UGC questions › Making ads that convert

What does AI UGC look like in 2027?

TL;DR

AI UGC ads are allowed on every major ad platform in 2027, but more of them now need a label. AI videos are cheap to make, often under a few dollars each. The catch is that most ads don't become winners, so cost per winning ad matters more than cost per video. Independent tests show AI ads can get clicks, but they lose a lot of them when labeled as AI.

Key findings

  • No US law bans AI people in ads. New York has required a clear label on ads with AI performers since June 9, 2026.
  • TikTok needs a label on ads made or heavily changed with AI. You can use its AI tag or your own clear note.
  • The FTC has no blanket ban on AI avatars. An AI "customer" sharing an experience that never happened is the risk.
  • In one study, AI ads got about 19% more clicks than human-made ads. Adding an "AI-generated" label cut clicks by about 31%.
  • Only 45% of Gen Z and millennials feel good about AI ads. Ad executives guessed 82%.
  • In our long-running ad data, 7.3% of ads use AI people. Among ads that ran a full year, only 1.8% do.
  • Popular claims like "AI UGC gets 3-5x ROAS" have no source we could find.

What counts as AI UGC?

TL;DR: AI UGC is any "creator-style" ad where software makes some or all of it. It ranges from a fully fake person to small AI edits on real footage.

We sort AI UGC into four kinds, because the rules and results differ for each.

KindWhat it isLabel risk
AI avatarA software-made person talks to cameraHighest: counts as fully AI-generated on TikTok and as a synthetic performer in New York
AI voice over real footageReal product shots or B-roll with an AI voiceoverMedium: an AI voice clone of the main subject needs a label on TikTok
AI product videoGenerated scenes of a product being usedMedium to high, depending on how real it looks
AI edits to real footageA real creator, with AI used for captions, background, translation or cleanupLow: minor edits are generally exempt

Our comparison of AI UGC and real creators covers when each kind fits. This guide covers what changed for 2027: the rules, the costs and the evidence.

Are AI UGC ads allowed on Meta, TikTok and Google?

TL;DR: Yes, on all three. TikTok asks for a label on AI-made ads. Meta and Google mostly require labels for political ads, but both now add AI info on their own.

This is how each platform's published rules read as of November 10, 2026.

PlatformWhat the rules saySource
Meta (Facebook, Instagram)Advertisers must disclose AI use in ads about social issues, elections or politics in certain cases. Meta adds information to ads it detects were made with third-party AI tools, shown in "About this ad". It says its approach to AI labels in ads "continues to evolve".Meta, February 2026, updated June 1, 2026
TikTokAI-generated or significantly edited content can run in ads only with TikTok's AIGC label or your own clear disclaimer, caption, watermark or sticker. Ads Manager has a "contains AI-generated content" setting. A label doesn't make a misleading claim acceptable.TikTok ads policy: misleading and false content
Google AdsSince July 2026, advertisers can add an AI label without breaking image-overlay rules. Election ads must still tick "altered or synthetic content".Google Ads policy help
YouTubeCreators must disclose realistic altered or synthetic content. Edits for looks, AI-written scripts and cloning your own voice are exempt.YouTube help

For a brand running AI UGC on Meta, there's no general self-label rule for commercial ads in Meta's published policy that we could find. Some blogs say otherwise, but we couldn't match their claims to Meta's own text. Meta may still add its own AI note to your ad.

What do the FTC, New York and the EU require?

TL;DR: The FTC cares whether the testimonial is true, not whether the person is real. New York and the EU now require clear AI labels in some ads.

RuleWhat it requiresIn force
FTC rule on fake reviews and testimonialsBans testimonials from people who don't exist or who didn't have the experience described. FTC staff say there's no blanket ban on AI avatars. Courts can impose civil penalties for knowing violations.October 2024 (FTC Q&A)
New York synthetic performer lawAds that knowingly use a "synthetic performer" must say so conspicuously. Penalties are $1,000 for a first violation and $5,000 after. Audio-only ads are excluded.June 9, 2026 (Cooley)
EU AI Act, Article 50Deepfake content, including realistic AI people, must be disclosed as AI-made.August 2, 2026 (AI Act, Article 50)

The New York law doesn't say what "conspicuous" means or where the label should go. Several other states have likeness and digital-replica laws, and those change quickly. This page isn't legal advice, so have your counsel check anything that runs in New York, the EU or a regulated category.

A simple test for the FTC rule: if your AI person says "I used this and it changed my skin", someone has to have actually had that result, and the claim has to be true. If nobody did, the ad is a fake testimonial whether the person is real or AI. An AI presenter explaining features ("this serum has 10% niacinamide") is a different kind of ad.

Can an AI person make health, finance or weight-loss claims?

TL;DR: The same rules apply as for a real person. An AI label doesn't fix a claim you can't back up.

There's no special exemption for AI. Health, money and weight-loss claims need the same proof they always did, and Meta and TikTok's health and financial ad policies still apply. An AI "doctor" or "patient" sharing results that never happened combines the two biggest risks: a fake testimonial and an unproven health claim.

In our data on long-running ads, 42% of the ads featuring AI people were also in compliance-sensitive categories like health and finance. Those are exactly the ads that get the closest review.

Do AI UGC ads actually work?

TL;DR: They can get clicks. But labeled AI ads lose a lot of those clicks, and AI ads can do less for your brand over time.

The best independent evidence we found, with what each study measured:

FindingWho and whenWhat it measured
Fully AI-made ads got about 19% more clicks than ads made by human experts (4.44% vs 3.73% click-through)NYU Stern and Emory, 2025 (coverage)Display ads in a live field test
Labeling ads "AI-generated" cut click-through by about 31.5%Same studyThe same ads, with and without the label
AI ads scored 14% lower on creative effect and 17% lower on brand effect, even when viewers couldn't tellIpsos and Syracuse University, via Harvard Business Review, 2026Survey-based ad testing
45% of Gen Z and millennials feel positive about AI ads. Ad executives guessed 82%.IAB, January 2026Survey of 505 consumers and 104 ad executives

Put together, AI ads can win the first click, especially when nobody knows they're AI. Labels are spreading, and the brand cost shows up even when viewers can't tell. Both make it harder to build a lasting ad account on AI people alone.

Two popular claims, "AI UGC gets 3-5x ROAS" and "AI UGC gets 2-4x higher CTR", have no study behind them that we could find. They mostly appear on pages selling AI tools. Our audit of 45 UGC statistics explains how to check numbers like these.

How long do AI UGC ads keep running?

TL;DR: AI people show up much less often in ads that have run a full year. That may be partly because AI ads are newer.

We looked at 1,571 long-running Meta video ads across 54 markets, each tagged for what appears in it.

GroupAdsAI people in the ad
All long-running ads1,571115 (7.3%)
Ads running 365+ days2174 (1.8% of these ads)
Share of AI-people ads that reached 365+ days1153.5%, against 14.6% for ads without AI people

The median AI-people ad had been live 60 days, against 123 days for the rest. One reason could be that realistic AI video is fairly new, so fewer AI ads have had time to reach a year. Even so, it's a useful check on the idea that AI ads can replace an account's long-running winners. Our part 2 run-length study shows how much run length depends on industry, and our first longevity study found no AI avatar presenter among ads that had lasted.

AI people were most common in apps and SaaS (10.1%) and dental and clinic ads (10.0%), and rarest in food and drink (1.1%) and finance (2.7%).

What does AI UGC really cost?

TL;DR: An AI video can cost a few dollars, compared with about $200 for a creator video on one big marketplace. Cost per winning ad is the number that matters.

These were the listed plans on the tools' own pricing pages on November 10, 2026.

ToolMonthly plans (US$)Notes
Creatify$15 to $195 ($10 to $136 billed yearly)Credit-based; a 15-second avatar clip uses about 5 credits
MakeUGC$29 to $79, billed yearlyCredits vary by model, length and resolution
HeyGen$29 Creator, $49+ Pro, $149 BusinessCredit-based
Synthesia$19 Starter, $89 ProLower with yearly billing
invideo$20 to $150 per seatLower with yearly billing

For human creators, Collabstr's marketplace reported an average UGC asking price of about $198 in its 2025 report. That's one marketplace's listed prices, and it doesn't include usage rights, editing or ad management. Our UGC cost guide breaks those costs down.

Cost per video hides two things.

  • Re-rolls. AI tools charge per generation. Buyers on Reddit describe burning through credits to get one usable video, because of stiff lip-sync, wrong product details or odd hands.
  • Hit rate. Most ads don't become winners. In our own account, about 1 in 13 new UGC ads became a winner over the last 90 days, close to the 7.6% industry figure from Motion.

The simple way to compare is:

Cost per winner = cost per finished ad ÷ hit rate

Say a finished AI ad costs $20 after re-rolls and editing, and 1 in 25 becomes a winner. That's $500 per winner. A human UGC ad at $300 with a 1 in 13 hit rate comes to about $3,900 per winner. These numbers are only an example of the method. Nobody has published AI UGC hit rates yet, so track your own for both kinds before you move budget.

Where does AI UGC win, and where does it lose?

TL;DR: AI works well for testing ideas quickly and for apps, explainers and translations. Real people still win where trust, touch and long-running ads matter.

AI tends to fitReal creators tend to fit
Testing many hooks and angles quickly before filmingTestimonials and before-and-after results
App walkthroughs and screen demosProducts people need to see handled, worn or tasted
Translating a winning ad into other languagesHealth, finance and other high-trust categories
Explainer ads with a presenter, not a "customer"Ads you want to keep running for months
Filling B-roll gapsAudiences that react badly to AI, like many Gen Z viewers

The pattern we see in buyer discussions on Reddit matches this. Many media buyers describe using AI to "audition" hooks cheaply, then paying real creators to reshoot the hooks that win. The common complaints about AI ads are stiff lip-sync in the first second, comments calling the ad "AI slop" and strong click numbers that don't turn into sales.

How should you use AI UGC in 2027?

TL;DR: Use AI to test fast, label it where rules require, keep claims true and reshoot your winners with real people.

  1. Decide the job first. Use AI for testing, translation or explainers. Use real people for proof and trust.
  2. Never fake an experience. An AI person can explain a product. It can't claim results nobody had.
  3. Label by platform and place. Turn on TikTok's AI setting. Add a clear on-screen label for ads served in New York or the EU.
  4. Track cost per winner, not cost per video. Count re-rolls, editing time and hit rate for AI and human ads side by side.
  5. Reshoot winners with real creators. An AI hook that wins is a cheap test result. A human version is usually what keeps running.
  6. Check your creator contracts. If you hire real creators, put any AI use of their face or voice in writing. Our usage rights guide covers the terms.
  7. Watch for rule changes. Meta says its AI labeling for ads is still changing, and more states are writing likeness laws.

What are the limits of this guide?

TL;DR: Rules change fast and this isn't legal advice. Some prices and studies came from secondary sources, and we say which.

  • Platform rules and laws are summarized as published on November 10, 2026. Check the linked pages before you launch.
  • We read the NYU and Emory figures from press coverage of the paper, and the Ipsos and Syracuse figures from Harvard Business Review's summary.
  • Tool prices change often, and credit costs per video vary by model and length.
  • Our long-running ad data tags AI people with a model, uses ads that were still live, and can't separate "AI ads fail" from "AI ads are newer".
  • The cost-per-winner example uses made-up numbers to show the method. It isn't a benchmark.

Sources

On Google

See more of She Converts in your search results

Add us as a preferred source and Google shows our UGC ad research more often in Top Stories and AI answers.

Add as preferred source

Want to see what this looks like for your ads?

Book a 30-minute call. We review your current ads and your offer and see if we're a fit. No retainer required.

Book a Call