Do I need to label AI content?

Only in some cases. Meta requires a label when you post photorealistic video or realistic-sounding audio that was made or changed with AI, and says it may apply penalties if you do not. It does not require a label on images, but it may add "AI content" itself when the file carries AI metadata.

The image or video most likely carries a C2PA or IPTC record from the AI tool or editing app that made it, and Meta labels content with those signals automatically. Meta now calls the label "AI content" on posts and still uses "AI info" on ads. If the file was only edited with AI, the note sits in the post's menu instead of on the post.

Not an automatic one. TikTok says an auto label cannot be removed, YouTube says its automatic labels cannot be changed, and Meta applies its label whenever it detects AI signals. A label you added yourself is your choice, but removing it does not change whether the rules required it.

Not on Instagram, Facebook, TikTok, YouTube, LinkedIn, X or Pinterest: their rules are about images, video and audio. Under the EU AI Act, AI text published to inform the public on matters of public interest needs a notice, unless a person reviewed it and your business takes editorial responsibility. Reading your caption before it goes out keeps it a no.

Since 2 August 2026, Article 50 requires a business that publishes a deep fake (AI image, audio or video that resembles real people, objects, places or events and could pass as real) to disclose it clearly, at the latest when people first see it. Evidently artistic or satirical work needs only a light notice. Fines for breaking the transparency rules can reach 15 million euros or 3% of worldwide turnover, and for small businesses the lower of the two applies.

It is a deadline for AI tool makers, not for businesses that post. The Digital Omnibus gave generative AI tools already on the market before 2 August 2026 until 2 December 2026 to mark their output in a machine-readable way. Your own duty to disclose deep fakes has applied since 2 August 2026.

Several states require a disclaimer on AI-made election ads; California's, for example, requires "Ad generated or substantially altered using artificial intelligence" on qualifying ads by political committees. New York requires ads to disclose AI-made people (synthetic performers) since 9 June 2026. There is no general federal rule requiring labels on business posts, though the FTC bans AI-generated fake reviews.

Yes, for content that uses AI or heavy edits to show realistic-looking people or scenes. You can use the AI-generated content setting or your own clear caption, sticker or watermark. Artistic styles like anime and small edits like color correction or cropping do not need it. Some content is not allowed even with a label, such as a public figure shown endorsing a product.