Was this image
made by AI?
Scan the file's metadata for generator signatures, official AI labels, and Content Credentials — locally, in seconds, with honest limits explained.
How can I tell if an image is AI-generated? Check its metadata: AI tools often leave generator signatures, an IPTC "trainedAlgorithmicMedia" label, or Content Credentials (C2PA). This checker reads those signals locally — but stripped metadata proves nothing, so treat a clean result as "no evidence", not "real".
AI image checker
We scan metadata, not pixels. Pixel-based AI detection needs large models with false positives of their own — we don't guess. Need the full metadata picture (GPS, camera, dates)? Open the EXIF viewer.
What we check — and what each signal means
| Signal | Strength | What it means |
|---|---|---|
IPTC trainedAlgorithmicMedia label | Strong | The creating tool explicitly declared the image AI-generated (used by Firefly, DALL·E, Google models) |
| Generator name in Software / CreatorTool | Strong | Midjourney, Stable Diffusion, ComfyUI, NovelAI, and others often sign their output |
| Stable Diffusion "parameters" chunk | Strong | Automatic1111 and forks embed the full generation recipe (prompt, sampler, seed) in PNG metadata |
| Content Credentials (C2PA) | Verify | A signed provenance manifest — could be AI (most generators add one) or a camera (Leica, Sony). Verify fully at contentcredentials.org/verify |
| Camera EXIF (Make / Model / exposure) | Context | A camera took the file — but AI tools can copy metadata, and editors keep it |
| No metadata at all | Nothing | Social platforms strip metadata; screenshots never had it. Absence is not evidence |
Why metadata is the honest way to check
AI-detection startups sell pixel-based "detectors" with error rates nobody publishes. Metadata is different: it's a statement the file carries, written by the tool that made it. When the label is there, you don't need a probability — you have the receipt. When it's absent (most of the time, after any social upload), the honest answer is "can't tell from the file".
The rules are catching up
Under the EU AI Act's transparency rules and similar efforts elsewhere, AI content labeling is becoming expected — that's why official IPTC labels and C2PA credentials are appearing in more tools each year. Checking for them is now a practical habit for editors, teachers, and anyone receiving suspicious images.
If the image does have AI markers
Keep the file as-is if you need the evidence — re-saving or screenshotting destroys it. Need the image smaller or a different format for sharing? Use the 1 MB compressor or JPG to WebP — but note both strip metadata, which is exactly what you may not want here. To share the evidence itself, keep the original alongside.
Can this tool tell if an image is AI-generated?
It detects AI markers written into the file: generator software names, IPTC digital-source labels, and Content Credentials (C2PA). Those are strong evidence when present — but most social platforms strip metadata, so a clean result is not proof the image is real.
What does the IPTC trainedAlgorithmicMedia label mean?
It's the official IPTC digital source type for AI-generated content (trainedAlgorithmicMedia). Tools like Adobe Firefly, DALL·E, and Google's image models can embed it in XMP metadata — when it's there, it's a direct statement from the creating tool.
Why does my AI image show no markers?
Many generators don't write AI labels, and social platforms (Instagram, WhatsApp, X) strip metadata on upload. Screenshots of AI images also have no metadata at all. Absence of markers means nothing either way.
Is my image uploaded to check it?
No. The file is parsed entirely in your browser tab — nothing is sent anywhere.