Why that distinction matters
Creators search “remove AI watermark” and land on EXIF cleaners. Those cleaners never saw SynthID. Canvas re-encode, JPEG quality, and screenshots may degrade a watermark, or they may not. This site will not pretend to wash it out.
Official detection, when a product offers it, is a separate pipeline. Optional OpenAI provenance checks on this site can include SynthID results from that vendor’s API after an explicit click. Local inspection does not invent a SynthID score from pixels in WebGL.
What a negative result means
Not found: the copy you sent did not yield a detection. It does not mean “human.” It does not mean “safe to mislabel.” Combine with C2PA if present, and stop there if both are empty.
More context: AI watermarks, C2PA.
Common questions
What is SynthID?
It is Google DeepMind’s imperceptible watermark in the media itself. It is not a C2PA manifest and not an EXIF tag. This site’s local inspection does not invent a SynthID score. Optional OpenAI verification can report a SynthID result from OpenAI’s API only after you click it.
Does every Gemini or Google image include SynthID?
Product behavior changes. Read current Google documentation for the generator you used. Do not copy a 2024 screenshot as a rule.
If I delete EXIF, is SynthID gone?
No. A pixel watermark such as SynthID remains after a metadata strip. A screenshot or heavy recompress might damage it, or it might not. This site does not claim to wash it out.