How C2PA Content Credentials Work

What Is C2PA? covers the what and why. This one covers the mechanism — how a manifest actually gets built, attached, and verified.

The manifest store

A C2PA-aware file doesn't carry just one flat block of data — it carries a manifest store, which can hold multiple manifests. One is marked as the active manifest (the most recent, representing the file's current state); others can represent earlier stages if the file's history is being tracked across edits.

What's inside a manifest

Each manifest is a structured record built from a few core pieces:

• Claim generator — identifies the software/hardware that produced this specific claim (e.g. a camera model, an editing app, an AI generator), often with a version number.
• Assertions — the actual claims: what happened (captured, edited, composited, generated), when, and sometimes by whom. C2PA defines a standard vocabulary of action types so different tools' claims are comparable.
• Ingredients — references to source assets that contributed to this file, letting a manifest describe a chain: "this image was generated from these two source images." An ingredient can itself carry its own manifest, so provenance can nest across several edits.
• Thumbnail — an optional preview embedded in the manifest for quick display without decoding the full asset.

Signing

Once a manifest is assembled, it's cryptographically signed — hashed and signed with a private key, producing a signature that can later be verified against the manifest's content and the signer's public certificate. This is what makes C2PA data tamper-evident: if anyone modifies the manifest after signing without redoing the signature, verification fails. It's the same general cryptographic principle behind HTTPS certificates or signed software packages, applied to media files instead.

Verification

When a reader (like our C2PA Checker) opens a file, it:

1. Extracts the manifest store from the file
2. Recomputes the hash of the manifest's content and checks it against the signature
3. Checks whether the signing certificate chains up to a recognized, trusted issuer

This produces the Trusted / Valid / Invalid result explained in What Does "C2PA Manifest Detected" Mean? Signature verification confirms the manifest wasn't altered after signing and, for Trusted, that a recognized authority vouches for the signer — it does not independently confirm that the claims inside the manifest are factually accurate.

Where the data actually lives in the file

C2PA data is embedded using format-specific mechanisms — a JPEG carries it in an APP11 marker segment (using a JUMBF box structure), PNG uses a dedicated caBX chunk, and video containers use their own equivalent boxes/atoms. Because it lives in a removable segment like any other embedded metadata, it can be stripped the same way EXIF or XMP can — see Does Removing Metadata Remove an AI Label? for what that does and doesn't accomplish.