Check an image: does it carry an AI label?

How do I spot AI images, AI videos and deepfakes?

No single trick always works. The safest approach is to combine several clues.

Look closely

• Hands, teeth, ears and jewellery: extra or merged fingers, mismatched earrings.
• Text in the picture: twisted letters or made-up words on signs and labels.
• Light and shadow: shadows pointing in different directions, reflections that do not match.
• Background: lines, railings or windows that lead nowhere, and repeating patterns.
• Skin and hair: too smooth, waxy, hair that blends into the background.
• Important: new AI models make these mistakes less and less. A flawless image is no proof that it is real.

Videos and deepfakes

• Mouth and sound: the lips do not quite match the speech, teeth look blurred.
• Face and edges: flickering edges at the hairline, chin or glasses, especially when the head turns.
• Eyes and expression: rare or unnatural blinking, frozen expressions.
• Voice: monotone, wrong emphasis, no breathing, choppy transitions.
• A call in a familiar voice asks for money or data? Hang up and call back on a number you know.

Check the source and search backwards

• Reverse image search: upload the picture to Google Lens, Bing Visual Search or TinEye. This finds older versions and the original context.
• Who published it first? An original usually comes from a traceable person, agency or newsroom.
• Do reputable media or fact-checkers report on it?
• For videos: take screenshots of single frames and search for those.

Provenance labels and watermarks

• Content Credentials (C2PA) are a digitally signed label inside the file. They name, for example, the software and whether AI was involved. Our tool above reads them.
• Invisible watermarks such as Google SynthID sit in the pixels themselves and often survive screenshots. Only the providers' services can check them.
• Labels are often missing: many platforms strip metadata on upload, and not every AI tool adds them.

Labels on platforms

• Instagram, Facebook, YouTube, TikTok and others show an AI notice on some posts.
• It relies on what uploaders declare or on detected labels. If it is missing, that means nothing.
• According to Google, Google Search, Chrome and the Gemini app also show for some content whether it was made with AI from Google or its partners.

If you find a deepfake

• Do not share it, not even as a warning. That only spreads it further.
• Report the post to the platform.
• Are you affected yourself? Save screenshots with link and date and contact a support service or the police.

What about AI text?

We deliberately do not offer a text detector. Such tools are often wrong and can wrongly accuse people, such as students or job applicants. There is currently no freely available, reliable check for text.