Two separate disclosure questions apply to AI-touched sponsored content, and conflating them is the actual compliance risk. If the endorser itself isn't a real person — a fully AI-generated persona, a cloned voice reading a script no one spoke, a virtual influencer — both the FTC's endorsement framework and, for ads reaching New York, a dedicated state law require a disclosure that the performer isn't real, separate from any paid-partnership tag. If the endorser is a real creator whose footage was AI-edited — retouched, background-swapped, dubbed — the standard #ad disclosure still covers you in most cases, and only tips into identity-disclosure territory once the edit changes what a viewer would believe actually happened. Ask "is the performer real?" and "was this paid?" as two separate yes/no questions, because that's exactly how regulators are drawing the line in 2026.
Which disclosure rule actually applies here?
Two regimes now sit on top of the disclosure rules practitioners already know. The FTC's updated endorsement guidance, operationalized in May 2026, rests on two pillars: material-connection disclosure (you paid for this endorsement — the #ad rule you already follow) and identity disclosure (the endorser is not a real person). The identity pillar is the new one, and it sweeps in virtual influencers, AI-generated personas, and any content built to look like it comes from a real endorser who doesn't exist.
New York layers a second, narrower rule on top for anyone whose ads reach the state: the Synthetic Performer Disclosure Law, in effect since June 9, 2026, requires a clear and conspicuous disclosure inside any advertisement — social, paid digital, video, influencer content, TV, display — that features an AI-generated synthetic performer. First violation: $1,000. Each one after that: $5,000. It's a narrower rule than the FTC's — it only covers fully synthetic performers, not garden-variety editing — but it applies based on the audience the ad reaches, not the advertiser's location, so a nationally distributed campaign reaching New York consumers is in scope regardless of where your brand is based.
Practically: run the material-connection check you already run (is this paid, gifted, or otherwise incentivized?), then add a second, independent check (is the performer a real person, edited lightly enough that a viewer's belief about who's actually speaking wouldn't change if they knew the truth?). A single post can need one disclosure, both, or — for an unpaid organic post from a fully synthetic brand persona — just the identity one.
When do you need a synthetic-performer disclosure, not just #ad?
The identity disclosure applies whenever the entity appearing to endorse the product wasn't the one who actually said or did it. In practice, that covers:
- A fully AI-generated virtual influencer posting sponsored content — even with a consistent name, face, and following, the audience is being asked to trust a testimonial from someone who doesn't exist.
- Voice-cloned narration standing in for a real spokesperson who never recorded those words — the audience hears a familiar voice and assumes the person actually said it.
- AI-generated "customer" testimonials that map to no real customer. The FTC has named this pattern a priority target under its Operation AI Comply enforcement track, and each non-compliant post counts as its own violation, with a 2026 inflation-adjusted maximum civil penalty of roughly $53,088 per violation.
- A synthetic performer delivering claims a real creator or spokesperson is contractually attached to, without disclosing that the delivery itself is AI-generated.
None of this requires the content to look obviously fake. The identity disclosure exists precisely because well-made synthetic content doesn't announce itself — the audience has to be told.
Does AI-editing a real creator's footage need extra disclosure?
Most of it doesn't. Color grading, background blur, noise cleanup, standard retouching, and — notably — AI used solely to translate a real performer's own words into another language are all excluded from New York's synthetic-performer rule by name, and none of it trips the FTC's identity pillar either: the person on screen is still the person who actually said and did what's shown.
The line moves once the edit changes what actually happened, not just how it looks. Two examples worth separating in your own head:
- A creator runs an AI voice-cleanup pass on their own audio, then posts the same words they actually said. Standard #ad disclosure covers it.
- A brand uses AI to redub a creator's voice into different phrasing, adjusts their mouth movements to match, or extends a clip with AI-generated frames the creator never filmed. The person on screen is real, but what's shown didn't happen the way it's shown happening — and that crosses into the same territory as a synthetic performer, even though a real creator's likeness is involved.
The practical test: would a viewer's belief about what actually occurred change if they knew exactly what was edited? If yes, disclose the AI involvement specifically — not just the paid partnership.
What does a compliant disclosure look like, and where does it go?
"Clear and conspicuous" is the bar both rules use, and it rules out a few habits that are still common:
- Buried in a caption's 15th hashtag, or only in a link-in-bio document — not conspicuous.
- A one-time disclosure on a landing page or press release meant to cover every post in a campaign — the rule attaches per post, not per campaign.
- A generic #ad tag standing in for identity disclosure — the two are separate requirements, and neither substitutes for the other. A synthetic-performer post that's also paid needs both a paid-partnership label and a synthetic/AI-content disclosure.
What holds up: on-screen overlay text or a spoken disclosure inside the video itself (not just the caption), a platform's native paid-partnership label plus an explicit AI/synthetic tag when applicable, and disclosure language that states plainly what a viewer needs to know — that the performer is AI-generated, or that the content was AI-edited in a way that changes what's shown — rather than a symbol or abbreviation a general audience wouldn't parse. This isn't the first compliance shift built around exactly this gap: Meta's Partnership Ads mandate already treats an undisclosed paid connection as a "Deceptive Practice," not a formality, and undisclosed AI content is one of the categories ad-review teams flag most often.
What changes in your briefs and contracts starting now?
- Add one question to every content brief: will any AI post-production touch this creator's face, voice, or the sequence of events, beyond basic color and audio cleanup? Get a yes/no per deliverable before filming, not after.
- Write a disclosure clause into the contract: the creator or vendor must flag any voice cloning, face/mouth editing, frame extension, or synthetic elements before the asset is delivered, so the brand can attach the right disclosure layer before it goes live.
- Name an accountable reviewer on the brand side, not just the agency's, for AI-disclosure sign-off before publish — the FTC's enforcement sweep reaches brands and agencies, not only whoever generated the content.
- If a campaign uses a fully virtual influencer, draft the exact identity-disclosure wording into the brief itself rather than leaving it to the vendor's discretion.
- Keep a lightweight internal log of which posts used AI post-production and what disclosure ran on each — cheap insurance against the per-post violation math above, and the first thing you'll want on hand if a platform or regulator asks. Add it alongside the contract clauses practitioners keep missing, before the next round of signatures, not after.
The compliance work here is really an inventory problem: you can't disclose what you don't know is AI-touched, and by the time a platform flags a post or a regulator asks, scattered DMs and file names won't answer fast enough. Hyperstar keeps every creator's deliverables, usage rights, and campaign notes in one record instead of ten threads — so when a disclosure question comes up, you already have the answer. Want to see what that looks like for your current roster? Get started.
