Invisible signal, not a corner logo
A visible watermark is a graphic or text overlay. SynthID is designed to be imperceptible and embedded in the generated content itself. Google describes it as resilient to common changes such as cropping, filters, and lossy compression.
SynthID and Google AI
Google uses SynthID across supported image, audio, video, and text experiences. Public media verification is available inside Gemini for signed-in users, but Gemini currently scopes its SynthID result to content created or edited by Google AI tools.
SynthID versus metadata
EXIF, IPTC, and XMP live in file containers and can disappear during an export. C2PA adds signed provenance credentials. SynthID is a different embedded signal, so stripping metadata alone is not equivalent to SynthID removal.
Can SynthID always be removed?
No responsible tool can promise that. Detection depends on the originating model, file, edits, compression, and the detector’s supported implementation. SynthID Remover therefore uses best-effort language and gates public claims on a documented evaluation set.
Read Google’s primary explanation at Google DeepMind SynthID, or use our provider-specific checker guide.