Someone sent us a photo that shows a valid C2PA Content Credential in a verify tool. What does that actually prove about the image: that it's real, who made it, whether AI was used? And what does it mean when a file has no credential at all?

What should I check, step by step?

1. Check the file in a validator and read the result, not just the badge: is the manifest valid (hashes and signature match), and is the signer trusted?

2. Look at who signed. The signer is the holder of the certificate the claim generator used, often a camera maker, an app or a platform, not necessarily the person who took the photo.
3. Note which trust list the checker used. Content signed under the frozen Interim Trust List and content from conforming products on the official C2PA trust list must be shown differently by validators during the transition.
4. Check for a trusted time-stamp. It lets a validator confirm the signature was made while the certificate was valid even after the certificate expires or is revoked.
5. Read the actions and the digitalSourceType. A generative AI tool that records its use labels the content as AI-generated there; if the field is absent, that only means it wasn't recorded.
6. Separate assertions the signer made (created_assertions) from ones it only gathered (gathered_assertions). Only the first set is attributed to the signer.
7. Treat ingredient history with care. An ingredient's credential was checked when it was added; the original ingredient file usually isn't included, so you can't re-verify it yourself.
8. If there's no credential, don't conclude the file is fake. Look for a durable credential through a watermark or fingerprint lookup where the tool supports it, ask the sender for the original file, and use other checks.

What does a valid C2PA Content Credential prove about a file?

A Content Credential, which the specification calls a C2PA manifest, is a set of assertions about an asset (how it was created, what actions were applied, whether AI was involved) that are hashed, bound to the content and digitally signed with an X.509 certificate. The explainer describes the combination of cryptographic hashes such as SHA-256 and a Merkle-tree-like structure as tamper-evident: any change to the asset or the provenance data breaks the match.

• So a valid result means three things: the provenance data hasn't been altered, the content still matches what was signed, and the signature came from the stated certificate. If the manifest includes a time-stamp from a time-stamping authority on the C2PA TSA trust list, a validator can confirm the signature was made while the certificate was valid, without checking revocation online at the time you view it.
• Generative AI shows up through the digitalSourceType field on the action that created the content. That's only as complete as the generating tool's implementation: the credential records what the claim generator chose to record.

What doesn't a C2PA Content Credential prove, and does a missing credential mean a file is fake?

The C2PA explainer says it plainly: Content Credentials do not provide value judgments about whether a given set of provenance data is true, only whether it's well-formed, free from tampering, valid and trusted. Provenance alone can't tell you whether the content is true, accurate or factual.

• Provenance isn't always complete. The explainer's example is a crop made in a tool that doesn't support Content Credentials: nothing records it, and if the image then goes through a supporting tool, the new signer implicitly attests to the cropped version.
• It isn't a person's identity. The core specification focuses on the content, not on the human or organization behind it; identity claims need an extension such as the Creator Assertions Working Group's identity assertion. And the signer is only attributed with the assertions listed in created_assertions, not those in gathered_assertions.
• Ingredients can't be fully re-checked. When an asset is composed from others, each ingredient's credential is validated at the time it's added and that result is recorded, because the ingredient's own data usually isn't carried along.
• No credential isn't evidence of fakery. Adding provenance is optional and metadata can be removed. The specification's answer to stripping is durable Content Credentials: a soft binding such as an invisible watermark or a fingerprint that lets a tool find the manifest in cloud storage, with a Soft Binding Resolution API added in version 2.2.

Why can the same C2PA Content Credential show as trusted in one verify tool and not another?

Trust in C2PA comes from the signer's certificate. The C2PA trust list is a list of certificate authorities that issue signing certificates to generator products that passed the C2PA Conformance Program, which opened in June 2025; conforming validators must check signatures against it. Conforming products are published on the C2PA Conforming Products List, and the C2PA FAQ says certificates carry the product's assurance level.

• The earlier Interim Trust List was frozen on January 1, 2026: nothing is added or updated, existing certificates stay valid for legacy support, and content signed while a certificate was valid is still considered valid under the legacy model. During the transition validators may consult both lists but must distinguish between them.
• The Content Authenticity Initiative's documentation says the C2PA Verify tool still uses the Interim Trust List, and shows the certificate owner and signing time only when the certificate is on it. So the same file can read as a known signer in one tool and an unrecognized one in another; note which tool and list you used when you record a result.