AI
AI image checker
Check what the file declares.
See whether an image carries the markers AI tools leave behind, and when the file records it, which app and model made it and the prompt. Nothing is uploaded.
What's checked
- IPTC digital source type, the standard “made with AI” label
- C2PA Content Credentials, used by Adobe, OpenAI, Microsoft and camera makers: which app and model made it
- Prompt, model and settings saved by Automatic1111, Forge, ComfyUI, NovelAI, InvokeAI, Fooocus, SwarmUI and Midjourney
- Generator names in the metadata, such as Midjourney, DALL·E or Firefly
Not checked: invisible watermarks such as Google's SynthID, which live in the pixels and need their maker's own detector. Nothing is uploaded; the file is only read.
Need a stronger check?
- Google's SynthID: upload the image to the Gemini app and ask if it's AI-generated. The watermark survives screenshots and cropping, but it only covers Google's own AI, and the image is sent to Google.
- Content Credentials signature: contentcredentials.org/verify, the official verifier, confirms who signed the record.
How did this tool do?
Good to know
How AI images label themselves, and when they don't
Many AI image tools now write a label into the files they produce. The most widely used is the IPTC digital source type: a code that says, in effect, “created by a trained algorithm”. Google, Meta, Adobe and others add it, and social networks increasingly read it to show an “AI info” tag.
A second, stronger kind is Content Credentials, built on the C2PA standard. It's a signed record of how an image was made: which tool, which edits. Adobe Firefly, OpenAI's image generators and Microsoft's tools attach them, and so do some cameras to prove a photo is a capture. Finding Content Credentials with no AI claim is itself useful information.
Images made with Stable Diffusion on your own computer often carry something more revealing still: the full generation settings, including the prompt, the model and the seed. Automatic1111, ComfyUI and InvokeAI all save them by default.
The limit is that all of this lives in metadata, which is fragile. A screenshot, a messaging app, a social network upload or a single re-save can strip it, so most AI images you meet online no longer carry a label. That's why this check can confirm an image is AI but can never confirm it isn't. Invisible watermarks such as Google's SynthID survive that stripping, but only their makers can read them.
Where you'll meet it
Before you share or cite an image
See whether the file itself says it was generated.
Checking a stock or client image
Content Credentials show how it was made and edited.
Recovering a prompt
Stable Diffusion PNGs often still hold the prompt and settings.
Guide
How to use it
- 01Drop or paste an image.
- 02Read the verdict: what the file says about how it was made.
- 03Open What was found for the exact markers, and any prompt saved in the file.
Good to know
AI image markers, answered
Can this tell if any image is AI?
No, and nothing reliable can. It reads the markers AI tools write into files. When one is present, that's strong evidence. When none is found, the image may still be AI: screenshots, messaging apps and social networks strip metadata, and a single re-save can remove it.
Can it find the prompt and model?
When the file still has them, yes. Automatic1111, Forge, ComfyUI, NovelAI, InvokeAI, Fooocus, SwarmUI and Midjourney save the prompt, the model and settings like seed and steps. Content Credentials name the app and model but never the prompt. If the metadata has been stripped, the original prompt can't be recovered from the pixels.
What are Content Credentials?
A signed record (the C2PA standard) that tools such as Adobe Firefly, OpenAI's image generators and some cameras attach to a file, describing how it was created or edited. This tool detects them and reads what they say; it doesn't verify the signature.
What about Google's SynthID?
SynthID is an invisible watermark in the pixels, not in the metadata, so it isn't checked here. Google's Gemini app can check it: upload the image and ask whether it's AI-generated. It only recognises images made or edited with Google's own AI, and the image is sent to Google.
Can I verify a Content Credentials signature?
Not here. This tool reads what the credentials say but doesn't check the signature. The Content Authenticity Initiative's verifier at contentcredentials.org/verify does.
Is my image uploaded?
No. The file is read in your browser and nothing is sent anywhere.