In short
- C2PA is an open technical standard for attaching a tamper-evident record of an image's origin and edits to the file. Content Credentials is the name of the user-facing version.
- Credentials show who signed which claims about a file. They don't show that the picture depicts what really happened.
- Credentials can be stripped, so their absence says little. Their presence is only as useful as your trust in the signer.
- They work best alongside other checks, such as detection and research into the image's source.
What C2PA is
C2PA stands for the Coalition for Content Provenance and Authenticity. It publishes an open technical standard that helps publishers, creators and consumers establish where digital content came from and how it was edited. Companies on its steering committee include Adobe, Amazon, the BBC, Google, Meta, Microsoft, OpenAI, Sony, TikTok and Truepic.
The user-facing version is called Content Credentials. The C2PA site compares it to a nutrition label for digital content: a way to see an item's history.
How it works
- A camera, editing tool or AI generator records claims about the file: who made it, with which tool, and which edits were made, including whether generative AI was involved.
- Those claims are bundled into a manifest and tied to the file's content with a cryptographic hash, so a later change to the pixels breaks the link.
- The manifest is digitally signed with a certificate belonging to the tool or organization making the claims.
- Anyone can check the signature and the hash with a verifier. If the file is edited again with a compatible tool, a new signed record is added, building a chain of history.
What credentials can and can't tell you
They can show who signed the claims, whether the file has changed since it was signed, and the stated history of the file, including generative-AI use when the creating tool declares it.
They can't show that a scene is true. A signed photograph can still be of a staged event. They also say nothing about a file that has no credentials.
Why a missing credential doesn't mean fake
Credentials live in the file's metadata. Many platforms and tools remove or rewrite metadata when you upload, resize or screenshot an image, and plenty of cameras and generators don't add credentials at all. So most real images today carry none. Some approaches pair credentials with invisible watermarks or fingerprints so a stripped file can be matched back to its record, but support varies.
Who signed it?
A valid signature proves that a particular certificate holder made the claims. It doesn't prove that holder is honest. Deciding how far to trust a credential is a judgment about the signer, in the same way you would judge a publisher's byline.
How to check an image
- Look for the credentials icon on platforms that display it, or open the file in a verifier such as the Content Credentials verification site.
- Read who signed it and which actions are listed.
- If there are no credentials, fall back on research into where the image first appeared, and on detection.
Credentials and detection work together
Credentials answer the question “what does the maker say about this file?” Detection answers “what does the image itself look like?” Neither replaces the other. Synas Detect 001 includes a metadata-mismatch signal for the cases where a file's metadata and its pixels disagree, and our guide to how AI image detection works covers the other methods.