AI testimonials and the fake reviews rule: what 16 CFR 465 bans, and what the FTC left legal on purpose

01Does the FTC's fake reviews rule ban AI avatars?

No, and the Commission says so twice in its own documents. The rule's Q and A page states plainly that there is no blanket prohibition on AI-generated avatars in marketing, and that section 465.2 "is drafted specifically so as to not prohibit companies from using virtual influencers." The prohibition runs on deception, not on the technology.

I build this creative for a living and I am not a lawyer, so what follows is a reading of the documents rather than advice about them; the FTC's questions page is linked below, so hand your counsel the source. The narrowing was deliberate. The Federal Register statement of basis and purpose records that the Commission "now recognizes that absent an express reference to material misrepresentations, the provision could be interpreted to prohibit other potentially non-deceptive speech, such as the use of virtual influencers." Two words, materially misrepresents, do the whole job.

Read those six together and the shape of US law here stops being about machines. Part 465 asks whether a viewer is being told about an experience that no person had. The Endorsement Guides ask whether a viewer would take the message as somebody's opinion. Neither one asks how the picture was made. That is also why the disclosure question is a separate post with a much smaller penalty, and why the platforms ask for something different again.

02The sentence that governs a generated spokesperson

Part 465 gets all the attention and the operative sentence sits somewhere else. It lives in the Endorsement Guides at 16 CFR 255.2(c), written long before anyone generated a face. 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."

Row two carries a second trap most brands walk into without noticing. Under 255.2(b), an endorsement about a consumer's experience on a key attribute is read as representing what consumers will generally achieve. The FTC tested the escape hatch and reported the result: "Neither disclosure adequately reduced the communication that the experiences depicted are generally representative." Results not typical does not buy back a typicality claim you cannot substantiate. That is a live problem for beauty and skincare creative in the US.

03What does 16 CFR Part 465 ban?

Six things, each declared an unfair or deceptive act or practice. Only the first is about who the reviewer is. The other five cover paid sentiment, insider reviews, company-run sites posing as independent, suppression of negative reviews, and the trade in fake indicators of social media influence.

465.2   Fake or false consumer reviews and testimonials:
        materially misrepresenting (1) that the reviewer or
        testimonialist EXISTS, (2) that they USED or had
        experience with the product, or (3) their EXPERIENCE

465.4   Compensation or incentives conditioned, expressly or
        by implication, on a review expressing a particular
        sentiment - positive or negative

465.5   Insider reviews and testimonials with no clear and
        conspicuous disclosure of the relationship, including
        officers', managers' and their immediate relatives'

465.6   Misrepresenting that a company-controlled site or
        entity provides independent reviews or opinions

465.7   Review suppression: unfounded legal threats, physical
        threats, intimidation, public false accusation - or
        showing a filtered set as though it were most or all

465.8   Selling, buying or procuring fake indicators of social
        media influence: bots, accounts not associated with a
        real individual, hijacked accounts

465.1(c) A disclosure is not clear and conspicuous if a consumer
        "must take any action, such as clicking on a hyperlink
        or hovering over an icon, to see it."

One prohibition catches brands that never wrote a review in their lives. Section 465.8 covers buying or selling fake indicators of social media influence, and 465.1(h) defines those as metrics generated by bots, by purported individual accounts not associated with a real individual, or by hijacked accounts. If somebody is selling you engagement on the creative, that is the paragraph they are selling you into.

04The FTC did two opposite things on one day

December 22, 2025 is the date to hold in your head, because a post written from 2024 sources gets this backwards. The Commission enforced the advertiser rule and dismantled its own flagship AI case in the same news cycle. Read together, the two actions move the exposure off the vendor that generates and onto the brand that publishes.

One more thing changed quietly in the same period. Four of the FTC's own 2023 business-blog posts on AI claims now return 404 at ftc.gov and survive only in the Internet Archive. I could find no Commission statement explaining the removal, so I report the 404s and nothing beyond them: the Endorsement Guides and Part 465 are both still in force.

A virtual influencer cannot say 'I am human' with an 'AI-generated' disclosure that contradicts the main claim.

Laura J. Protzmann, National Advertising Division, BBB National Programs, June 9, 2026

05How do you check an ad against the rule?

Watch the cut once with the sound on and answer one question: is anybody telling me what this product did for them? If nobody is, Part 465 stays out of it and you are checking claims instead. If somebody is, the next four questions decide which paragraph you are in, and the scoring below is how I run it on our own files before delivery.

The check that catches the most in our own review is the second question, and it costs nothing to ask. A generated person who resembles nobody is a manageable problem. A generated version of a real creator is a different conversation entirely, and it belongs in the contract before it belongs in the edit, alongside what usage rights cost a buyer and the five briefs worth taking to a real person. The first question is the cheaper one to design around: a presenter who demonstrates the product without narrating a result stays outside the rule, and getting a hand and a product to meet convincingly is a craft problem rather than a legal one.

06What it costs, and who is on the hook

$53,088 per violation is the statutory maximum in 16 CFR 1.98 today, and the FTC used that figure in its own December 2025 letters. It is a ceiling, not a price list: courts weigh the factors in section 5(m)(1)(C), may impose far lower per-violation amounts, and decide how violations get counted in the first place. Anyone quoting $53,088 times the number of reviews is selling you something.

The label answers a question nobody at the FTC is asking. The regulator wants to know whether a viewer was told something untrue about who is speaking and what happened to them, and a machine can generate a face without generating a single false claim. Write the script so no invented person reports an experience nobody had, and the hardest rule in this subject stops applying to you at all.