AI advertising disclosure: what you have to label

This is general information as of November 2026 and not legal advice. I'm a producer who has to get this right on real projects, so I read the sources linked here, and I'd still have a lawyer check any campaign with money behind it. These rules are new and some of their edges haven't been tested.

The rules side by side

FTC: the question is deception

I couldn't find an FTC rule requiring a label on AI ads as such. What applies is the general ban on deceptive practices, plus specific rules that reach AI. The rule on fake reviews and testimonials covers AI-generated fakes. The FTC's Q&A on that rule says it has no blanket prohibition on AI avatars in marketing, and that a stock avatar isn't giving a consumer review as the rule defines one. The same page warns that using actors to portray testimonialists could still violate the FTC Act, while an obviously fictional dramatization involves no testimonial at all.

The way I read it for production: a generated person holding your product in a clearly staged scene is one thing. A generated "customer" describing results, or a generated before-and-after, is another. A small "made with AI" line wouldn't make a fabricated result true.

New York: synthetic performers

General Business Law section 396-b was signed in December 2025 and, according to Cooley's summary, took effect on June 9, 2026. It requires a conspicuous disclosure when someone producing an ad for a commercial purpose has actual knowledge that it contains a synthetic performer. Penalties are $1,000 for a first violation and $5,000 for later ones.

A synthetic performer, under the statute, is a digitally created asset meant to look like a human performance by someone who isn't recognizable as any identifiable real person. The details that matter when you're making the ad:

• A generated stranger counts. A digital double of a real, identifiable person falls outside this definition and into consent law, including New York's Civil Rights Law section 50, which requires written consent for advertising use, from a parent or guardian for a minor.
• Audio-only ads are exempt. So is AI used only to translate a human performer's language.
• Ads for expressive works such as films and games are exempt if the performer is used as it is in the work.
• Media outlets that merely carry the ad are exempt, so the duty stays with those who make it.
• "Conspicuous" isn't defined. Practice will have to settle what size and duration are enough.

I'd raise two gaps with counsel. Cartoons: the definition speaks of a human performance, so a talking hen seems to sit outside it, while a stylized human is harder to call. And reach: the sources I read address companies advertising in New York and don't say how the law treats a national social campaign. I'd assume a paid US campaign that reaches New York needs the disclosure.

EU: deep fakes, from August 2026

Article 50(4) of the AI Act requires deployers of AI that generates a deep fake to disclose that the content is artificially generated or manipulated. It has applied since August 2, 2026. Under paragraph 5 the information must be clear and distinguishable by the first exposure. Article 2 reaches deployers outside the EU when the output is used in the Union, and it excludes purely personal, non-professional use.

Paul Weiss's summary of the EU guidelines describes the test as objective: intent to deceive isn't needed, photorealism isn't decisive, and the question is whether the subject is plausibly real. Wholly fantastical content falls outside. For evidently artistic or creative work the duty shrinks to disclosing that AI was used, in a way that doesn't spoil the work. The same summary says a provider's hidden watermark doesn't discharge the deployer's duty, that the Commission has three EU icons deployers may use, and that fines can reach €15 million or 3% of worldwide turnover.

That memo doesn't discuss advertising directly. A realistic generated person in a commercial looks to me like a clear case for the plausibly-real test, and whether an ad can claim the creative-work allowance is a question for a lawyer.

Platforms

TikTok

The ad policy allows AI-generated or significantly edited content if it carries the AIGC label or your own clear disclaimer, caption, watermark or sticker. Undisclosed AI content gets the ad rejected or restricted. TikTok's AI content page also says it doesn't allow content containing the likeness of people under 18, which matters for any AI work based on a real child.

Meta

Meta does the labeling. Its help page says AI info appears in the three-dot menu, and next to "Sponsored" when an image made with its tools includes an AI-generated photorealistic human. Its transparency post says Meta is moving to detect and label third-party AI tools automatically, while that help page still lists third-party tools as a case without the label, so the two are out of step. Advertisers must disclose digitally created or altered media in ads about social issues, elections or politics. I found no Meta page requiring self-disclosure on an ordinary commercial ad, and blog posts on the subject disagree with each other.

YouTube

YouTube's rule is written for uploads: realistic altered or synthetic content has to be disclosed at upload, while clearly unrealistic content and minor aesthetic edits don't. The page doesn't address ads, so a brand publishing the film to its channel should use the setting and check Google's ad policies separately for paid placements.

How I'd handle it on a campaign

On my site every AI work carries a caption that says so. That's the easy case. For a client ad, here's how I'd run it:

1. List every territory and platform in the brief, before the script. The answer changes with both.
2. Note whether a realistic human appears, and whether that human resembles anyone real. If yes, get written consent first.
3. Strip out generated testimonials, results and demonstrations. Those need real evidence.
4. Write into the agreement who adds the disclosure and who approves its wording. Cooley's practical advice is to allocate this with agencies, studios and AI vendors, and it's a one-line job at the start.
5. Design the label into the frame. I hold key on-screen text for at least two seconds, and I'd treat a disclosure the same way.
6. Check it in every cutdown. A line placed for 16:9 can be cropped out of the 9:16 version.
7. Keep a record of which AI tools made which shots, so that "actual knowledge" isn't a guess later.

If you'd like an ad made with this built in from the first draft, that's part of the AI film work I do.