| Where | What it asks | Worth knowing |
|---|---|---|
| TikTok | Ads with realistic AI-generated images, video or voices need a disclosure label (rule since 21 July 2026). | TikTok also reads C2PA content credentials to detect AI content. AI-generated political content is banned in ads. Use the AI-generated content toggle in Ads Manager or add a visible label. |
| Meta (Facebook, Instagram) | Disclose photorealistic video or realistic-sounding audio made or altered with AI, using Meta's AI disclosure tool. | Meta may also label content it detects. Many AI tools do not embed provenance data, so you often have to declare AI use yourself. |
| YouTube | Creators disclose realistic synthetic or altered content. Labels now sit below the player on long videos. | Mass-produced, templated videos can lose monetisation under the platform's inauthentic content policy. |
| EU AI Act, Article 50 | Providers of generative AI must mark outputs in a machine-readable way. Businesses that publish deepfakes must label them. | The marking duty has until 2 December 2026 for systems already on the market. The Act can reach companies outside the EU whose content reaches EU users. |
| USA (FTC) | Existing rules on deception, endorsements and fake testimonials apply to AI content in full. | The FTC updated its AI endorsement guidance in May 2026. |
Some advertisers report that disclosure does not reduce performance, and YouTube has said its disclosure does not affect ranking. Treat that as encouraging rather than guaranteed. Being caught without a label costs more: removed ads, reduced reach, or lost monetisation.
Yes, if the ad contains realistic AI-generated images, video or voices. TikTok has required a disclosure label on such ads since 21 July 2026, and it also detects AI content through content credentials.
It can. The Act reaches companies outside the EU whose AI output or content reaches people in the EU, and duties differ for the tool provider and for the business publishing the content. Ask a lawyer about your own case.
Platform rules focus on realistic AI images, video and audio. AI-assisted writing such as captions, scripts and hashtags is generally outside those labelling rules, but check each platform's current policy.
Last updated 3 November 2026. We update this page as rules change.