FAQ & troubleshooting

Can proofmark detect if something was AI-generated?

No — detection after the fact is unreliable. proofmark proves the opposite direction: who signed a claim (created / edited / aiGenerated / reviewed) bound to the exact bytes. It's attestation, not detection.

Is this real C2PA?

It shares C2PA's model of signed assertions but uses portable W3C VCs in a detached sidecar instead of a manifest embedded in the media. See proofmark vs full C2PA.

verifyManifest says contentMatches: false

The bytes you passed don't hash to the manifest's contentHash — the content changed (even one byte), or you verified against a different file. That's the tamper-evidence working.

Where do I store the manifest?

Anywhere: next to the asset (e.g. photo.jpg.proofmark.json), in a database, or in a header. It's just { contentHash, assertions }.

Does a valid signature mean the claim is true?

No. It proves who signed it. Whether to believe the claim depends on whether you trust that signer's DID — keep an allow-list of known tool/organization DIDs.

Does it work on images and binaries?

Yes — stamp/verifyManifest accept Uint8Array, so any bytes (images, PDFs, files) work, not just text.