Explainer
What Content Credentials actually record
Content Credentials are a way of attaching a signed, tamper-evident record to a media file describing where it came from. The technical standard behind them is C2PA, from the Coalition for Content Provenance and Authenticity, a group that includes Adobe, Microsoft, the BBC, Sony, Nikon and Leica among others. You will see the little "cr" marker on images from Adobe tools and on output from several AI generators.
What is in a manifest
A C2PA manifest is embedded in the file and cryptographically signed. It typically records:
- What produced the file, such as a camera model or a named AI model
- Whether the content was generated or modified by AI
- A list of edits, described in broad terms such as colour adjustment or cropping
- Optionally, the identity of the person or organisation that signed it
- A hash of the content, so that later changes to the pixels break the seal
The signature is the interesting part. It means a manifest cannot be quietly rewritten. If the content changes after signing, verification fails. This is genuinely useful for newsrooms and anyone who needs to show a photograph is what it claims to be.
What it is not
A Content Credential is a claim by whoever signed it, not an independent fact. It says "this signer asserts these things." A file with no credential is not suspicious, and a file with one is not automatically trustworthy. Most media on the internet carries no manifest at all, because most software does not write one.
It is also not a watermark. The manifest lives in the file's metadata. Re-encoding an image, screenshotting it, or passing it through software that does not understand C2PA usually discards the credential without anyone intending to. That fragility is a known property of the design, and it is why some vendors pair C2PA with a pixel watermark such as SynthID, which is a completely separate mechanism.
Reading one
Drop a file into the Metadata tab. If a manifest is present it is reported, and where the supporting tool is available the manifest is described rather than merely detected, so you can see what it actually claims about the file.
Removing one
The Files and Images tabs remove C2PA manifests as part of a normal clean. There are honest reasons to want this. The manifest can carry a signer identity you did not intend to publish. It can name internal tooling. It can record an edit history that is nobody else's business. And on a personal photo it is simply another block of data attached to a file you are sending to a stranger.
Be clear about the effect. Removing a credential does not forge one, and it does not make an AI-generated image look camera-original. A verifier will report the file as having no credential, which is the same state as the overwhelming majority of files. What it cannot do is produce a valid credential saying something untrue, because that would require the signer's private key.
What survives removal
If the file came from a generator that also embeds a pixel watermark, removing the C2PA manifest does nothing to that watermark. Google's SynthID in particular is embedded in the content and is not affected by any metadata operation. The full list of what can and cannot be removed is here.