AIWatermarkCheck
A clearer signal in a noisy world

Check AI watermarks.

Upload a file to inspect supported SynthID, C2PA Content Credentials, metadata, and provenance signals. Text and website checks remain available as separate experimental tools.

Your content is checked, then discarded.

Drop an image here

or choose a file from your device

JPEG, PNG, or WebP · up to 10 MB · paste from clipboard · checked and discarded

Built for better questions about what we see online.

Evidence over certainty Nothing retained Free to try
A little more context

Three layers.
One honest answer.

We look for the signals that can be verified, then show you what the softer signals suggest. When the evidence is thin, we say so.

01 / PROVENANCE

Look for a signed origin

Content Credentials and supported watermarks can carry a verifiable record of how a file was made.

02 / SIGNALS

Read the fingerprints

Metadata and forensic models add context, but they are signals—not proof of who made something.

03 / CONTEXT

Keep the caveats visible

Editing, compression, translation, or a missing credential can change what a check can see.

Curious how it works?

See exactly which checks we run, which models support them, and why an inconclusive answer can be the responsible one.

Read the methodology
Clear limits

Questions worth asking.

Does no watermark mean human-created?

No. Metadata can be removed, signals can degrade, and many generators do not expose a supported detector.

Can this detect every SynthID watermark?

No. We only show a SynthID result when a supported authoritative provider check is available.

What are Content Credentials?

They are C2PA provenance manifests that can carry signed creation and editing information.

Are uploaded files stored?

No. Files are processed transiently and are not retained by default.