SynthID Checker

SynthID and C2PA solve different problems

Google DeepMind describes SynthID as an imperceptible watermark embedded directly into generated or altered content — images, audio, text, and video — with detection handled by SynthID technology rather than by an open standard.

C2PA takes the opposite approach. It stores signed provenance assertions alongside the asset, using open cryptography that any compatible validator can check. That openness is why this site can validate C2PA locally and honestly.

The trade-off is precise. C2PA is independently verifiable but trivially stripped by re-encoding. SynthID survives editing far better but can only be checked by the party that holds the detector. An asset can carry one, both, or neither.

• C2PA — open, cryptographic, anyone can verify, easily removed.
• SynthID — robust to compression and cropping, invisible, only Google can verify.
• Metadata strings — readable and writable by anyone, removed by most platforms.

Why no third-party page can check SynthID

The signal is embedded in the content itself — in the pixel values of an image, the waveform of audio, or the token choices of text. It is not a field that a parser can read, and detecting it requires the specific model and parameters used to embed it.

Those are not published, and that is a deliberate design choice rather than an omission. A watermark whose detector is public is a watermark that becomes straightforward to strip, because an attacker can iterate against the detector until it stops firing.

The practical consequence for you: if a website offers a SynthID scanner, it is either running a generic AI-image classifier and mislabelling it, or checking metadata and calling that SynthID. Neither is what it claims.

What you can check here instead

The local file checker on this site validates C2PA Content Credentials and reads ordinary EXIF, IPTC, and XMP metadata. Both of these are genuinely inspectable, and both are worth checking first because they are free and instant.

Some content that carries SynthID also carries C2PA credentials, because a provider may apply both. In that case the credential gives you a verifiable statement about origin even though the perceptual watermark itself stays unreadable to you.

Removing metadata does not remove a perceptual watermark, and reading metadata does not reveal one. The two live in different layers of the file and behave independently.

How to check responsibly when it matters

Where Google offers a verification flow for the media type you are checking, use it — that is the only path to an authoritative SynthID answer, and it is the provider's own detector rather than a guess.

Run the local C2PA and metadata check as a first pass. It costs nothing, and a verified credential often settles the question before you need a watermark check at all.

Treat any single negative result cautiously, especially after heavy editing, format conversion, or screenshotting. And remember that SynthID covers Google's own systems — content from other providers will not carry it no matter how synthetic it is.

• Use Google's official detection flow for an authoritative SynthID claim.
• Use the local checker here for C2PA credentials and editable metadata.
• Do not treat a negative from any single method as clearance.
• Remember SynthID only marks Google-produced content.