Meta labels content on Facebook, Instagram and Threads with “AI info” when it detects industry signals such as C2PA or IPTC, or when users disclose AI themselves. For content only edited with AI, the notice sits in the menu.
In ads, “AI info” only appears for Meta's own AI tools. Images from other tools are yours to label.
What Facebook and Instagram require
Posts
- Meta adds the label “AI info” (called “Made with AI” until July 2024) when it detects industry signals such as C2PA or IPTC in the image, or when users disclose AI themselves.
- For photorealistic video and realistic-sounding audio that was digitally created or altered, Meta requires users to disclose it.
- If content was only edited with AI, the notice has been in the post's menu since September 2024 rather than directly on it.
Ads
In ads, “AI info” only appears when the image was created or edited with Meta's own generative AI tools. For images from other AI tools, Meta adds no automatic label, except for political ads with their own disclosure duty.
How this relates to your Article 50 duty
Deep fakes that you generate or manipulate with AI in your business, or have a contractor create to your instructions, must be disclosed (Article 50(4) AI Act, guidelines paras. 12 and 14). This applies in your own store and on every platform.
Under the Code of Practice, the “AI” should be visible without interaction. A notice in the menu only shows after a click.
The Commission lists Facebook and Instagram as very large online platforms. There, deployers may rely on the platform's labelling tools under para. 126 of the guidelines if they allow a clear and distinguishable disclosure. How far this carries is not settled, especially for notices that appear only sometimes or only after a click.
Steps
- Check every AI image with the labelling check.
- For ads with images from other AI tools, use a copy with the label burned in from Label an AI image or a clear notice in the creative. Catalogue ads pull images from your product feed, so check the feed as well.
- In posts, use Meta's disclosure where it is offered. If the notice would otherwise only sit in the menu, also put a label in the image.
- Do not remove C2PA or IPTC data. Meta uses it to detect AI.
Not legal advice. The current wording of the platform rules is what counts. A label does not make a misleading image lawful.
Frequently asked questions
Is “AI info” enough for Article 50?
Not automatically. A notice that only sits in the menu or does not appear reliably is no safe basis for relying on Meta under para. 126 of the guidelines.
May I remove metadata so that no label appears?
We advise against it. Under the Code of Practice, AI providers prohibit the intentional removal of such markings in their terms of use.
Sources
Para. refers to the paragraph number in the European Commission's guidelines on Article 50, C(2026) 5054.
- Meta: AI images in adsMeta · Platform · 3 Oct 2026
- Meta: identifying and labelling AI contentMeta · Platform · 1 May 2026
- Meta: Ansatz zur Kennzeichnung KI-generierter InhalteMeta · Platform · 5 Apr 2024
- Meta: KI-Bilder auf Facebook, Instagram und Threads kennzeichnenMeta · Platform · 6 Feb 2024
- European Commission: list of designated very large online platforms and search enginesEuropäische Kommission · Commission · 7 Sept 2026
- Commission Guidelines on the transparency obligations under Article 50 AI Act, C(2026) 5054 finalEuropäische Kommission · Commission · 20 Jul 2026
- Code of Practice on Transparency of AI-Generated Content (final version)Europäische Kommission · Code of practice · 10 Jun 2026

