Image Tools
AI Image Detector
Check an image for AI-generation signals — Content Credentials and metadata read locally in your browser, plus a pixel-level AI scan that ignores metadata entirely.
What the report shows
- Content Credentials (C2PA)Found — signed by issuer
- AI metadataNot found
- Pixel-level AI scanAI probability with a generator-family hintLikely AI
Example of the report layout — each signal gets its own row and badge.
What it detects
Three independent signals, read separately so one stripped channel can't hide another.
Works on screenshots and re-uploads
Social platforms strip metadata. The pixel-level scan reads the pixels themselves, so a screenshot of an AI image still produces a signal.
Names the likely generator family
When the pixel scan finds AI signals, the report can name a generator family — Google Imagen and Nano Banana, GPT image and DALL·E, Midjourney, Flux, Stable Diffusion, Adobe Firefly, Ideogram, Recraft, Seedream, Qwen, Grok Imagine, Hunyuan, Kling or Reve.
Reads signed credentials first
If the file carries a Content Credentials manifest, the report shows the signer, the generator app and what the claim declares — the strongest signal when present.
Free checks stay in your browser
Credential and metadata parsing runs locally — the file is never uploaded. Only the optional pixel scan needs a signed-in account and credits.
How to check an image
Start with the original file
Download the image instead of screenshotting it when you can — originals keep Content Credentials and metadata that screenshots lose.
Drop, paste or pick the image
Drag a JPG, PNG or WebP up to 10 MB into the dropzone, paste it from your clipboard, or use the file picker.
Read the three signal rows
The report checks Content Credentials and AI metadata in your browser and shows each result as its own badge — found, not found, or not readable.
Run the pixel scan when nothing is embedded
A stripped file still shows whether pixels look generated. Sign in, then run the scan — its credit cost is shown before it starts.
Confirm Google-made images with Google
If signals point to a Google model, verify the SynthID watermark itself in Google's official detector — no third party can read it.
When to check an image
A viral photo before you share it
Screenshots circulating on social have no metadata left — the pixel scan is the check that still works.
A marketplace or dating profile picture
Profile photos get re-compressed and re-uploaded many times. The scan looks for generation artifacts that survive that.
A newsroom source image
Before publishing, check whether a tip or leak photo carries a signed credential or shows pixel-level AI signals.
Moderation of AI-labeled content
When platform rules require AI images to be disclosed, the report helps reviewers triage suspicious uploads.
Limits and honest edge cases
Nothing found is not proof of real
An image with no credential, no metadata and an uncertain scan can still be AI-generated — a clean result is absence of evidence, not evidence of a camera.
Degraded files lower confidence
Heavy re-compression, strong filters or screenshots with lots of UI chrome reduce what the pixel scan can say.
AI edits to real photos
A real photo retouched or inpainted with AI tools can score as AI — the signal describes generation, not truth.
Face swaps are a different problem
Deepfake faces composited into real photos are not what this detector specialises in — use a dedicated deepfake check for that.
AI Image Detector FAQ
Is the AI image detector free?
Content Credentials and metadata checks are free and need no account — they run in your browser. The pixel-level AI scan requires sign-in and credits; its cost is shown before the scan runs.
Does it read the SynthID watermark?
No. The SynthID watermark is proprietary — only Google's official detector can read it. This tool checks public signals (credentials, metadata, pixels) and links you to Google for watermark confirmation.
What does an uncertain result mean?
It means the pixel signals are inconclusive — common for heavily edited, cropped or degraded images, and for real photos processed with AI tools. Treat it as 'no reliable signal' rather than 'probably real'.
Is my image uploaded or stored?
Credential and metadata checks run entirely in your browser — nothing is uploaded. The pixel scan sends the image to our server for analysis; see the privacy policy for retention details.
Which generators can it identify?
The pixel scan can hint at a generator family — Google Imagen and Nano Banana, OpenAI GPT image and DALL·E, Midjourney, Flux, Stable Diffusion, Adobe Firefly, Ideogram, Recraft, Seedream, Qwen, Grok Imagine, Hunyuan, Kling and Reve. It is a hint, not a certificate — confirm Google output in Google's detector.
Why does it differ from Google's SynthID Detector?
Different signals: Google reads its invisible watermark; this tool reads credentials, metadata and pixels. A file can have a SynthID watermark and still show 'nothing found' here if the scan can't see it — only Google can confirm the watermark.