Do You Have to Disclose AI-Generated Content? FTC Rules and Example Wording

The short version: disclose whenever a viewer would reasonably think an AI person is real, that a testimonial came from an actual customer, or that a realistic scene really happened. Never use AI to fake reviews or results, because no label makes that legal. Then add whatever label each platform and jurisdiction requires on top. This page covers each rule with a link to the primary source, a table of common situations with example wording, and the pre-publish checklist we use. It is general information, not legal advice.

Situations, rules and example wording

The rows below cover the setups we see most often in AI UGC ads. “Required” means a law or a platform rule calls for it. “Recommended” means nothing names it explicitly, but a label lowers the risk that the ad reads as deceptive.

What US federal law actually says

FTC Act Section 5: deception, not AI, is the violation

Section 5 of the FTC Act prohibits unfair or deceptive acts or practices. It doesn't mention AI, and there is no federal statute that requires an “AI-generated” tag on ordinary ad creative. The test is the overall impression the ad leaves on a reasonable consumer. An AI model holding your serum in a bathroom is a staged ad, the same as a hired model. The same AI model saying “this cleared my skin in two weeks” is a claim about an experience that never happened.

The Endorsement Guides (16 CFR Part 255, revised 2023)

The FTC revised its Endorsement Guides in 2023. Three provisions matter most for AI UGC:

• Endorsers must be real users. Under § 255.1(c), if an ad says the endorser uses the product, the endorser “must have been a bona fide user” of it. An AI-generated person has never used anything, so first-person experience scripts can't be true.
• “Actual consumers” must be actual consumers, or you say they aren't. Under § 255.2(c), ads that present endorsers as actual consumers should use real ones, or clearly and conspicuously disclose that the people shown are not. That is where “AI-generated actor, not a real customer” comes from.
• Virtual endorsers count. § 255.0 defines the endorser as a party that “could be or appear to be” an individual, so a computer-generated influencer is held to the same rules, including disclosing paid relationships.

The Guides also say how to disclose: clearly and conspicuously, in the ad itself, not in a comment or behind a link. Put the label on screen long enough to read, in contrasting text, and say it out loud if the ad is mostly audio.

The Consumer Reviews and Testimonials Rule (16 CFR Part 465)

The FTC's final rule on fake reviews took effect on October 21, 2024. It bans reviews and testimonials that misrepresent that they come from someone who does not exist (the FTC names AI-generated fake reviews as an example) or from someone who had no real experience with the product. It also bans buying reviews that express a set sentiment, undisclosed insider reviews and fake social-media indicators. Courts can impose civil penalties for knowing violations.

The FTC's Q&A on the rule is useful for AI UGC: it says the rule “has no blanket prohibition” on AI-generated avatars in marketing. An avatar delivering a testimonial is a problem only if the testimonial behind it is fake or false, although the ad can still be deceptive under Section 5. So the line is clear: AI faces are allowed, invented experiences are not.

One related case was reversed. In September 2024, as part of Operation AI Comply, the FTC brought a case against Rytr, an AI writing tool with a review-generator feature. In December 2025 the FTC set that order aside, saying it went too far in punishing the tool. That decision concerned the tool maker, not the advertiser: a business that posts fake reviews is still covered by Part 465.

State laws: New York requires a label on AI performers in ads

New York amended General Business Law § 396-b. Since June 9, 2026, an advertiser who knows an ad contains a “synthetic performer” must conspicuously disclose that in the ad. A synthetic performer is a digitally created asset, made with generative AI or other software, that appears to perform and is not recognizable as any real person. That describes most AI UGC actors. The law applies to ads distributed in New York in any medium except audio-only, wherever the advertiser is based. Penalties are $1,000 for a first violation and $5,000 for each one after that. Ad platforms and publishers are exempt; the advertiser is responsible.

Practically, a national Meta or TikTok campaign will reach New York unless you exclude it, so a brand running AI actors in US paid social should label them everywhere.

Other state rules to know:

• Election deepfakes. Around half of US states now regulate deceptive AI media in election communications (see Public Citizen's state tracker). This matters if you run issue or political ads.
• California's AI Transparency Act (SB 942, amended by AB 853). It became operative on August 2, 2026 and puts the duty on large generative AI providers (embedded provenance plus a free detection tool), not on advertisers.
• Right of publicity. Using a real person's face or voice without permission is a separate problem from AI labeling, and a label doesn't fix it. Get written consent.

EU AI Act Article 50 (applies from August 2, 2026)

Article 50 has two parts that matter to advertisers:

• Art. 50(2), providers: companies that run generative AI systems must mark outputs in a machine-readable, detectable way (watermarks, metadata). The 2026 AI Omnibus gave systems already on the market before August 2, 2026 until December 2, 2026 to comply (Goodwin summary).
• Art. 50(4), deployers (you, the brand): if you publish a deepfake, meaning AI image, audio or video that resembles real people, objects, places or events and would falsely appear authentic, you must disclose that it is AI-generated. For evidently artistic, satirical or fictional work, the duty is lighter: disclose in a way that doesn't spoil the work. The Omnibus did not delay this part.

A photoreal AI person recommending a product to EU viewers fits that definition, so label it. Fines for transparency violations can reach €15 million or 3% of worldwide annual turnover.

Platform rules

• TikTok requires creators to label realistic AI-generated content with its AI-generated content label, and auto-labels uploads that carry C2PA Content Credentials from other platforms.
• YouTube requires creators to disclose realistic altered or synthetic content in YouTube Studio: content that makes a real person appear to say or do something they didn't, alters footage of real events or places, or generates a realistic scene that didn't occur. Clearly unrealistic content, beauty filters, color fixes and AI help with scripts or thumbnails don't need it.
• Meta (Facebook, Instagram, Threads) requires people to use its disclosure tool when they post organic photorealistic video or realistic-sounding audio that was digitally created or altered, and applies an “AI info” label when it detects industry AI signals such as C2PA or IPTC metadata. Political and social-issue advertisers must disclose photorealistic, digitally created or altered content during ad setup.
• Google Ads requires verified election advertisers to prominently disclose synthetic content that inauthentically depicts real or realistic-looking people or events. Separately, Google Merchant Center requires AI-generated product images to keep metadata that marks them as AI-generated (such as the IPTC DigitalSourceType tag), and the main image must show the exact item being sold.

Platform labels sit next to the post, not in it. Once an ad is screenshotted, reposted or embedded in an email, the platform label is gone. That is why the checklist below also puts a label in the creative itself.

Compliance checklist (run it once per asset)

1. Classify the asset. Fully synthetic (AI person and AI scene), hybrid (real product or real footage with an AI person or background), or AI-assisted (real content with light edits like color correction or cleanup). Light edits rarely need a label; a realistic AI person almost always should have one.
2. List every destination. TikTok, Meta, YouTube, Google Shopping, your site, email, marketplaces. Note whether the ad can reach New York or the EU.
3. Set the platform label at upload. TikTok's AI-generated content setting, YouTube Studio's altered or synthetic content question, Meta's AI disclosure for photoreal video and audio, the synthetic-content disclosure in political ad setup.
4. Keep provenance data in the file. Don't strip watermarks or metadata on export. Merchant Center requires AI metadata to stay in product images, and Meta and TikTok read C2PA and IPTC signals. Images from Google’s Gemini API carry an invisible SynthID watermark (Google docs). See also C2PA.
5. Audit every claim and quote. Quotes must come from real customers who used the product. Results must be what buyers typically get. No AI before/after, no fake experts, no invented star ratings.
6. Put a label in the creative when a viewer could think the person is real. Required for New York and EU deepfakes, and the safest default everywhere else. Keep it short, on screen, in contrasting text.
7. Get consent for any real person. If a real face, body or voice is used, even only as a reference or a voice clone, keep a signed release that covers AI use. A face generated from scratch involves no model release.
8. Disclose paid relationships separately. “#ad” or the platform's paid-partnership tag. An AI label is not a material-connection disclosure.
9. Log it. Asset ID, tool and model, synthetic tier, destinations, labels applied, consent records, claim sources, date. It is your answer if a platform, partner or regulator asks.
10. Re-check the rules every quarter. The fake-reviews rule, New York's synthetic-performer law and the EU deepfake rule all took effect within the last two years.

Writing labels that don't hurt the ad

• Name what is AI: “AI-generated actor. Product shown is the real item.” beats a vague “AI-enhanced.”
• Put it where the eye already is: lower third in the first seconds of a video, a corner of a still, and in the first line of the caption.
• Don't script around it. If the label says “not a real customer,” the voiceover can't say “I've been using this for months.” Rewrite into demonstration (“here's how it works”) or third-person framing (“people who switch usually notice…”, only if it's true and typical).

For the bigger picture on what counts as AIGC, see what AIGC means. Related reading: AIGC detection and the AIGC rate, the AI content disclosure glossary entry, and building a brand with AI influencers. If you make AI UGC with ppl.studio, the creator UGC video and product UGC photos workflows put your real product in the shot; pricing is $1.99 a week.

This article is general information, not legal advice. For a specific campaign, ask an advertising lawyer.