The FTC's Fake Testimonial Rule and AI Actors: Where the Line Actually Is

Quick answer:The FTC does not ban AI actors in ads. Its staff guidance on the Consumer Reviews and Testimonials Rule states: "The rule has no blanket prohibition on the use of AI-generated avatars in marketing." The line runs through the script, not the face. If what the AI actor says is a message "that consumers are likely to believe reflects the opinions, beliefs, or experiences of a consumer who has purchased, used, or otherwise had experience with a product", it is a consumer testimonial. A testimonial from a person who does not exist and never used the product is exactly what 16 CFR 465.2 prohibits, with civil penalties for knowing violations. If the actor presents the product, explains what it does and what it costs, and never claims personal experience, the FTC's own worked examples say there is no testimonial. An AI UGC ad written in the first shape is the problem. The same ad can be written in the second.

Two rulebooks, one test

Two FTC documents govern this. The first is a trade regulation rule: 16 CFR Part 465, the Consumer Reviews and Testimonials Rule, which the Federal Register notice says is "effective October 21, 2024". A rule carries civil penalties. The second is the Endorsement Guides, 16 CFR Part 255, revised effective 26 July 2023. The Guides are interpretive: they tell you how the FTC reads Section 5 of the FTC Act, and an ad that breaches them is attacked as a deceptive practice rather than as a rule violation.

Both turn on the same question, and it is not "was this made with AI". It is "what are consumers likely to believe about the person speaking".

What the rule actually prohibits

Section 465.2(a) makes it "an unfair or deceptive act or practice and a violation of this part for a business to write, create, or sell a consumer review, consumer testimonial, or celebrity testimonial that materially misrepresents, expressly or by implication: (1) That the reviewer or testimonialist exists; (2) That the reviewer or testimonialist used or otherwise had experience with the product, service, or business that is the subject of the review or testimonial; or (3) The reviewer's or testimonialist's experience with the product, service, or business that is the subject of the review or testimonial." Section 465.2(b) extends the same three limbs to a business that disseminates such a testimonial when it "knew or should have known" of the misrepresentation.

Read those three limbs against an AI actor delivering a first-person story. The testimonialist does not exist (limb 1). It never used the product (limb 2). Its described experience is fiction (limb 3). And because you wrote the script, the "knew or should have known" standard in (b) is met by definition. The FTC's press release names the scenario: the rule "addresses reviews and testimonials that misrepresent that they are by someone who does not exist, such as AI-generated fake reviews, or who did not have actual experience with the business or its products or services, or that misrepresent the experience of the person giving it." Everything therefore depends on whether the ad is a "consumer testimonial" at all, and that is a defined term.

The definition that decides it

Section 465.1(f) reads in full: "Consumer testimonial means an advertising or promotional message (including verbal statements, demonstrations, or depictions of the name, signature, likeness, or other identifying personal characteristics of an individual) that consumers are likely to believe reflects the opinions, beliefs, or experiences of a consumer who has purchased, used, or otherwise had experience with a product, service, or business." And 465.1(o): "Testimonialist means the individual giving or purportedly giving a consumer testimonial or celebrity testimonial."

Three words carry the weight. "Likeness" means a synthetic face counts as a depiction. "Purportedly" means the testimonialist does not need to be real. And "likely to believe" means the test is the viewer's impression, not your intent and not the production method.

The FTC staff Questions and Answerspage (dated November 2024) walks through this for AI avatars, and the full passage matters because summaries usually stop after the first sentence: "The rule has no blanket prohibition on the use of AI-generated avatars in marketing. As discussed in our Federal Register notice, Section 465.2 of the rule is drafted specifically so as to not prohibit companies from using virtual influencers. Further, AI-generated ‘stock avatars’ would not be giving ‘consumer reviews’ as that term is defined in the rule. It's possible that a company's use of such an avatar might be considered a ‘testimonial’ under the rule. That would be prohibited under the rule only if the underlying testimonials were fake or false. The use of avatars could also be deceptive in violation of the FTC Act."

So the avatar is fine. The avatar giving a testimonial is fine if a real customer actually said it. The avatar giving a testimonial nobody gave is the violation. The same page answers the human-actor version identically: "If the television commercial is an obviously fictional dramatization, then the actors are not giving testimonials."

The Endorsement Guides draw the same line with worked examples

The Guides define an endorsement in 255.0(b) as "any advertising, marketing, or promotional message for a product that consumers are likely to believe reflects the opinions, beliefs, findings, or experiences of a party other than the sponsoring advertiser, even if the views expressed by that party are identical to those of the sponsoring advertiser." The 2023 revision added that the endorser "could be or appear to be an individual, group, or institution." The Federal Register notice says why: "the new language is intended to also encompass the writers of fake reviews and non-existent entities that purport to give endorsements." A synthetic presenter is a non-existent entity.

The examples in 255.0 then do the work directly. Example 3: an announcer who "does not purport to speak from personal experience, nor on the basis of their own opinions, but rather in the place of and on behalf of the drug company" is not endorsing. Example 6 puts the same actor on both sides of the line. In 6(i), an actor hosting a fitness infomercial says "This is the most effective and easy-to-use home exercise machine that I have ever tried." The Guides conclude: "Even if the actor is reading from a script, the statement would be an endorsement, because consumers are likely to believe it reflects the actor's personal views." In 6(ii), the same actor instead explains that the machine was designed by exercise physiologists, isolates five muscle groups, is used fifteen minutes a day, and then gives the price. Conclusion: "As the actor does not say or do anything during the infomercial that would lead viewers to believe that the actor is expressing their own views about the machine, there is no endorsement." Swap "actor" for "AI actor" and nothing in the reasoning changes. The face is a delivery mechanism. The script is the legal object.

The two script shapes, side by side

The left column is the testimonial shape. The right column sells the same product and stays inside Example 3 and Example 6(ii).

Every line on the left implies purchase, use and a result experienced by the speaker. Every line on the right conveys the brand's claims through a presenter who never says "I used it." If you want the left column anyway, 255.2(c) sets the price: "Advertisements presenting endorsements by what are represented, expressly or by implication, to be ‘actual consumers’ should utilize actual consumers in both the audio and video, or clearly and conspicuously disclose that the persons in such advertisements are not actual consumers of the advertised product." That is a disclosure in the creative, not the caption. And a disclosure only cures the Guides problem. It does not manufacture a real customer, so a first-person story delivered by a disclosed AI actor is still a testimonial nobody gave. Rewrite it, or put a real customer's words behind it.

The claim inside the script is a separate liability

Moving a line from the left column to the right does not launder the claim it carries. 255.1(a) states that "an endorsement may not convey any express or implied representation that would be deceptive if made directly by the advertiser." And 255.2(a) is blunt about what a consumer story is worth as evidence: the advertiser must hold "adequate substantiation, including, when appropriate, competent and reliable scientific evidence", and "Consumer endorsements themselves are not competent and reliable scientific evidence."

For supplements, skincare and anything health-adjacent, the FTC's Health Products Compliance Guidancecloses the ‘Results not typical’ escape hatch: "Testimonials that report results more dramatic than users can generally expect are likely to be deceptive. Moreover, attempts to disclaim dramatic results with statements like ‘Results not typical’ don't cure the deception. Those testimonials should be accompanied by a clear and conspicuous disclosure of the results a typical consumer can actually expect." If you would not put the number in your own brand voice on the product page, it does not belong in a synthetic mouth either. Which platform rejects the claim first is covered in the Meta and TikTok policy posts, and the health-specific rules each one applies to a supplement or skincare script are quoted in supplement and skincare claims in AI UGC ads.

Celebrity and creator likenesses

465.1(b) defines a celebrity testimonial with the same "likeness, or other identifying personal characteristics" language, applied to "a well-known individual". The staff Q&A spells out the consequence: "It would also violate the rule for someone to use a celebrity avatar without the celebrity's permission to speak favorably about a product, if reasonable consumers would think that the celebrity actually gave a testimonial for the product." Prompting a model toward a recognisable creator's face or voice is the one AI-specific move that lands inside the rule however the script is written. State disclosure statutesstack on top of it. The voice half of the same question, including Tennessee's ELVIS Act, is in AI voices in ads.

What this means for an AI UGC workflow

Two facts about our own pipeline decide which shape you are in before a frame renders. Every avatar model we run starts from an image you supply, so the person on screen is a synthetic actor you chose. And the script you type reaches the model as written; we do not rewrite it. The 465.1(f) test is applied to those words, and nothing about the model or the credit cost changes the answer.

• Write AI actor scripts in the presenter shape by default: features, mechanism, price, offer, a demonstration with the product in hand. The five-beat structure still works; the payoff beat becomes what the product is designed to do rather than what "I" experienced.
• Never script "I bought", "I used", "I lost", "my skin" for a synthetic actor unless a real customer said it and you can produce that customer. An AI actor reading a genuine review, disclosed as such, is defensible. An AI actor pretending to be the reviewer is not.
• If a test genuinely needs the first-person shape, burn the disclosure into the creative in the audio and on screen. 255.2(c) is what triggers the disclosure; 255.0(f) is what places it: "if the representation is made through both visual and audible means, the disclosure should be made in the communication's visual and audible portions." Our export test found provenance metadata does not survive a remux; text in the pixels and words in the audio are the only disclosure you control.
• Keep the substantiation for every claim in the script, whichever shape it takes.
• Do not prompt toward a real person's likeness or voice, and keep a record of which actor image and script each ad used. Both rulebooks turn on what you knew.

A media buyer who asks "is AI UGC legal" before buying a tool is usually asking about this rule, and the answer is not a ban. It is a rewrite. None of this is legal advice, and the FTC says the same of its own Q&A: "Our staff guidance isn't definitive or comprehensive, and it doesn't provide a safe harbor from potential liability." A lawyer should read your scripts if you run health claims at volume.

Sources

• 16 CFR Part 465, Rule on the Use of Consumer Reviews and Testimonials: 465.1 Definitions and 465.2 Fake or false consumer reviews, consumer testimonials, or celebrity testimonials (Cornell LII)
• Federal Trade Commission, "The Consumer Reviews and Testimonials Rule: Questions and Answers" (staff guidance, November 2024)
• Federal Trade Commission, "Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials" (press release, 14 August 2024)
• Federal Register, Trade Regulation Rule on the Use of Consumer Reviews and Testimonials (22 August 2024; effective 21 October 2024)
• 16 CFR Part 255, Guides Concerning the Use of Endorsements and Testimonials in Advertising: 255.0 Purpose and definitions (including 255.0(f), the "clear and conspicuous" definition), 255.1 General considerations, 255.2 Consumer endorsements (Cornell LII)
• Federal Register, Guides Concerning the Use of Endorsements and Testimonials in Advertising (26 July 2023; the notice explaining the "could be or appear to be" language)
• Federal Trade Commission, Health Products Compliance Guidance

All seven were read directly on 4 September 2026. The CFR text is stable; the staff Q&A page is the one most likely to be revised, and its date line is how you check.

The FTC decides whether the script is a testimonial. The platforms decide whether the ad runs at all: TikTok puts the label on you, and the EU AI Acttreats a realistic synthetic presenter as a deep fake whatever it says. The UK's CAP Code reaches the same script test by a different route, covered in the UK post. The talking-head anatomyshows where the energy comes from once the "I" is gone. UGC Vids AI renders the presenter-shaped scripts in the right-hand column from an image you supply, from $49/month, free for 3 days.