AI UGC disclosure rules in 2026: FTC, New York, the EU, Meta, TikTok, and YouTube

This is not legal advice, and the rules move

This guide summarizes the public rules as we read them in September 2026, with links to the primary sources so you can check them yourself. It is not legal advice. Advertising law depends on where you operate, where your audience is, and what your ad claims, so confirm the specifics with counsel before a campaign that matters.

The good news is that the rules converge on one idea: do not let anyone believe a synthetic performer is a real person with real experience. Build your creative around that and most of the rest is paperwork. The bad news is that the paperwork now has teeth in several jurisdictions at once, and the platform toggles that people assume cover them mostly do not.

United States: the FTC

The FTC's Endorsement Guides were revised on June 29, 2023, and the revision explicitly clarified that the definition of an endorsement covers virtual influencers and fabricated endorsers. In plain terms, a computer-generated creator who recommends your product is held to the same standard as a human who does: claims must be truthful and substantiated, and material connections must be disclosed clearly and conspicuously.

Separately, the FTC's rule on fake reviews and testimonials, announced August 14, 2024 and in effect since October 21, 2024, prohibits testimonials from people who do not exist, including AI-generated ones, and from people who did not actually use the product. Civil penalties are assessed per violation and were above fifty thousand dollars per violation when the rule took effect; the figure is adjusted annually.

The Guides also define what clear and conspicuous means. A disclosure has to be hard to miss, in a simple font with contrasting background, and understandable to an ordinary reader. For video, the FTC's guidance favors simultaneous written and spoken disclosure when the endorsement itself is both seen and heard. A line in the caption that a viewer has to tap to expand does not meet that bar.

New York: the synthetic performer law

On December 11, 2025, New York's governor signed an amendment to General Business Law section 396-b that takes effect June 9, 2026. It requires anyone who produces or creates an advertisement for a product or service, for a commercial purpose and with actual knowledge, to conspicuously disclose that a synthetic performer appears in the advertisement. Reported penalties are $1,000 for a first violation and $5,000 for each subsequent one.

Law-firm summaries note that the definition targets digitally created or altered human likenesses that a reasonable viewer would take for a real person, and that some categories are carved out. If your ads can reach New York, and most paid social ads can, the safe assumption is that a visible on-screen disclosure belongs on any creative with an AI-generated human. The summaries linked below walk through the definitions and the exemptions in more detail.

Other states have moved on adjacent issues, including digital replicas of real people and deceptive synthetic media in political advertising. Those are narrower than New York's rule, but they add to the case for treating on-screen disclosure as a default rather than a jurisdiction-specific exception.

European Union: transparency obligations

The EU AI Act's transparency obligations, including the requirement to disclose artificially generated or manipulated image, audio, and video content that resembles real people, places, or events, apply from August 2, 2026. Obligations fall both on providers of AI systems, who must mark outputs in a machine-readable way, and on deployers, who must disclose deepfake content to the people who see it. An advertiser running synthetic-performer creative in an EU market is a deployer.

If you run creative in EU markets, treat a visible synthetic-content disclosure as a baseline there as well, and confirm the implementation details for your member state. National advertising standards bodies are layering their own guidance on top of the Act.

Platform rules

Platform labels do not replace legal disclosures. A label hidden behind a three-dot menu is not a conspicuous disclosure under New York's law, and it says nothing about whether your claims are substantiated. Think of the platform toggle as the floor and the on-screen text as the actual compliance.

What a compliant disclosure looks like

There is no mandated wording in most of these rules, only a standard: conspicuous, understandable, and present where the viewer is looking. The following patterns meet that standard in practice. Adapt the wording to your counsel's preference.

• On-screen text in the first three seconds and again at the close: "AI-generated presenter" or "This video features a synthetic performer."
• A persistent small caption for the full runtime: "AI presenter" in a corner, readable at phone size.
• For paid partnerships, a separate line: "Paid partnership" or the platform's built-in paid partnership tag. Do not merge the two disclosures into one.
• In the caption, repeat both: the AI disclosure and the paid relationship.

A working checklist

1. Decide what the AI creator is. Spokesperson, presenter, or demonstration are fine. Customer is not.
2. Strip testimonial language from AI scripts. No first-person experience claims, no before and after stories the character did not live.
3. Substantiate every number and claim exactly as you would for a human creator. Health, financial, and environmental claims carry their own rulebooks.
4. Add a visible on-screen disclosure that the performer is AI-generated, in a font and position a viewer can actually read, and keep it up long enough to be read.
5. Add the paid-partnership or advertising disclosure separately. Two facts, two disclosures.
6. Use the platform's AI disclosure setting wherever one exists, and never remove content credentials.
7. Keep the reference footage, character source, and rights attestations for every render. If a question comes up later, the trail is your answer.
8. Re-check this list quarterly. Every rule above changed in the last three years.

None of this slows down testing. A disclosure is a text layer, and the rest is a script discipline. The teams that get into trouble are the ones that used a synthetic performer to fake a customer, not the ones that used one to present a product clearly.