Google-focused · honest boundaries
Gemini & SynthID
SynthID is invisible and baked into the pixels; Google runs the official detection doors. What we can legitimately verify in your browser is the other layer Google attaches — C2PA content credentials.
Runs entirely in your browser — your file is never uploaded.
Drop an image here, or browse files
JPEG · PNG · WebP — up to 25 MB. Files are read locally and never uploaded.
Quick answer
SynthID pixel detection = Google's portals (Gemini app; SynthID Detector, rolling out). C2PA metadata = anyone can read it — including the tool above, entirely in your browser. If an image shows a Google-family C2PA generator claim, that's strong, signed provenance evidence. If it shows nothing, that's not evidence it's natural: screenshots destroy C2PA while SynthID may still survive — and only Google's gates can check it.Where SynthID verification officially lives
| Surface | Checks | Access |
|---|---|---|
| Gemini app | SynthID on images/video/audio in context | Public, with a Google account |
| SynthID Detector portal | Cross-format SynthID checks | Rolled out via waitlist (prioritizing journalists & researchers) |
| About this image (Search) | Provenance-ish context for indexed images | Public |
| C2PA readers (like ours) | Metadata layer co-existing with SynthID | Public, any file, no account |
One nuance this page exists for: “no C2PA manifest” and “no SynthID found” are different facts. Because they are independent layers, a file can (a) have both, (b) have C2PA only, (c) have SynthID only — e.g. after the metadata was stripped — or (d) have neither. Our tool reports layer (metadata) honestly and links out where layer (pixels) must be checked.
FAQ
How do I check if Gemini generated an image?+
Google's official SynthID verification exists in the Gemini app, and the standalone SynthID Detector portal is being rolled out (with priority for journalists and researchers). For third parties, the practical check we offer is the C2PA metadata layer Google attaches to generated images — readable above without any Google account.
Can any website detect SynthID from pixels?+
SynthID detection is gated: Google provides the detection interfaces and, for most SynthID layers, keeps the verification capability close to its models. Be skeptical of random 'SynthID detector' sites that don't explain their method. This site reads the independent open-standard layer (C2PA), which is honest about its limits.
Does SynthID survive screenshots?+
Google says SynthID is designed to survive common edits like cropping, compression, and color changes — that's the point of a pixel watermark — but no vendor claims invariance to every transformation, and extreme edits can degrade or remove it. The C2PA layer, by contrast, is destroyed by screenshots.
What about Gemini text?+
Gemini text output uses SynthID-Text; the open-source variant of that method lets outside developers watermark their own models, but verifying Gemini's own served traffic requires Google's key — same story as Claude's text watermark.
Sources
- Google DeepMind — SynthID — accessed 2026-09-29
- Google — Making it easier to understand how content was created — accessed 2026-09-29
- C2PA — Technical specification — accessed 2026-09-29