How To Detect AI Generated Photos In 2026

The old tricks, counting fingers and looking for melted text, work less often now. The best image generators produce photos that pass a casual look. Detection has shifted from spotting flaws to checking where an image came from.

What's changed

• Fewer visual errors: hands, text and faces are much better than in 2023.
• More provenance data: OpenAI, Adobe, Google and Microsoft label images from their tools with Content Credentials or IPTC tags, and some cameras and phones now sign real photos.
• More platform labels: Meta, TikTok, LinkedIn, YouTube and Pinterest show labels when they detect AI signals or creators disclose.
• New rules: the EU's AI Act requires deepfakes to be labeled from August 2026.

What still works

1. Check provenance. Read the file's metadata and Content Credentials. In Chrome, AI Image Detector (made by Patrick Bushe, who runs this site) reads them when you right-click an image; Content Credentials Verify checks the full signed history.
2. Check the source. Who posted it first? Reverse image search with Google Lens or TinEye. A real event usually has many photos from different people.
3. Check platform labels, such as Meta's "AI info" or LinkedIn's "CR" icon.
4. Use detectors as one signal. Services such as Hive give probability scores; they're useful on unlabeled images but not proof.
5. Still look closely. Small errors in reflections, shadows, background text and jewelry still happen.

The biggest gap

Most images people share have had their metadata stripped by screenshots or social media. Provenance helps only when it survives, and a missing label doesn't mean a photo is real.

Change the question

Instead of "is this AI?", ask "can I confirm this is real?" If you can trace an image to a trustworthy source, you don't need to prove it isn't AI. If you can't, treat it as unverified.

See how to tell if an image is AI and fake AI photos on social media.