How to check if an image is AI generated
Most major AI image generators now label their output. OpenAI, Adobe, Microsoft and Google embed C2PA Content Credentials, a signed record stating that an image was produced by AI. Stable Diffusion interfaces write the full prompt, seed and model into PNG text chunks. Stock agencies and editors add the IPTC tag trainedAlgorithmicMedia. AIScan Lens reads all of these directly from the file in your browser.
What the results mean
- AI-generated content detected: the file contains a recognised AI signature. This is strong evidence.
- Likely authentic capture: camera EXIF data (make, model, exposure) was found and no AI signature.
- No AI signature found: the file has no provenance either way, which is common after screenshots or social media uploads.
How to spot a fake AI image by eye
When metadata is missing, look for visual tells: inconsistent hands and teeth, warped text and signage, mismatched earrings, unnatural reflections, overly smooth skin and backgrounds that melt into patterns. Run a reverse image search to find the earliest copy and its source.
AI image detector vs Google SynthID
Google's SynthID Detector (free since October 2026) finds Google's invisible watermark in Nano Banana, Gemini and partner images, but returns "not detected" for Midjourney, FLUX and local models. Our metadata scan covers many of those. Use both for the most complete check. Read the comparison.
Read more: How to tell if an image is AI-generated · What are Content Credentials?