed to generate and test different audience segments at scale. academy.madgicx.com

Motion is particularly interesting if your bottleneck is creative iteration. It analyzes creative performance across Meta and TikTok, breaks down winning visuals/hooks/formats, and identifies what to make next. It can test multiple creative elements simultaneously, but it's more of a creative intelligence/optimization layer than a fully autonomous media-buying engine. motionapp.com

Smartly is the enterprise-grade option. It combines creative production/variation with campaign management and optimization, including dynamically generating many creative combinations for different audiences. docs.smartly.io

And TikTok itself now has surprisingly sophisticated native split testing: you can test targeting, creative, creative assets, placement, bidding, budget strategy, and more, with audience groups separated to reduce interference between variants. ads.us.tiktok.com

The distinction I'd pay attention to

If what you mean is:

“Give the system 20 creatives × 10 audiences, automatically run the experiments, identify statistically meaningful winners, kill losers, generate new variants from winners, and continuously repeat the process.”

That's a more specific product category, and there isn't one universally dominant platform that perfectly does that end-to-end across Meta + TikTok.

The stack today tends to look more like:

Creative generation → experiment orchestration → performance analysis → budget optimization → new creative generation

with different vendors owning different pieces.

If you're evaluating this because you're considering building/buying something like this, I can also map out the current competitive landscape and show you which parts are already commoditized vs. where there is still a real product gap.

Yes. There are several platforms now that go beyond basic A/B testing and automate the creative × audience × budget experimentation loop.

The strongest options

Smartly is probably the closest match if by "at scale" you mean thousands of creative variations across multiple ad networks. It combines creative production, campaign management, testing, and optimization. Its platform can generate/personalize creative variants, launch campaigns in bulk, analyze creative elements, and automatically rotate/optimize campaigns. www.smartly.io

Smartly advertising platform

Hunch is particularly interesting if your bottleneck is creative experimentation. It can generate thousands of personalized variations from templates, launch/manage campaigns across Meta, TikTok and Snapchat, and automate campaign workflows. www.hunchads.com

Hunch

Madgicx is more appealing if you're primarily running Meta and want a comparatively accessible automation layer. It has automated rules for killing/ scaling ads, creative analytics, creative generation, and an Audience Studio that can identify promising interests/lookalikes and launch audience combinations. madgicx.com

Madgicx

One important distinction

There are really two different products hiding inside your question:

1. Automated optimization

"Here are 50 creatives and 10 audiences. Spend money intelligently and shift budget toward winners."

2. True experimentation

"Run controlled tests where creative A vs. B, or audience X vs. Y, are isolated enough that I can confidently conclude why one won."

The first is much more common. The second is harder.

For example, TikTok's current Split Testing explicitly supports testing targeting, creative, bidding, placement, budget strategy, catalog, and custom combinations, with audience groups separated to reduce contamination. ads.us.tiktok.com Meanwhile, TikTok Smart+ automatically explores creatives and audiences and shifts delivery toward predicted winners. ads.tiktok.com

So if you're imagining a system like:

Generate 100 creative variants → test them across 20 audience hypotheses → identify winning creative/audience combinations → automatically kill losers → generate new variants based on the winners → repeat

Yes, that category exists, but most platforms don't completely automate that closed-loop scientific experimentation process. Smartly/Hunch get closest on the enterprise creative-operations side; Madgicx is attractive for Meta-centric automation; and the ad platforms themselves increasingly provide the optimization engine.

If you're thinking about building a product in this space, there's actually a fairly interesting gap between "automated ad optimization" and true autonomous experimentation—especially around automatically forming hypotheses, controlling test contamination, statistical confidence, and feeding learnings back into creative generation.

Yes. There are a few platforms built specifically around this, although they differ in how much of the testing loop they automate.

Best fits

Madgicx is probably closest to what you're describing. Its workflow can launch multiple creative variations, test them, analyze performance, and then take winning creatives into different audiences. Its Audience Studio specifically supports creating/testing combinations of interests and lookalikes and launching them from the platform. academy.madgicx.com

Smartly is the more enterprise-grade version. It combines creative generation/variation, personalization, campaign activation, testing, and optimization across multiple social channels. It can generate large numbers of creative variations and connect them to audience personalization and automated workflows. www.smartly.io

Motion is different: it's excellent if your problem is "We have 500 ads running—tell me which hooks, visuals, formats, and concepts are actually winning." It automatically groups creatives and analyzes performance across Meta, TikTok, YouTube and LinkedIn, but it's less of an end-to-end audience experimentation engine. motionapp.com

If you're imagining a system like this

Creative A/B engine
→ generate 20 hooks × 5 visuals × 3 CTAs
→ automatically launch variations
→ allocate enough spend to establish signal
→ kill losers
→ promote winners

Audience engine
→ winner Creative A × Broad
→ winner Creative A × LAL 1%
→ winner Creative A × LAL 3%
→ winner Creative A × Interest cluster B
→ automatically identify the best creative × audience combination
→ scale it

That's much closer to Madgicx/Smartly than a normal analytics platform.

One important caveat: on Meta in particular, "A/B testing" isn't always a clean scientific split test. Meta's delivery algorithm can redistribute spend toward variants it predicts will perform better, so some platforms give you automated optimization rather than a statistically controlled experiment. Madgicx itself acknowledges this distinction with Meta's DCO/Multi-Text Options. academy.madgicx.com

If you're asking because you're evaluating/building a product in this space, I can also map out the current landscape and identify what's still missing / where there may be a SaaS opportunity.

Yes. There are several platforms now that get close to “automated experimentation for paid social”, but they differ substantially in whether they optimize creative, audiences, or the whole campaign.

The strongest options

Smartly is probably the closest match to what you're describing. It combines creative generation/variation, campaign management, audience targeting, automated workflows, optimization, and reporting across Meta, TikTok, Pinterest, Snapchat and other channels. Its DCO capabilities can personalize creative based on audience/data signals and generate large numbers of variations. docs.smartly.io

Marpipe is more interesting if the core problem is experimentation. It can take modular elements—images, headlines, CTAs, layouts, etc.—generate combinations, launch structured tests, and identify which individual creative variables are actually driving performance. www.marpipe.com

Motion is more of a creative intelligence/testing layer. It pulls Meta/TikTok performance data, analyzes creatives and their elements, and surfaces patterns such as which hooks, visuals, formats, demographics and retention points are working. help.motionapp.com

The interesting gap

If what you're imagining is something like:

Generate 50 creative variants × 10 audience hypotheses → automatically launch experiments → allocate spend → detect winners → kill losers → generate the next experiment → repeat

that's a more specific product category than ordinary “creative optimization.”

Most platforms today automate parts of that loop:

Creative generation → testing → analysis → optimization

but fewer provide a genuinely automated creative × audience experimentation engine where the system treats both as experimental variables and continuously designs the next test.

For example, you could have:

• Creative variables: hook, opening frame, UGC creator, offer, headline, CTA, format
• Audience variables: broad, lookalike, interest cluster, age, gender, geography, first-party segment
• Business outcome: CAC, ROAS, qualified leads, LTV
• Experiment engine: factorial testing / Bayesian allocation / sequential testing
• Automation: launch → monitor → reallocate → graduate winner → create next test

That last category is where I'd look if you're asking because you're evaluating whether such a platform exists vs. thinking about building one.

If you tell me which channels you're primarily running (Meta, TikTok, Google, LinkedIn) and roughly how much you're spending per month, I can narrow this down to the 3–5 platforms that actually fit your use case, including pricing and how automated their experimentation really is.

Yes. There are several platforms that automate parts (or most) of the workflow of creating ad creative variants, launching A/B tests, analyzing winners, and scaling winning combinations of creative × audience. The best fit depends on whether you want a media buying optimizer, a creative testing engine, or a full creative production + campaign automation system.

Some notable options:

A typical scalable setup looks like:

1. Creative generation
   • Generate 20–200 variants:
     • hooks
     • first 3 seconds of video
     • thumbnails
     • headlines
     • offers
     • testimonials
     • creator styles

2. Experiment design
   • Test dimensions separately:
     • Creative A vs B
     • Audience 1 vs 2
     • Placement
     • Copy angle
     • Landing page

3. Automated allocation
   • Shift spend toward winners
   • Pause underperformers
   • Refresh fatigued creatives

4. Learning layer
   • Identify patterns:
     • “UGC + problem/solution hook wins for cold audiences”
     • “Product demo beats lifestyle images”
     • “Audience X responds to benefit Y”

For a performance marketing team spending seriously on Meta/TikTok, I’d usually separate the stack:

• Creative intelligence/testing: AdStyle, AdCreative.ai, Adobe GenStudio-type tools
• Campaign automation/optimization: MediaScale, AdStellar AI, native Meta automation
• Enterprise workflow: Adobe GenStudio or similar DAM + campaign systems

One thing to watch: many “AI ad testing” tools generate lots of variants, but fewer actually solve the harder problem—controlled experiments that isolate why a creative wins (e.g., changing only the hook while keeping everything else constant). If you are trying to build a repeatable growth engine, prioritize platforms that track creative attributes and audience segments, not just generate more ads. www.adstyle.co

If you mean $50k+/month Meta spend, DTC/ecommerce, SaaS lead gen, or agency client management, the shortlist changes quite a bit.

Yes. There are several platforms built specifically to automate paid social creative testing, audience testing, and scaling winning combinations. The right choice depends on whether you want more of a media-buying optimizer, creative testing engine, or AI creative factory.

Here are some of the stronger options:

A typical high-scale workflow these tools enable:

1. Upload 5–20 base creatives
2. Generate variations:
   • hooks
   • first 3 seconds of video
   • headlines
   • thumbnails
   • CTAs
   • UGC styles
   • models/scenes
3. Create audience cells:
   • broad
   • lookalikes
   • interest clusters
   • demographic segments
4. Launch hundreds of combinations
5. Automatically shift budget toward winners
6. Feed learnings back into the next creative batch

A few distinctions matter:

• If you spend <$10k/month on ads: native Meta/TikTok testing + a creative generator may be enough.
• If you spend $10k–$100k+/month: look at Marpipe, AdStyle, AdStellar, MediaScale-type tools.
• If you spend $500k+/month across many brands: enterprise platforms like Smartly.io or Adobe-style creative operations tools become more relevant.

One thing to watch: many "AI ad testing" tools mostly generate more creatives; fewer actually run statistically clean experiments and separate creative effects from audience effects. If your goal is finding why an ad wins (e.g., "testimonial hook beats product demo for women 35–44"), prioritize platforms that support structured experiments rather than just automated optimization. www.adstyle.co

If you tell me your ad spend level, channels (Meta/TikTok/YouTube/etc.), and whether you’re DTC, SaaS, or lead gen, I can narrow this to the best-fit 3–5 platforms.

Yes. There are platforms that get pretty close to “continuously run creative × audience experiments, identify winners, and scale them automatically.”

The strongest options I’d look at are:

1. Madgicx — probably closest to what you're describing

Madgicx lets you launch multiple creative variations, analyze which creative elements perform, and then test winning creatives across different audiences. Its Audience Studio can identify high-performing audiences, suggest interests, construct lookalikes/intersections, and launch audience combinations. academy.madgicx.com

So you can conceptually run:

Creative A/B/C/D
×
Audience 1/2/3/4
→ collect performance
→ identify winning creative/audience combinations
→ scale winners
→ introduce new variations

It also has automated ad launch/management and creative-refresh functionality. madgicx.com

Madgicx

2. Smartly — better if you're operating at serious scale

Smartly is more of an enterprise advertising operating system. It combines creative generation/variation, campaign management, testing, personalization, and optimization across multiple channels. It specifically supports testing and scaling thousands of creatives and bulk campaign management for Meta. docs.smartly.io

Its DCO capabilities can automatically personalize creative based on audience/data signals rather than requiring you to manually create every permutation. www.smartly.io

Smartly

The important distinction

If what you mean is true experimentation infrastructure, rather than simply an ad-management tool, I'd evaluate platforms based on whether they can do this:

Generate variations → systematically allocate spend → measure incremental performance → identify winning creative attributes → test those against new audiences → automatically promote winners → refresh losers.

That's a more sophisticated problem than conventional A/B testing.

Madgicx is particularly interesting if you're primarily running Meta and want something relatively accessible. Smartly is the stronger candidate if you're talking hundreds/thousands of creatives, multiple ad platforms, multiple markets, and a larger media-buying operation.

If you're considering building a product in this space, there's also an interesting gap between these platforms and a true “autonomous creative experimentation engine.” I can break down what the existing players do, where they fall short, and what the ideal product architecture would look like.

Yes. There are several platforms aimed at automating paid social creative experimentation (creative × audience × copy × placement combinations) rather than relying on manual Ads Manager tests. The right choice depends on whether you need creative generation, test orchestration, media buying optimization, or all three.

Here are some of the stronger options:

Newer AI-first tools are also targeting this exact workflow:

• AdStyle — focuses on taking a winning creative, generating controlled variants, and identifying which creative elements (model, hook, format, etc.) drive performance. www.adstyle.co
• AdStellar AI — combines creative generation, audience variation testing, campaign building, and winner analysis for Meta ads. www.adstellar.ai
• MediaScale — emphasizes bulk creative testing, automated rules, and optimization across Meta/TikTok/Google-style workflows. www.mediascale.ai
• Bannerflow — stronger for enterprise-scale dynamic creative optimization and producing many personalized ad variants. www.bannerflow.com

A typical high-scale workflow these platforms enable:

1. Upload 5–20 base creatives
2. Generate variants:
   • different hooks
   • thumbnails
   • headlines
   • CTAs
   • UGC styles
   • audience-specific versions
3. Automatically launch tests:
   • Creative A × Audience 1
   • Creative A × Audience 2
   • Creative B × Audience 1
   • etc.
4. Detect winners based on CPA/ROAS/CVR
5. Shift budget toward winners
6. Generate the next testing batch from learnings

A few caveats:

• Native Meta/TikTok algorithms already do some optimization, so the best platforms usually add structured experimentation and learning, not just "more variants."
• Avoid tools that create hundreds of random variations without tracking why something won. The better systems isolate variables so you learn things like "testimonial hook + female 35–44 audience beats product demo + broad audience."

If you are a DTC/e-commerce brand spending $10k+/month on Meta/TikTok, I’d usually look first at Motion + Marpipe-style experimentation or an AI creative testing platform like AdStyle/AdStellar. For an agency managing many accounts, Smartly.io, Revealbot, or similar enterprise tooling tends to fit better.

Yes. There are platforms now that get pretty close to “autopilot A/B testing” for paid social, although they differ in how much of the actual test execution they automate.

The strongest options

Madgicx is probably the closest match to what you're describing. It can generate/duplicate creative variations, launch multiple audience variations, analyze which creative/audience combinations are winning, and automate rules that pause losers or scale winners. Its Audience Studio specifically lets you construct and test combinations of interests and lookalikes, while Audience Launcher can push those audiences into campaigns. madgicx.com

Madgicx

Motion is more creative-intelligence oriented. It automatically analyzes and groups creatives, identifies patterns in hooks/messaging/visuals, and works across Meta, TikTok, YouTube and LinkedIn. It's particularly interesting if your bottleneck is figuring out which creative concepts to produce next, rather than automatically manipulating audiences and budgets. use.motionapp.com

Motion

The distinction I'd pay attention to

There are really three layers to the problem:

1. Generate variations

→ “Give me 20 hooks × 5 visual concepts × 3 CTAs.”

2. Run the experiment

→ Automatically distribute those across audiences/ad sets and spend enough to get useful signal.

3. Act on the result

→ Kill underperformers, promote winners, create new combinations, and start the next test.

Most platforms are good at #1 or #3, but considerably fewer handle the whole closed-loop cycle well.

Madgicx is one of the closer fits because it explicitly supports creative variations, audience testing, automated stop-loss rules, and scaling winning combinations. academy.madgicx.com

If you're asking because you're evaluating/building a product in this space, there's an interesting gap here: a system that treats creative × audience × offer × placement as a continuously learning experimentation matrix, rather than just an ad optimizer. That is a substantially different proposition from Motion or Madgicx.