Are AI UGC Ads Legal? What the Rules Say in 2026

"Are AI UGC ads legal?" sounds like one question. It is two, and the AI part is the easier one.
The first question is whether your ad presents a generated person as a real customer. The second is whether viewers are told the person is generated. Different laws answer each, and most advice online blurs them into "it depends".
This guide covers US federal rules, New York and two more states with dates coming up, the EU AI Act and three platforms' ad labels. It does not cover every state or country, and some of it is unsettled. Where it is, I say so. It is not legal advice.
The short answer
Using an AI presenter in an ad is not banned by the FTC. Having that presenter claim an experience no customer had is where the exposure sits, and an AI label does not fix that. Separately, New York and the EU now ask you to tell viewers when a realistic generated person appears, Pennsylvania and California follow from December 2026 and January 2027, and TikTok and YouTube have their own disclosure controls for ads.
Question 1: is the person presented as a real customer?
The FTC's rule on consumer reviews and testimonials is 16 CFR Part 465(opens in new tab). It took effect on October 21, 2024(opens in new tab). Section 465.2 makes it a violation for a business to write, create or sell a consumer testimonial that materially misrepresents, expressly or by implication, that the testimonialist exists, that they used the product, or what their experience was.
The rule's definition of a consumer testimonial includes "depictions of the name, signature, likeness, or other identifying personal characteristics of an individual" that consumers are likely to believe reflects a real customer's experience. So a generated face saying "I've used this for three weeks and my skin cleared up" can be a testimonial in the rule's sense. The FTC's press release(opens in new tab) names this case: testimonials "by someone who does not exist, such as AI-generated fake reviews". It also says the rule lets the agency seek civil penalties against knowing violators. I found no per-violation amount worth quoting, because the figure is adjusted over time. Check the current one.
What the FTC says about AI avatars
The FTC's own questions and answers on the rule(opens in new tab) cover AI stock avatars (the example it uses is TikTok's Symphony Avatars) directly. The FTC does not address a custom generated actor by name, so applying the answer to one is an extension, though a natural one. The rule "has no blanket prohibition on the use of AI-generated avatars in marketing". An avatar's statements might be a testimonial, and that would be prohibited "only if the underlying testimonials were fake or false". The same answer adds that using an avatar could still be deceptive under the FTC Act.
The generated presenter is allowed. A made-up customer story is not, and an AI label on the video does not make the story true.
The Endorsement Guides say the same thing earlier
The Endorsement Guides, 16 CFR Part 255(opens in new tab), are guidance, not a statute, so a departure from them is not automatically a violation. But the FTC uses them to judge what is deceptive. Three passages matter here.
Section 255.2(c)(opens in new tab): advertisements presenting endorsements by what are represented, expressly or by implication, to be "actual consumers" should use actual consumers in audio and video, or "clearly and conspicuously disclose that the persons in such advertisements are not actual consumers of the advertised product".
Section 255.1(c)(opens in new tab): when an ad represents that the endorser uses the product, the endorser must have been a bona fide user when the endorsement was given.
Section 255.2(b): a real customer's result can still mislead if it is not what people generally get. The advertiser needs evidence that the result is typical, or must clearly disclose what viewers can generally expect.
A generated presenter cannot be a bona fide user. So a line like "I've used this for three weeks" needs a real customer who had that experience. If no customer did, cut the line. The disclosure in 255.2(c) only tells viewers who is speaking.
What works instead: give the presenter a brand message you can back. If the presenter reads a genuine customer's words, verify the experience, get the customer's permission, keep the meaning, and say what the viewer is watching, for example "AI-generated presenter reading a real customer's words." Disclose any payment or free product the customer got that viewers would not expect. A result claim still needs evidence, whoever reads it.
One more check belongs to the UGC format itself. It is built to look like a post. The FTC's guide on native advertising(opens in new tab) asks that viewers can tell an ad is an ad. In the EU, the Unfair Commercial Practices Directive(opens in new tab) bans a trader posing as a consumer (Annex I, point 22) and submitting or commissioning false consumer reviews or endorsements to promote products (point 23c).
Who is on the hook: the Rytr order
In December 2024 the FTC finalized an order against Rytr, which sold an AI writing service that let subscribers generate fake reviews. The order banned Rytr from offering any AI service that generates reviews or testimonials. On December 22, 2025(opens in new tab) the FTC reopened and set the order aside, saying the complaint did not meet the FTC Act's requirements and that "condemning a technology or service simply because it potentially could be used in a problematic manner" is inconsistent with the law. In the same release it said it will keep holding accountable "actors that use AI to violate the law or deceive consumers about the capabilities of their generative AI", and added that Rytr was not such a case. The case page(opens in new tab) holds the documents.
The order fell on the facts alleged in that complaint. It does not say a tool can never be responsible. Responsibility depends on what each party did. Section 465.2(a)(opens in new tab) reaches any business that writes, creates or sells a false testimonial. Section 465.2(b) reaches a business that disseminates a testimonial about itself or its products that it knew or should have known was false. The advertiser should verify the claims and the disclosures. An agency or a service that writes the story, makes the video or arranges its distribution can face its own exposure. This post sits on the blog of a company that makes such a tool, which is one more reason to say that plainly.
Question 2: are viewers told the person is generated?
New York
New York's synthetic performer law, S.8420-A / A.8887-B(opens in new tab), amends General Business Law section 396-b. The bill history shows it signed on December 11, 2025 as chapter 617, effective 180 days later. The Governor's office announced it in effect on June 9, 2026(opens in new tab).
What the current text of section 396-b(opens in new tab) says:
- A "synthetic performer" is a digitally created asset, made with generative AI or a software algorithm, intended to create the impression of a human performer who is not recognizable as any identifiable natural performer. A generated actor fits that description.
- A person in the business of dealing in property or services who produces or creates an advertisement for a commercial purpose must "conspicuously disclose" the synthetic performer in it, where they have actual knowledge.
- The civil penalty is $1,000 for a first violation and $5,000 for each later one.
- Audio-only ads are excluded, and so is AI used solely to translate a human performer. Subdivision 4 excludes ads for certain expressive works, under conditions. Subdivision 8 excludes a medium that carries the ad as a distributor, so the duty sits with whoever produces or creates the ad.
What is unsettled: the text I read does not define "conspicuously", and it does not spell out how far it reaches for an ad produced outside the state. One law firm's reading(opens in new tab) is that it covers visual and audiovisual ads distributed to a consumer located in New York. Treat that as a reading until a court or the state says so.
California and Pennsylvania
Two more states have passed laws that reach generated people in ads. California's SB 1050(opens in new tab) requires a disclosure in an advertisement that prominently features a synthetic performer, and unlike New York's law it can cover a synthetic voice. It was signed in September 2026 and prints no date of its own, so under California's default rule for new statutes it takes effect on January 1, 2027. Pennsylvania's Act 99 of 2026(opens in new tab) takes effect on December 21, 2026. It covers qualifying synthetic advertising content that creates a false perception about the nature, origin or authenticity of the goods or services advertised.
A campaign launched in November that runs through the winter will still be live when both apply. I have not surveyed all fifty states.
EU AI Act, Article 50
Article 50(opens in new tab) applies from 2 August 2026 (per Article 113(opens in new tab)). Two paragraphs matter for ads:
- 50(2), providers: whoever builds a system that generates synthetic video or audio must make the outputs machine-readable and detectable as artificially generated, with technical solutions that are effective, interoperable, robust and reliable as far as technically feasible. This is the tool's duty, not yours. A later amendment, Regulation (EU) 2026/1744(opens in new tab), gives providers of systems already on the market before 2 August 2026 until 2 December 2026 to comply with this paragraph.
- 50(4), deployers: whoever uses an AI system to generate or manipulate image, audio or video content that constitutes a deep fake must disclose that it is artificially generated or manipulated. Paragraph 5 says the disclosure must be clear and distinguishable, at the latest at the first exposure. Fines for Article 50 breaches reach EUR 15 million or 3% of worldwide annual turnover for a company, whichever is higher(opens in new tab) (Article 99(4)). For SMEs, including start-ups, Article 99(6) applies whichever of the two figures is lower, and Article 99(6a) extends that to small mid-cap companies.
The Act 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". Read alone, that seems to leave a wholly invented presenter out. The Commission's guidelines on Article 50(opens in new tab) do not. They include realistic synthetic people who could plausibly exist, and give the example of a synthetic influencer demonstrating a sponsor's product. When the content qualifies, the business using the AI system must clearly disclose its artificial origin by the viewer's first exposure. The test is the whole scene, its context and what the audience would reasonably take to be real, not whether a real model sat for it. The Commission also answers common questions in its Article 50 FAQ(opens in new tab).
Platforms
Platform rules are contracts, not laws, but they decide whether your ad runs. For a paid ad, the advertiser pages apply, not the creator upload settings.
| Platform | What the page says | Source |
|---|---|---|
TikTok | For non-Spark ads with fully generated or substantially AI-edited material, use the AI disclosure control in Ads Manager where it is offered. The ad policy also describes clear disclosures in the creative itself. For Spark Ads, the disclosure follows the underlying post | TikTok Ads Manager help, TikTok ad policy |
YouTube | For Promote, review the ad's own label when you create the campaign. A disclosure set at upload may not carry over. Google Ads has asset-level disclosure settings | YouTube Promote help, Google Ads help |
Meta | Ads made or edited with some of Meta's own generative AI tools get an "AI info" label automatically. Meta says it is beginning to detect ads made with third-party AI tools through industry-standard signals, such as C2PA metadata, and label them | Meta newsroom |
Read the current policy for your placement before you launch, and check where viewers will actually see the label. If the platform has to detect a file mark to apply the label, you do not control it.
Where Clipwright.io fits
The person in a Clipwright.io presenter video is generated, so the ad must not pretend to be a real customer's testimonial. Write scripts in the brand's voice ("we make", or a real quote from a real customer who agreed), not "I switched last month and loved it."
Clipwright.io's AI disclosure page(opens in new tab) is plain about the split. Every completed video run includes an AI disclosure in the API response. Clipwright.io embeds a disclosure in supported files and flags a skipped tag on current runs. Re-encoding may remove an embedded tag, and the file carries no visible caption and no C2PA manifest. The visible label on the published ad is yours to add.
Clipwright.io can also animate a photo you upload, which is a different case. A generated actor is not drawn from a real person; an uploaded photo is. The Acceptable Use Policy(opens in new tab) prohibits showing a real, identifiable person speaking your script without their consent, given knowingly and in advance, and lets you upload an image only if you hold the rights for this use. My advice goes further than the policy: get the image rights and the person's written permission covering the AI-generated performance and its advertising use. In New York, Civil Rights Law section 50(opens in new tab) already requires a living person's written consent to use their name, portrait, picture, likeness or voice in advertising. The same policy also prohibits presenting a generated speaker as a doctor, an election authority or a licensed adviser, and fabricated testimonials about earnings.
A pre-flight check for every AI UGC ad
- Whose experience does the script claim? If the presenter says "I used it", a real customer must have had that experience. If no one did, cut the line. An AI label does not fix it.
- Is the presenter reading a real customer's words? Verify the experience, get permission, keep the meaning, and say so on screen, for example "AI-generated presenter reading a real customer's words". Otherwise write in the brand's voice.
- Would viewers recognise this as an ad when they first see it? If the commercial purpose is unclear, name the ad and the brand prominently.
- Is the label in the video, or only in the file? Put a visible "AI-generated" line in the picture, and also use the ad platform's own disclosure control where it has one.
- Who sees it, and when? New York and the EU require a label now. Pennsylvania's law applies from December 21, 2026 and California's from January 1, 2027. If you cannot target around them, label everywhere.
- Is it a real person's face or voice? If you animate a photo of a real person, get the image rights and their written permission for the AI performance and its advertising use.
- Is there a claim the presenter cannot back? Health, income and results claims need substantiation whoever says them. A generated white coat does not help.
FAQ
Is it legal to use an AI avatar in an ad?
Yes, in general. The FTC's own answer, given about AI stock avatars, is that its fake-testimonial rule has no blanket prohibition on AI-generated avatars. The problems are false testimonials, deception under the FTC Act, and in some places a missing disclosure. An AI label does not make an invented customer experience acceptable.
Do I have to label an AI UGC ad?
In New York, the law now requires a conspicuous disclosure for a synthetic performer in an advertisement. The EU AI Act requires disclosure for deep fakes from 2 August 2026, and the Commission's guidelines include realistic synthetic people. Pennsylvania's law applies from December 21, 2026 and California's synthetic performer disclosure from January 1, 2027. TikTok and YouTube have their own disclosure controls for ads. If you sell into those places, label.
Is the tool or the advertiser responsible?
It depends on what each party did. The EU puts marking duties on the tool's provider and disclosure duties on the deployer. In the US, the FTC set aside its Rytr order on the facts alleged in that case, but its rule still reaches any business that writes, creates or sells a false testimonial, and a business that disseminates one it knew or should have known was false. The advertiser should verify the claims and disclosures. An agency or service that creates a false testimonial can face its own exposure.
Sources
- FTC, 16 CFR Part 465 and section 465.2: https://www.ecfr.gov/current/title-16/chapter-I/subchapter-D/part-465/section-465.2(opens in new tab)
- FTC press release, August 2024: https://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials(opens in new tab)
- FTC rule questions and answers: https://www.ftc.gov/business-guidance/resources/consumer-reviews-testimonials-rule-questions-answers(opens in new tab)
- FTC Endorsement Guides, 16 CFR Part 255: https://www.ecfr.gov/current/title-16/part-255(opens in new tab)
- FTC native advertising guide: https://www.ftc.gov/business-guidance/resources/native-advertising-guide-businesses(opens in new tab)
- FTC, Rytr case page: https://www.ftc.gov/legal-library/browse/cases-proceedings/232-3052-rytr-llc-matter(opens in new tab)
- New York General Business Law section 396-b: https://www.nysenate.gov/legislation/laws/GBS/396-B(opens in new tab)
- New York Civil Rights Law section 50: https://www.nysenate.gov/legislation/laws/CVR/50(opens in new tab)
- California SB 1050: https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202520260SB1050(opens in new tab)
- Pennsylvania SB 806 (Act 99 of 2026): https://www.palegis.us/legislation/bills/2025/SB0806(opens in new tab)
- EU AI Act Article 50: https://artificialintelligenceact.eu/article/50/(opens in new tab)
- Regulation (EU) 2026/1744: https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng(opens in new tab)
- European Commission, guidelines on Article 50: https://digital-strategy.ec.europa.eu/en/library/guidelines-transparency-obligations-providers-and-deployers-ai-systems(opens in new tab)
- EU Unfair Commercial Practices Directive 2005/29/EC: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A02005L0029-20260927(opens in new tab)
Checked against these sources on 18 November 2026. Laws and policies change, and this is not legal advice. For a campaign with real money behind it, ask a lawyer who knows advertising law in the markets you sell to.
If you are building the ad with an agent, the product on screen guide and the live shots guide cover making the video itself.
Your API result says it is AI. Your viewer cannot read it.
A machine-readable mark is not a visible label, and the laws above are written for viewers. The visible label is yours to add at publish. See what Clipwright.io marks and what it leaves to you.
Read what the file carriesAbout the Author
Founder at Dimantika
Co-founder of Dimantika, the company behind Clipwright.io and ViralFaceless. Builds both with coding agents. Previously ran GlockSoft with a partner for about 15 years. Writes here about handing video production to agents.
View all postsMore articles you might like.

How to find viral reels for ads when views lie
The biggest reel in your niche is often an ordinary day for a big account. Find the one that beat its own creator, check it's still gaining, and build your ad on how it's made.

How to create a video ad with AI from one brief
How to create a video ad with AI where you approve every step: a short brief, a reel that took off, a beat plan, your notes, then the price before it runs.

Put the product on screen in your AI UGC video ads
Most AI UGC video ads hand back a talking head and leave the product for an editor. Templates put a before-and-after, a number and an end card on the words that sell, in the same request.
