C2PA Metadata Explained
What C2PA metadata means
"C2PA metadata" is a useful informal phrase, but it can understate what the technology does. C2PA provenance data is carried in a Content Credential, represented and processed as a C2PA Manifest. That manifest can contain claims, assertions, ingredients, signature information, and bindings to the asset.
Technically, the credential is more than a conventional metadata block. Its signed structure and cryptographic asset bindings make it possible for a validator to check whether the recorded provenance and the bytes of the asset still correspond.
The important distinction is that these fields are part of a tamper-evident provenance record. Conventional metadata can describe a file; a Content Credential also carries signature information and hash-based bindings.
C2PA vs EXIF vs XMP vs IPTC
EXIF commonly carries camera capture settings and dates. XMP is an extensible metadata framework used by creative tools. IPTC describes editorial and rights information such as captions, creators, and licensing terms.
| EXIF | XMP | IPTC | C2PA | |
|---|---|---|---|---|
| Typical purpose | Camera data | Extensible metadata | Editorial and rights metadata | Verifiable provenance |
| Common fields | Camera, date, and GPS fields | Creative and workflow data | Caption, creator, and licensing | Claims, assertions, actions, and ingredients |
| Cryptographically bound | No, not by EXIF itself | Not inherently | Not inherently | Yes, through manifest signatures and asset bindings |
| Tracks recorded changes | No | Not inherently | Not inherently | Yes, for actions and ingredients included in the record |
| Can coexist with others | Yes | Yes | Yes | Yes |
C2PA can reference or include metadata-like assertions, but it is not a replacement for every one of those systems. A C2PA manifest may coexist with EXIF, XMP, and IPTC blocks in the same JPEG. Conversely, a file can have rich EXIF and XMP but no C2PA manifest.
Conventional metadata fields are not, by themselves, a cryptographically signed provenance record. They may be modified or removed without the kind of cryptographic verification provided by C2PA.
What a C2PA credential can contain
A viewer may expose several layers of the credential:
| Part | What it represents |
|---|---|
| Manifest store | The collection of manifests in or associated with the asset |
| Active manifest | The manifest that describes the asset currently being inspected |
| Assertions | Structured statements such as actions, capture data, or declared digital source |
| Ingredients | References to source assets used during production |
| Signature information | Signing algorithm, signer, issuer, and related certificate details |
| Asset hash and binding | Cryptographic information that associates the record with the exact asset |
| Source disclosure | A producer declaration such as a digital-source type |
Not every producer includes every field. A credential only reports the claims, assertions, and ingredients that were actually recorded.
Fields you may see
A C2PA viewer can show the credential status, signer, issuer, signing algorithm, signed time, claim generator, actions, ingredients, assertions, manifest identifier, and asset hash or binding information. It may also show a declared digital source when the producer recorded one.
Each field has a different meaning. A signer proves that a credential was signed by a certificate, not that the signer observed the scene. An action proves that an action was recorded, not that every other edit was absent. A source declaration reports what the producer declared, not an independent AI detector result.
What happens when C2PA metadata is stripped
Screenshots, re-encoding, and some export workflows create a new file without copying the original manifest. That file can accurately report no C2PA metadata even if an earlier version had a valid credential.
Removing a credential from a copy does not prove anything about the original. The original capture or production file may still be available, still contain its credential, and still validate under a suitable policy. This is why the provenance question is tied to a specific file, not merely to a story or filename.
Some workflows also use durable Content Credentials or soft-binding techniques to help a credential be discovered after it has been removed from an asset. Whether that recovery path works depends on the implementation, the discovery service, and the support available in the viewer. Absence of an embedded credential should therefore be reported as a property of the inspected copy, not as proof that no provenance ever existed.
C2PA metadata vs traditional metadata
Traditional metadata is often descriptive and editable. It can be accurate, useful, and professionally maintained, but ordinary EXIF, XMP, and IPTC fields do not by themselves establish a signed provenance chain.
C2PA changes the question from "what does this file say about itself?" to "what signed provenance record is bound to these exact bytes, and does it validate?" Those are complementary questions, not identical ones.
Inspect only what is actually supported
C2PA Lab parses the C2PA manifest through its browser SDK. It does not separately parse or validate all EXIF, XMP, or IPTC blocks. That boundary is intentional: presenting unsupported metadata as verified would be misleading.
For a structured field-by-field view, use the C2PA Metadata Inspector. If you need every assertion and the raw JSON tree, open the C2PA Manifest Viewer.
Inspect C2PA metadata
Open the metadata inspector in your browser and review the fields actually present in your own file. No upload is required.
Open Metadata InspectorOfficial reference
For the representation of claims, assertions, signatures, manifests, and asset bindings, see the official C2PA specifications.
FAQ
Is C2PA just another metadata format?
No. C2PA uses structured data, but its manifest is a signed provenance structure with hash-based bindings to the asset. That is why validation can detect changes that would silently affect ordinary descriptive metadata.
Can EXIF, XMP, IPTC, and C2PA exist in the same file?
Yes. They serve different purposes and can coexist. A file may also contain any combination of them or none of them.
Does missing C2PA metadata prove manipulation?
No. Most files still do not carry a credential, and a screenshot or re-encoded copy can remove one from a file that originally had it. Absence describes the inspected file; it is not an authenticity verdict.
Related guides
Background: What is a C2PA manifest? and What are Content Credentials?.