AI UGC disclosure has moved from a grey area to an enforced advertising requirement. If you run synthetic creators or AI-generated video in your paid social, the platforms and the regulators now expect you to say so, and the rules differ enough between them to catch teams out.
The requirements are more consistent than they look. This is a practical explainer for advertisers running AI UGC on Meta and TikTok in 2026: what Meta's rules cover, how TikTok differs, what the FTC and UK ASA expect on top, and the evidence that disclosure does not cost you performance. One caveat: this is operator guidance, not legal advice, so check current platform policy before shipping anything sensitive.
Key takeaways
- AI UGC disclosure is required when AI generates or substantially alters a realistic person, voice or event. Behind-the-scenes AI help, like scripting or captions, is generally exempt.
- Meta and TikTok both read C2PA provenance metadata and can label content automatically, but many ad tools do not embed it, so you still declare AI use yourself.
- The FTC expects layered disclosure for paid content: one label for the advertisement, a separate one for the AI. An AI tag alone is not enough.
- The UK ASA runs a media-neutral code. There are no AI-specific rules, but existing rules on misleading content apply regardless of how an ad was made.
- Disclosure has not been shown to reduce ad performance in testing, so there is no performance reason to avoid it.
What AI UGC disclosure actually is
AI UGC disclosure is the act of telling the viewer that synthetic media was used to create advertising that could otherwise be mistaken for a real person or a genuine testimonial. The trigger across every regime is the same question: would the audience be misled if they did not know AI was involved?
That tells you what counts. A synthetic presenter reading a script is exactly what these rules exist for. AI used to write that script, generate captions or clean up audio is not, because the viewer is not deceived about anything they can see. Realism is the line, not tooling. Our guide to AI UGC versus real UGC covers where each fits in a mobile app account.
Meta's AI labelling requirements
Meta treats AI labelling as an enforced advertising standard on Facebook and Instagram, not an optional courtesy. Ad creative that uses AI to generate or materially alter visual or audio content is expected to carry an AI-generated label. When an ad features a photorealistic synthetic human, Meta's policy pushes that label toward a visible overlay rather than a detail hidden in a menu.
Detection runs two ways. Meta reads C2PA signals, the provenance standard from the Coalition for Content Provenance and Authenticity that many AI tools embed, and labels content automatically when it finds them. For creative built with tools that do not embed C2PA data, the responsibility shifts to you: declare the AI use during campaign setup rather than hoping automated detection covers it.
TikTok's AI content policy
TikTok was the first major platform to integrate C2PA Content Credentials, in January 2025, and it now auto-labels detected AI content from a long list of generation tools. Its AIGC label covers realistic AI depictions of people, places and events, and it explicitly extends to AI-generated voiceovers even when the footage itself is real. Once TikTok applies that label, the creator cannot remove it.
The exemptions mirror Meta's: AI-written captions, descriptions, hashtags, text overlays and script assistance do not require a label. Usefully, TikTok has said switching on the AI-generated content setting does not reduce distribution as long as the video follows its community guidelines, which removes the main excuse teams use for skipping disclosure.
The FTC and the layered disclosure rule
In the US the FTC applies its existing rules on deception, endorsements and fake testimonials to AI-generated content with full force, and updated its AI endorsement guidance in May 2026. The practical consequence for sponsored AI UGC is layered disclosure: one disclosure that the content is an advertisement or paid partnership, and a separate one that AI was used to create it.
The two do not substitute for each other. A "#ad" tag does not satisfy the AI requirement, and an "AI-generated" tag does not satisfy the sponsorship one. Both must be clear and up front. The stakes are per-piece: the FTC's maximum civil penalty rose to 53,088 dollars per violation in 2026, and California, New York and Texas have layered their own synthetic-endorsement rules on top.
The UK position: a media-neutral code
The UK is quieter. The CAP Code contains no AI-specific rules, and the ASA has confirmed it is media-neutral: an AI-generated ad is held to the same standards as any other, a position that June 2026 guidance reinforced rather than extended.
The test CAP gives advertisers is the one to carry into any market. Ask whether the audience would be misled if the use of AI were not disclosed, and whether disclosure would clarify or contradict the ad's overall message. If leaving AI unmentioned would mislead, you disclose. That single question resolves most cases without a rulebook for each platform.
Does disclosure hurt performance?
This is the real worry behind most reluctance to label, and the evidence is reassuring. A MediaScience and Adelaide University study of 900 US participants tested several labelling approaches against an unlabelled control and found no significant drop in brand recall, ad liking or brand choice. It measured brand and attention metrics rather than click-through specifically, so treat it as strong directional evidence rather than a guarantee, but the direction is clear.
In our experience audiences have largely made their peace with synthetic media, and a clear label reads as honesty rather than a red flag. The real performance risk in AI UGC is fatigue and sameness, a creative problem, not a compliance one.
A practical disclosure checklist
- If a synthetic person, voice or event appears in the ad, label it. Assume yes unless you can clearly argue no.
- Turn on the platform's own AI-content toggle at upload on both Meta and TikTok, rather than relying on automated detection.
- For sponsored content in the US, disclose both the ad relationship and the AI use, separately and up front.
- Keep a record of which tools generated which assets, so a provenance query later is a lookup, not a scramble.
- When in doubt, apply the CAP test: would the audience feel misled if they knew? If yes, disclose.
None of this changes which tools are worth using. If you are still building your stack, our roundup of the best AI UGC tools in 2026 covers what each is genuinely good at, tested on real ad accounts.
Frequently asked questions
Do I need to disclose AI UGC in my ads?
If AI generates or substantially alters a realistic depiction of a person, a voice or an event, then yes. Meta and TikTok both require a visible AI label, and the FTC treats an undisclosed synthetic endorsement as a deceptive testimonial. An AI-assisted script over real footage is generally exempt, but a synthetic presenter never is.
What AI UGC is exempt from disclosure?
Behind-the-scenes uses that do not change what the viewer believes they are seeing. AI-written scripts, suggested hooks, captions, hashtags, text overlays and light audio cleanup on real footage do not usually trigger a label. The line is realism: if a viewer could mistake synthetic content for a real person or event, it needs a label.
Do Meta and TikTok add the AI label automatically?
Sometimes. Both read C2PA Content Credentials, the provenance metadata embedded by many AI tools, and label content automatically when they detect it. You cannot rely on this alone: many ad-native tools do not embed C2PA data, so you still declare AI use yourself at campaign setup.
Does an 'AI-generated' label satisfy the FTC?
Not on its own for paid content. The FTC expects layered disclosure: one label that the content is an advertisement, and a separate one that AI was used. An 'AI-generated' tag does not satisfy the sponsorship rule, and a '#ad' tag does not satisfy the AI rule. For sponsored AI UGC in the US you need both, clearly and up front.
Does disclosing AI reduce ad performance?
The available evidence says no. A MediaScience and Adelaide University study of 900 US participants found AI disclosure labels caused no significant drop in brand recall, ad liking or brand choice against an unlabelled control. Disclosure is a compliance requirement first, but the fear that labelling quietly taxes performance is not supported by the data.
Want this run for you?
Disclosure is the easy part. Building an AI UGC system that stays compliant and keeps producing genuinely different creative, instead of ten labelled versions of the same idea, is the work that moves cost per install.
If you run a mobile app spending real money on Meta and want a creative engine built for how the platforms work now, apply to work with us. We take a small number of mobile app clients per quarter.