AI labels on social media: Instagram, TikTok, LinkedIn, YouTube and the EU icon

In February 2024, Meta said it would label AI images on Facebook, Instagram and Threads by reading the signals AI tools write into their files: C2PA Content Credentials and IPTC metadata. From April 2024 the label read “Made with AI”. On 1 July 2024 Meta renamed it “AI info”, after lightly retouched photos were flagged as made with AI.

Since September 2024, content Meta detects as AI-generated keeps “AI info” on the post, saying whether it came from industry signals or from the person who posted. Content only edited with AI tools gets the label in the post's menu instead. Meta still describes this system in a post updated on 1 June 2026. On Instagram, the label sits under your username, above the picture, not on it (TechCrunch, 2024).

Adding it yourself: Meta lets you disclose AI when you post: look for the AI label option as you share. Disclosure is required for photorealistic video and realistic-sounding audio, and Meta says it may apply penalties when people don't.

On 9 May 2024, TikTok said it had become the first video-sharing platform to use C2PA Content Credentials: it labels AI content uploaded from elsewhere when the file carries them. In November 2025 it said it was testing an invisible watermark that only TikTok can read, and that its labelling had covered over 1.3 billion videos.

TikTok asks you to label realistic AI-generated content yourself when you post. Unlabelled realistic AI content may be removed, restricted or labelled by TikTok, and creators can't remove the labels it applies automatically. TikTok also warns that labels may be lost when content is reuploaded or edited elsewhere. Third-party guides place the label bottom left, below your username.

LinkedIn shows a small Cr icon, top left, on images and videos that carry signed C2PA Content Credentials. Click it to see whether AI generated or edited the content, with the app, the signer and the date. LinkedIn says it can't yet label all AI content. Our tool leaves out the original's credentials (their signature no longer matches a changed picture), so a labelled copy shows no Cr icon: the EU icon tells people instead.

YouTube asks creators to disclose realistic content that is meaningfully altered or synthetic, in YouTube Studio when they upload: a real person saying something they didn't, or a realistic scene that never happened. Unrealistic content and beauty filters are exempt. YouTube may also label content itself: when it carries C2PA, comes from YouTube's AI tools or is detected by its systems. Its separate “Captured with a camera” label, launched in October 2024, says the opposite: the video carries C2PA 2.1+ credentials from the camera.

Pinterest has shown an “AI modified” label since 30 April 2025. It relies on IPTC metadata, its own classifiers and your disclosure.

X: we couldn't confirm from X's own help pages how it labels AI images. Press reports and a blog test from 2026 describe a “Made with AI” label you apply yourself. X removes a photo's EXIF metadata on upload, according to Lumethic's comparison of platforms (updated 13 August 2026). On X, count only on the picture and your post's text.

Platforms read metadata at upload, then often drop it: Meta serves files with neither C2PA nor IPTC data, according to Lumethic. The Commission's EU icons page says the icon must stay visible when content is reshared or downloaded; our Detect AI-generated images guide explains why metadata is so easily lost.

For deepfakes published by professionals, Article 50 of the AI Act asks for disclosure that is clear and distinguishable at the latest at first exposure. The Code of Practice wants a label embedded in the content, seen without a click and clear of other overlays. The Commission's Guidelines add that the AI tool's machine-readable metadata doesn't meet the publisher's duty (para 117), and that a disclosure hidden in menus can be easily overlooked (para 142).

Every app lays names, buttons and badges over the picture. The top right corner, the Code of Practice's example, is clear of most of them. Figures are approximate, from third-party guides that don't always agree (checked in October 2026), and apps change their layouts.

The EU AI image label tool puts the icon top right by default, and its Clear of app buttons on tall images option keeps corner labels on 9:16 pictures below the top 14%, above the bottom 22% and 6% from the sides (approximate figures too).