AI Disclosure Rules for Ads: What Agencies Actually Have to Do Now

If you have used an AI-generated voice, face, or full performer in a client's video or audio ad this year, you need to know whether the law requires you to say so. As of September 2026, the answer is yes in two states, and yes federally too, depending on what the ad actually shows and who is watching it. California just became the second state to pass a synthetic performer disclosure law, New York's version has already been live since June, and the FTC has its own rule running underneath both of them that most shops still aren't following correctly. None of this is theoretical anymore. It's dated, it's enforceable, and it applies whether your agency is in Sacramento or Singapore, as long as the ad reaches a consumer in one of these states.

What Actually Counts as a "Synthetic Performer" (and What Doesn't)

Both state laws use a version of the same core definition: a synthetic performer is a digitally created figure, voice, or representation, built in whole or in part with generative AI, that creates a realistic impression of a human performance by someone who isn't a real, identifiable person. That last clause matters more than it looks. These laws are not about deepfakes of real people, that's a separate area of law built on right-of-publicity claims. They're about the AI-generated spokesperson, host, or voiceover artist who never existed at all.

That framing also tells you what's out of scope. Neither law touches AI-generated backgrounds, product renders, or scenery. Nobody's flagging your AI-cleaned-up storefront shot or your generated product spin. The trigger is specifically a performer: a face, a body, or a voice standing in for a human doing the selling. If your team runs AI upscaling or color correction on footage of an actual actor, that's not a synthetic performer either. The law cares about AI creating the performance, not AI touching up a real one.

Where it gets more specific is audio. New York's law only covers visual and audiovisual ads. California's covers audio-only ads too, which closes a gap agencies have been quietly using: an AI voice reading a script for a radio spot or a podcast pre-roll never had to disclose anything under New York's rule. Once California's law takes effect, that same AI voiceover triggers a disclosure requirement if the ad reaches a California listener, full stop, no picture required.

California vs. New York: Same Idea, Different Rules

Governor Newsom signed SB 1050 on September 16, 2026, with the bill's author, Senator Angelique Ashby, at SAG-AFTRA's headquarters, the union that sponsored it. It takes effect January 1, 2027. New York's law, signed by Governor Hochul back in December 2025, has been in force since June 9, 2026, as an amendment to the state's General Business Law. They read like the same bill until you put the details side by side.

New York's penalty is a flat civil fine: $1,000 for a first violation, $5,000 for each one after that. California's is lower per incident, up to $500 per violation, but the enforcement mechanism is broader. The Labor Commissioner or a public prosecutor can bring the case, and the statute also opens the door to injunctive relief and punitive damages on top of the fine, plus it bars the advertiser from continuing to run an ad once it's found in violation. So New York hits harder per infraction; California hits with more tools and can actually pull the ad.

Both laws share the "reach, not residency" rule. It doesn't matter where your agency is headquartered or where the ad was produced. If the ad is distributed to a consumer in California or New York, the law applies. An agency working entirely out of another state, or another country, is covered the moment a client's media buy includes either state in its targeting.

California carved out two exemptions worth knowing. Ads for expressive works, movies, TV, streaming series, documentaries, video games, are exempt as long as the synthetic performer's use in the ad matches how it's used in the actual work, so a movie trailer using the film's own AI-assisted footage isn't a separate violation. And AI used solely for language translation or dubbing doesn't count as introducing a synthetic performer, so localizing an existing human performance into another language isn't a trigger either.

The FTC Rule Most Agencies Miss: One Disclosure Doesn't Cover Both Things

Here's the part that catches even shops who think they're already compliant. The FTC finalized updated Endorsement Guides in early 2026 addressing synthetic media directly, and it treats commercial-relationship disclosure and AI-involvement disclosure as two separate obligations. A "#ad" tag tells a viewer this is sponsored content. It says nothing about whether the person in the ad is real. An "AI-generated" label tells a viewer the content was AI-made. It says nothing about whether money changed hands. Under the FTC's current guidance, you need both, stated separately, whenever a reasonable consumer would otherwise assume they're watching a real person's real endorsement or a real product demo.

This is a federal rule, so it applies regardless of what state the ad runs in, and it isn't limited to synthetic performers the way California and New York's laws are. It reaches AI-generated testimonials, AI-edited creator content that changes what's actually being shown, and virtual influencers presented without context. The FTC stood up a dedicated AI enforcement unit in January 2026, and the max penalty per violation is now $53,088, with each individual non-compliant post or ad counted separately. A client running the same undisclosed AI-generated testimonial across fifty ad units isn't looking at one violation.

What the Disclosure Actually Has to Say, and Where It Goes

California's statute is specific about the language: something "substantially similar to" a statement that the performance features a synthetic digital performer, placed in close proximity to the performer, not buried in a description field or a terms page. New York requires a "conspicuous" disclosure without dictating exact wording, but the practical standard courts and regulators are applying is the same: a viewer shouldn't have to hunt for it, and it has to appear at or near the moment the synthetic performer appears, not just once at the start of a longer video.

For the FTC's dual requirement, that means two visible statements, not one blended line. "Paid partnership. This performer is AI-generated" reads clean and covers both. Splitting them into "#ad" plus a separate, visible "AI-generated performer" label anywhere in the frame also works. What doesn't work: a single disclosure that only names one of the two facts, or a disclosure that only shows up in a caption when the substance of the ad is in the video itself.

Who Actually Eats the Fine: You or the Client

The statutes name "the person who creates the advertisement" alongside the advertiser, so an agency that built and placed the creative is exposed directly, not just the brand whose name is on it. This is worth putting in your contracts now rather than after a complaint gets filed. Two things are worth adding to a standard SOW or MSA: a clause specifying who's responsible for adding and verifying the disclosure on delivered creative, and a requirement that the client confirm, in writing, which elements of a campaign use AI-generated performers before the agency starts producing at scale.

In practice, the agency is usually the one who knows whether a performer is synthetic, since that's a production decision made in-house or by a vendor the agency hired. Don't assume the client's legal team is tracking this. Most aren't yet, which means the agency is the only party in the relationship who actually knows there's something to disclose.

A Compliance Checklist Before Your Next AI-Assisted Ad Goes Live

Run this before any campaign with an AI voice, face, or full performer ships:

• Confirm whether the performer is fully synthetic, partially AI-assisted, or a real actor with AI enhancement. Only the first two trigger these laws.
• Check the media plan's geographic targeting. If California or New York consumers are in scope, both state disclosure rules apply regardless of where the ad was made.
• Add two disclosures if the content is both sponsored and AI-generated: one for the commercial relationship, one for the AI involvement. Don't combine them into a single ambiguous line.
• Place the disclosure at or near the performer's appearance, not only in a caption or a linked landing page.
• Confirm the contract states who's responsible for adding and verifying the disclosure, and get the client's sign-off on which assets use synthetic performers before production starts.
• If the ad is audio-only, don't assume it's exempt. California's law reaches it even though New York's doesn't.

More States Are Writing This Bill Right Now

New York went first, California just followed with a broader version that also catches audio, and neither state legislature is done. Bills covering the same ground have already been introduced or drafted in several other states this year, largely modeled on whichever of these two laws passed most recently in that state's legislative session. The pattern so far is states expanding scope each time, not narrowing it, and the FTC's own enforcement numbers are moving the same direction: ad enforcement cases were already up 40% in 2025, before the dedicated AI unit even existed.

The agencies that come out ahead here aren't the ones scrambling to retrofit disclosures onto a live campaign after a complaint lands. It's the ones treating "is this performer synthetic, and does this client know" as a standard production question, asked and answered before a shoot, the same way you'd already ask about usage rights or music licensing. That's a five-minute conversation now. It's a much longer one with a client's legal team later.

FAQ

Do I have to disclose an AI-generated voice in an ad?

In California, yes, once SB 1050 takes effect on January 1, 2027, because California's law covers audio-only ads that reach a California listener. New York's law only covers visual and audiovisual ads, so an AI voice in a radio spot or podcast pre-roll doesn't trigger it, although the FTC's separate rule can still apply if listeners would assume they're hearing a real person's endorsement.

Do AI-generated product images or backgrounds need an ad disclosure?

Not under the California and New York synthetic performer laws. Both are triggered only by an AI-generated performer, meaning a face, body or voice standing in for a human, so AI backgrounds, product renders, scenery and AI touch-ups of footage of a real actor are out of scope. The FTC rule is broader, and an AI-made product demo a viewer would take as real still needs an AI label.

Is "#ad" enough when an ad uses an AI-generated person?

No. The FTC treats the commercial relationship and the AI involvement as two separate disclosures, so a sponsored ad with an AI-generated performer needs both, for example "Paid partnership. This performer is AI-generated." A single label that names only one of those two facts doesn't cover it.

Do the California and New York AI ad laws apply to agencies in other states?

Yes. Both laws follow where the ad is distributed, so an agency based in another state or another country is covered as soon as a client's media buy reaches consumers in California or New York. The statutes also name the person who creates the advertisement alongside the advertiser, which puts the agency that built the creative on the hook directly.

What is the fine for not disclosing an AI performer in an ad?

New York charges $1,000 for a first violation and $5,000 for each one after that. California allows up to $500 per violation plus injunctive relief, punitive damages and an order to stop running the ad, and the FTC's maximum is $53,088 per violation, with each non-compliant post or ad counted separately.