Content Credentials (C2PA)
OpenAI, Adobe, Microsoft and Google attach signed provenance manifests to AI images. We read them and name the generator.
Upload a photo, video or document. AIScan Lens reads hidden provenance data and writing patterns to tell you whether AI was involved, and which tool made it.
Estimated time:
Results are probabilistic evidence, not proof. Missing metadata does not mean an image is real, and text estimates can be wrong. Don't use a single result to accuse anyone.
OpenAI, Adobe, Microsoft and Google attach signed provenance manifests to AI images. We read them and name the generator.
Stable Diffusion web UIs and ComfyUI save the prompt, seed, sampler and model inside PNG files. We surface them.
The trainedAlgorithmicMedia tag is the industry standard label for AI imagery, used by stock sites and generators.
Real EXIF data such as camera make, exposure time and aperture points toward a genuine photograph.
For text, we measure burstiness, vocabulary diversity and stock LLM phrasing to estimate AI involvement.
Nothing is uploaded. Your files are processed locally and disappear when you close the tab.
| Generator | Signal used | Reliability |
|---|---|---|
| OpenAI DALL·E / ChatGPT images | C2PA manifest | High when present |
| Adobe Firefly / Photoshop Generative Fill | C2PA + IPTC tag | High when present |
| Microsoft Designer / Bing Image Creator | C2PA manifest | High when present |
| Stable Diffusion (A1111, Forge) | PNG "parameters" chunk | High when present |
| ComfyUI / FLUX workflows | PNG workflow chunk | High when present |
| Nano Banana / Gemini (Google) | Metadata here + SynthID at synthid.com | High with both checks |
| OpenAI Sora (video) | C2PA manifest in MP4 | High when present |
| Midjourney, Leonardo, Ideogram | IPTC / XMP / software tags | Medium |
| ChatGPT, Claude, Gemini (text) | Stylometric estimate | Indicative only |
Yes. AIScan Lens is completely free with no sign-up and no daily limit. The site is supported by advertising.
No. The analysis runs entirely inside your browser using JavaScript. Your image, video or text is never sent to us or stored anywhere.
It reads provenance metadata embedded in the file: C2PA Content Credentials, the IPTC "trainedAlgorithmicMedia" tag, Stable Diffusion and ComfyUI generation parameters, and EXIF/XMP software fields. When a generator such as DALL·E, Adobe Firefly, Midjourney or Stable Diffusion left a signature, the tool names it.
Any tool that writes identifiable metadata, including OpenAI DALL·E and ChatGPT images, Adobe Firefly, Microsoft Designer / Bing Image Creator, Stable Diffusion (Automatic1111, Forge, ComfyUI), FLUX, NovelAI, Midjourney, Google Imagen / Gemini, Leonardo.Ai, Ideogram, Runway, Sora and others.
Metadata is easily removed. Screenshots, social media uploads, messaging apps and many editors strip it. "No signature found" means the file carries no proof either way. It does not mean the image is real.
Paste an essay, article or email into our AI text detector.
Open AI Text Detector