Are AI UGC ads legal? FTC rules and likeness rights

When does an AI actor ad count as a fake testimonial?

On our reading, when viewers are likely to take the AI actor for a customer it is not.

The FTC rule defines a consumer testimonial (16 CFR 465.1(f)) as 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, service, or business”.

Section 465.2(a) makes it a violation “for a business to write, create, or sell” a testimonial that “materially misrepresents, expressly or by implication”. That covers whether the testimonialist exists (a)(1), used or otherwise had experience with the product (a)(2), or their experience with it (a)(3).

The FTC’s announcement of 14 August 2024 gives “AI-generated fake reviews” as its example of a review by “someone who does not exist”.

Section 465.2(b) reaches a business that disseminates, or causes the dissemination of, a testimonial “about the business or one of the products or services it sells” that it “knew or should have known” materially misrepresented those points. Paragraph (a) has no such product link, so on our reading it also reaches the agency or freelancer who writes the script.

A first-person claim from either kind of AI actor falls within 465.2, on our reading, under different paragraphs. A fully synthetic face is a testimonialist who does not exist, (a)(1). A licensed replica is a real, paid person who, unless they used or otherwise had experience with your product, falls under (a)(2).

The FTC’s announcement also says the rule lets it “seek civil penalties against knowing violators”. The ceiling is $53,088 per violation (16 CFR 1.98(d)).

The FTC’s Endorsement Guides draw the presenter line (255.0, Example 6). An actor calling a machine the best “that I have ever tried” endorses it “Even if the actor is reading from a script”. One who describes the machine and its price does not: nothing leads viewers “to believe that the actor is expressing their own views about the machine”.

The Guides are the FTC’s “administrative interpretations” of the law (255.0(a)). “Whether a particular endorsement or testimonial is deceptive will depend on the specific factual circumstances of the advertisement at issue.” On these texts, what decides is whether the actor claims use or voices an opinion as their own.

Section 255.2(c) gives a disclosure route. Ads with endorsements by people represented, “expressly or by implication”, as “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.” Part 465 has no matching disclosure clause.

The Guides’ hidden-camera example (255.2, Example 6) is the nearest to a UGC-style ad. With unnamed speakers, “the net impression conveyed to consumers may well be that these are actual customers. If actors have been employed, this fact should be clearly and conspicuously disclosed.”

Do you have to disclose an AI actor?

Under the FTC’s Guides, when the ad presents the actor as an “actual consumer” (255.2(c)). The three disclosure laws below turn instead on what kind of actor it is, and they point different ways.

A paid human creator falls under another clause. Under 255.5(a), a connection to the seller “that might materially affect the weight or credibility of the endorsement, and that connection is not reasonably expected by the audience”, “must be disclosed clearly and conspicuously”. Such connections “can include monetary payment or the provision of free or discounted products”.

New York’s S8420-A has required a conspicuous disclosure of a synthetic performer, one “not recognizable as any identifiable natural performer”, since 9 June 2026 on our count.

It binds a business that deals in the advertised product and “for any commercial purpose produces or creates” the ad, where it has “actual knowledge”. Whether it also reaches that business’s agency or freelancer is unclear (question 6).

California’s SB 1050 makes it unlawful to “create and cause to be published” an ad that “prominently includes a synthetic performer without a clear and conspicuous disclosure”. Its text names no start date. Under the state constitution (article IV, section 8(c)(1)), it starts on 1 January 2027, on our count.

The EU AI Act’s Article 50(4), applying since 2 August 2026, makes deployers of the AI system that generates a deep fake disclose it.

Article 3(60) defines a deep fake as content that “resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful”. Whether an AI actor ad meets that test is unclear.

New York and California describe a performer “not recognizable” as any real person. On our reading, a fully synthetic face fits those two definitions, and a real performer’s replica sits closer to the EU’s.

AI disclosure in advertising, platform by platform quotes each law’s exemptions, and New York’s and the EU’s penalties. It also quotes the Commission’s draft guidelines, which list two AI UGC formats as deep fakes.

TikTok’s ads policy (last updated April 2026) allows AI-generated ads that “Apply the AIGC label” or carry “a clear disclaimer, caption, watermark, or sticker of your own”. If TikTok finds undisclosed AI content, the ad “will be rejected or restricted”.

TikTok’s AI-generated content label for ads covers Spark Ads and the labels TikTok adds on its own.

Meta labels ads made with its own generative tools, and its update of 1 June 2026 said it would “begin automatically detecting ads created or edited using third-party AI tools”. Meta’s AI disclosure rules has the detail.

Likeness: the actor may be a real person who can withdraw

Arcads’ and Maxfusion’s terms describe their actors as real creators who can withdraw (clauses below), and Synthesia’s policy calls a stock avatar’s performer “the real-life actor”. On Arcads and Maxfusion, a withdrawal can pull your live ads.

California’s Civil Code 3344(a)(1) makes anyone who “knowingly uses another’s name, voice, signature, photograph, or likeness” for advertising “without that person’s prior consent” liable for damages. On our reading, the vendor’s licence is how that consent reaches you. Whether a withdrawal ends it for ads already live is unclear.

New York’s counterpart, Civil Rights Law section 50, makes a misdemeanour of using “for advertising purposes, or for the purposes of trade, the name, portrait, picture, likeness, or voice of any living person without having first obtained the written consent of such person, or if a minor of such minor’s parent or guardian”.

Arcads’ terms (in force from 22 September 2026, 13.3.2) say FRESHR, the company behind Arcads, “may ask the Client to delete any Videos from all supports where they are used, in the event of a withdrawal request made by the Creator”. The deadline is “the timeframe requested by FRESHR”.

HeyGen puts the same duty on custom avatars in its moderation policy (effective 25 July 2024, 2.1): “You are responsible for honoring any request by or on behalf of an Actor to remove content depicting the Actor wherever you have published such depictions.”

Synthesia’s Acceptable Use Policy (last updated 23 February 2024) bars stock avatars from “promoted”, “boosted”, or “paid” social ads “absent written express consent from Synthesia”. A stock avatar used so viewers think “the real-life actor personally holds, endorses, or actually has your stated opinions, characteristics or conditions” is allowed “only on enterprise Service plans”.

Synthesia’s help centre (updated 22 July 2026) is stricter. Stock avatars are for content “Free from personal opinions, endorsements, or sensitive claims”, and it names no plan exception. On our reading, a stock avatar voicing your opinion breaks both documents on other plans; on enterprise, they disagree.

Captions’ Prompt to Video article (modified 17 August 2026) calls its actors “completely virtual and free from licensing restrictions”. If so, no performer can withdraw, but on our reading these are the actors New York’s and California’s definitions describe.

Captions’ own terms split by product. In the app, avatars are Mirage Content (terms 1.3 and 2.1, 4 September 2025), and 3.1(xiv) bars “any commercial use of the Services including the Mirage Content except as authorized hereunder without our prior written consent”. On our reading, an ad using app actors needs that consent.

The Captions on Web terms (29 September 2025, 1.1) allow avatar output “for advertising or other purposes”.

AI UGC meaning tells a licensed replica from a synthetic character, and both from AI edits of a filmed take.

Contract clauses that put a price on misuse

Of the six tools below, only Maxfusion and Arcads put a fixed sum on misuse. The other four can end your licence or account instead.

All six bars concern what an ad claims. HeyGen, MakeUGC and Maxfusion also require the AI disclosure the law asks for. Captions’ web terms (29 September 2025, 1.2) ask you to “disclose that AI was used” and that outputs are ads, “as and when appropriate”.

Arcads’ 13.3.2 lump sum applies “in addition to the general provisions”. It is due “In the event of non-compliance with the conditions and/or Restrictions of the applicable license”. Clause 13.3.1 calls its whole list “Restrictions”, the bar on “misleading or deceptive advertising” included. Whether the sum covers those bars for actor videos is unclear.

Arcads’ 13.3.1 also bars videos in content that “infringes the rules of ads platforms”, “infringes the rules of influence” or “requires the addition of compulsory information (e.g. sponsored content)”. TikTok rejects or restricts undisclosed AI ads (above), so on our reading an actor video with no label or disclaimer falls within the first bar.

Whether a label the law requires counts as “compulsory information” is unclear (question 2 below). Suppose it does, and New York’s or California’s law reaches a licensed replica (question 1). Then an actor video carrying that label falls within the bar, and the 13.3.2 penalty could follow if it covers 13.3.1.

So an Arcads actor video on TikTok falls within 13.3.1 without a label, on our reading, and may with one (unclear, question 2).

Under Maxfusion’s terms (19 July 2026, 13.3), each prohibited use “incurs liquidated damages” of 500,000 euros “per violation”. The prohibited uses include “deceptive advertising” and any video used “contrary to applicable advertising-platform rules”. On our reading, the same unlabelled actor video on TikTok falls within that last bar too.

Maxfusion’s terms set no period to fix a breach. For unauthorised use of a Video Model its 500,000 euros are “payable on demand and without prior notice” (13.2).

MakeUGC’s Fair Use Policy says the platform generates “avatar-led testimonials” (section 1), then forbids using outputs to “fabricate testimonials” (section 7). On our reading, an avatar claiming use nobody had is what section 7 bars.

Creatify’s policy adds that “Changes are effective immediately upon posting”. On our reading, a live ad has to meet each new version from the day it is posted.

The AI UGC hub sets these tools side by side on price and ad rights. The rights matrix quotes each plan’s clauses for Arcads and Maxfusion.

Questions to take to a lawyer

1. Do New York’s “natural performer” and California’s “natural person” tests cover a licensed replica of a real actor, or only a fully synthetic one?
2. Would a legally required AI label, or TikTok’s AIGC label, count as “compulsory information” under Arcads’ 13.3.1, which also bars content that “infringes the rules of ads platforms”? And does 13.3.2’s lump sum apply to a 13.3.1 breach?
3. Does a clear disclosure that the speaker is not an actual consumer (255.2(c)) keep a first-person AI script outside 465.2?
4. How far do Arcads’ “all supports” and Maxfusion’s 48-hour clauses reach: a client’s ad accounts, archived ads, affiliates’ reposts?
5. Is a replica of a real actor a deep fake under Article 3(60) when viewers do not know the person?
6. Does a freelancer who makes an ad but does not run it “create and cause to be published” it under California’s law? Under New York’s, does the law’s “agent or employee thereof” bring an agency or freelancer that makes the ad for the business into the disclosure duty?