SynthID Detector: Check AI Images for Google's Watermark
SynthID is an invisible watermark woven into image pixels. This page does not claim to detect it locally: it first searches the file for familiar C2PA marker strings, clearly labels that result as an unverified hint, and then hands you to the official provenance checker for the verification step.
If the image is watermarked, you proceed straight to the SynthID remover to clean both the pixel signal and the C2PA layer.
Why a 'SynthID Scanner' Can't Be Instant
Only Google can read SynthID back, so any tool claiming an instant local verdict is guessing.
SynthID embeds a statistical pattern across an image's pixels that requires a suitable detector. A browser widget that only searches metadata cannot return that verdict. This tool is explicit about the boundary: the local C2PA step is only a marker-string hint, and the official checker provides the separate verification step.
The Honest Two-Step Check
Start with a private marker hint, then use the official checker for verification.
First, the local scan looks for common C2PA and JUMBF marker strings. A hit is not proof of AI origin and a miss is not proof that credentials are absent. The tool then opens the official content-provenance checker, where you can upload the same image for the verification step before deciding whether to clean it.
Removing SynthID Once Confirmed
Confirmed SynthID needs the pixel-level signal dealt with — a metadata wipe will not do it.
Confirmed SynthID needs the pixel-level signal dealt with, not a metadata wipe — the signal lives in the pixels and survives crops, screenshots and re-saving. The remover this detector routes to disrupts that signal where it lives and strips the C2PA manifest in the same pass, returning a clean file with no visible change. Which AI images carry SynthID: Gemini (including Nano Banana / Nano Banana Pro), Imagen, and ChatGPT stills.