The short version: an AI-generated presenter cannot legally deliver a testimonial about your product in US advertising. A testimonial is a claim about a real person's actual experience, and a synthetic avatar has not had one. The FTC Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465) took effect on 21 October 2024 and explicitly covers testimonials that misrepresent being made by someone who does not exist, whether the fake was produced by a human or by generative AI. What is still perfectly legal is paying a real creator: if they used the product, the opinion is honest, and the paid relationship is disclosed clearly and conspicuously, you are on the right side of the line.
This question comes up constantly now, because AI UGC tools got good enough in the last eighteen months that a synthetic person holding your product looks convincing in a nine-second vertical video. The technology question has been settled. The one that decides whether you can actually run the output is legal, and a surprising number of brands are getting it wrong in the same specific way.
What the FTC rule actually prohibits
The rule bans six categories of conduct. The one that catches AI testimonials is the first: creating, buying, or disseminating fake consumer reviews and testimonials. "Fake" is defined by two failure modes, and it is worth separating them because brands usually only think about the first.
- The person does not exist. An AI-generated persona, a stock face with a fabricated name, an invented customer. This is where synthetic avatars land, regardless of how they were produced.
- The person exists but had no actual experience. A real hired actor who never used the product, reading a script that presents their words as a genuine customer experience. This one surprises people, because everyone involved is real.
Both are treated the same way. Civil penalties run up to $53,088 per violation under the FTC's 2025 inflation adjustment, and that figure still applies in 2026 because the January adjustment was canceled after the appropriations lapse stopped the October 2025 CPI data from being published. Per violation is the phrase to pay attention to. On a campaign running one asset across several placements, the arithmetic gets unpleasant quickly.
What is banned and what is allowed
The distinction is not "AI bad, humans good." It is narrower and more useful than that.
| Scenario | Status | Why |
|---|---|---|
| AI avatar says the product worked for them | Prohibited | Testimonial from a person who does not exist |
| Actor reads a script as though they were a customer | Prohibited | Real person, no actual experience with the product |
| Paid creator who used the product, disclosed | Allowed | Real experience, honest opinion, material connection disclosed |
| Unpaid customer review you did not solicit selectively | Allowed | Genuine consumer speech |
| AI voiceover over product b-roll, no personal claim | Allowed | Not a testimonial, nobody is vouching for a result |
| Clearly labeled dramatization with a paid actor | Allowed | Not presented as a real customer experience |
| AI used to edit, caption, or translate real footage | Allowed | The underlying testimonial is still real |
The row that trips up the most brands is the actor. Hiring someone to perform a dramatization is legal and has been standard practice in advertising for decades. The violation happens at the framing: the moment that performance is presented as a customer's own experience rather than a portrayal, it becomes a fake testimonial, and the exposure is identical to using a synthetic face. If you use actors, label the dramatization and do not let the ad copy imply otherwise.
Can I use AI in UGC-style ads at all?
Yes, in most of the pipeline. The constraint is specifically on who is making the claim, not on which tools touched the file.
Generating b-roll, cutting the edit, writing hook variants, burning in captions, translating a video for another market, and cleaning up audio are all fine. None of them involve a synthetic person asserting a personal experience. Plenty of high-performing UGC ads are demonstrations rather than testimonials: the product being used, a before and after, an unboxing, a problem being solved on camera. Those formats do not require anyone to vouch for a result, which is exactly why they carry no testimonial risk.
Where it breaks down is the format brands most want to automate, because it is the one that converts best. Testimonials work precisely because a real person is putting their credibility behind a claim. Strip out the real person and you have removed the mechanism that made the format effective, then added legal exposure on top. That trade is bad on both ends.
The platform rules are a second, separate problem
Even where the FTC rule leaves you room, the ad platforms have their own policies, and they enforce faster than any regulator because enforcement is automated and the penalty is your account rather than a fine.
Meta requires disclosure of AI-generated or digitally altered realistic content and applies its own labeling to material it detects. TikTok requires creators to disclose realistic AI-generated content, and TikTok Shop's seller guidance for product video is stricter still: it calls for a visible face, a physical product, a real environment, and genuinely recorded speech, explicitly rejecting AI-generated voice, still-image slideshows, and product page screenshots. If TikTok Shop is a meaningful channel for you, synthetic presenters are not a compliance grey area, they are a rejected asset.
Policies here change more often than statute does, which is an argument for keeping the obligations that attach to each campaign somewhere you actually track them against the controls you have in place, rather than in the head of whoever briefed the last flight of ads. The disclosure requirement that applied when a video was made travels with that video every time it is reused.
How to run testimonial creative correctly
Four things, and none of them are expensive.
- Ship the product before you brief the testimonial. The creator needs an actual experience to describe. This also produces better footage, because specific details only come from real use.
- Disclose the material connection clearly and conspicuously. In the video itself where practical, not only in a caption a viewer has to expand. Paid partnership tools on each platform are the floor, not the whole obligation.
- Let creators keep the caveat. Universally glowing footage reads as bought, because genuine customers almost never produce it. A small admission that one thing did not work perfectly raises credibility and keeps the opinion honest, which is what the rule requires anyway.
- Get the licensing in writing before filming. Rights are renegotiated after delivery far more often than brands expect. The UGC usage rights page covers what each licence type actually grants.
Nothing on that list requires a compliance department. It requires deciding up front that the person on camera is real and that you are willing to say the video was paid for.
Does disclosure hurt performance?
Less than people fear, and the fear is usually a proxy for a different problem. A disclosed testimonial from someone who genuinely used the product and can describe a specific result still works, because the persuasive weight sits in the specificity rather than in the viewer's belief that nobody was paid. Audiences in 2026 assume most creator content is commercial. What they punish is being misled about the substance, not being told about the arrangement.
If a testimonial only performs while the audience believes it was unpaid, the testimonial was weak. That is worth knowing early, and it is cheaper to learn from three honest videos than from one enforcement action.
Where this leaves your creative pipeline
Split your briefs into two buckets and the whole question gets simpler. Demonstrations, unboxings, b-roll, and explainers carry no testimonial claim, so you have wide latitude on tools and talent. Testimonials assert a personal experience, so they need a real person who genuinely had one, plus disclosure. Most brands need both, and confusing them is what creates the risk.
For the testimonial bucket specifically, the sourcing requirement is narrow: identity-verified real people who will actually use the product before filming. That is the whole premise of how UGC testimonials are commissioned here, and it is why every profile on the UGC marketplace belongs to a verified human who is 18 or older rather than an AI persona. If you want polished on-camera customer stories for a product page or a sales deck instead of feed-native ad creative, video testimonials cover that format.
The short answer to the question in the title has not changed since the rule took effect: you can use AI almost everywhere in the production of an ad, and nowhere in the mouth of the person vouching for your product.
This article is general information about US advertising rules, not legal advice. Regulations and platform policies change, and the specifics of your category may add requirements. Check with counsel before running claims you are unsure about.
Last updated July 2026.
See how UGCMarketplace works for your kind of brand on the use cases page.
UGC MARKETPLACE // POST A BRIEF
Run this on your next campaign
Post a brief, match with vetted real creators, and receive licensed short-form videos and photos you own and run as your own paid ads.