Guides
What Is C2PA? Content Credentials Explained Simply
C2PA Content Credentials are a signed record of who made an image or video, with which tool, and how it was edited. What they record, how to inspect them.
C2PA (Coalition for Content Provenance and Authenticity) is an open technical standard for Content Credentials: a cryptographically signed record, attached to an image, video, audio file or document, that says which device or software created it and what was done to it afterwards, including whether generative AI was used. Anyone can inspect it with a verify tool. It tells you the file's history as recorded by the signer; it doesn't tell you whether the content is true.
Who is behind it
C2PA was announced in February 2021 by Adobe, Arm, the BBC, Intel, Microsoft and Truepic. It brought together two earlier efforts: Adobe's Content Authenticity Initiative (CAI) and Project Origin, backed by Microsoft and the BBC. It's a project of the Joint Development Foundation, a non-profit, and publishes its specification openly. The newest version listed on the C2PA specification site, 2.3, is dated January 2026.
"Content Credentials" is the consumer-facing name, shown as a small CR pin on sites and apps that support it.
What a Content Credential records
A Content Credential (in the specification, a C2PA manifest) is made of a few parts:
| Part | What it holds |
|---|---|
| Assertions | Statements about the file: actions taken (created, opened, edited, cropped), the tool used, a digital source type such as "trained algorithmic media", thumbnails |
| Ingredients | The source files used to make this one, each with its own history, forming a provenance tree |
| Claim | The list of assertions the signer stands behind, plus hashes that tie them to the exact file contents |
| Signature | A digital signature with an X.509 certificate identifying the hardware or software that signed |
The hard binding is a set of hashes of the file itself. If someone changes even part of the image after signing, the hash no longer matches and a verifier reports it.
Since version 2.0, the specification focuses on identifying hardware and software signers rather than individual people. Personal identity can be added through extensions, and the specification notes the privacy implications of doing so.
Where it's stored
Usually the manifest is embedded in the file, inside a container format called JUMBF (ISO/IEC 19566-5). In a JPEG it sits in APP11 segments. A manifest can also be external, referenced by a URL, which lets a large history live in the cloud.
Because an embedded manifest is just data in the file, it disappears when that data is removed: a screenshot, a re-encode, or an upload to a service that strips metadata. The specification addresses this with soft bindings, such as invisible watermarks or fingerprints, that let a verifier look up a credential stored in the cloud even after it was removed from the file.
Who signs, and what "trusted" means
The signer is the claim generator: the camera, phone, editing app or AI service that wrote the manifest. It signs with an X.509 certificate, often with a trusted timestamp.
A verifier checks three things:
- Well-formed: the manifest follows the specification.
- Valid: the signature checks out and the hashes still match the file.
- Trusted: the signing certificate chains to a list the verifier trusts, such as the C2PA Trust List.
Google, for example, says it surfaces C2PA information in About this image only when the manifest is signed by a certificate from a Certification Authority on the C2PA Trust List.
Who uses it
Support keeps growing, but it's far from universal:
- AI image tools: OpenAI adds C2PA metadata to ChatGPT and API images; Google adds it to images from some of its models; Adobe Firefly adds it.
- Cameras and phones: a number of camera makers and some phones sign photos at capture.
- Platforms and search: Google reads C2PA for "About this image"; Meta reads C2PA and IPTC signals to apply "AI info" labels; LinkedIn shows the CR pin on some images. The Gemini app reads Content Credentials when you ask whether an image is AI-made.
Many apps and sites still strip metadata on upload, so most images you see online carry no credential at all.
How to inspect Content Credentials
In the browser:
- Open the Verify tool at contentcredentials.org/verify (it redirects to the Content Authenticity Initiative's verifier).
- Drop in the original file.
- Read the summary: who signed it, when, which app or model, the actions recorded and any ingredients.
On a computer: the open-source c2patool command-line tool prints a JSON report of the manifests in a file.
On websites: hover or tap the CR pin where it appears.
Without uploading: our AI image checker finds a C2PA manifest in the file and lists the app or model, the recorded actions and the declared source type, in your browser. It doesn't verify the signatures, so use the Verify tool when authenticity matters.
If you just want to see ordinary EXIF fields (camera, date, GPS), our photo metadata viewer reads them in your browser without uploading the file. How to check photo metadata covers the iPhone side.
What C2PA doesn't do
- It doesn't judge truth. The C2PA's own explainer says provenance alone can't tell you whether content is true or accurate. A staged photo signed by a real camera is still staged.
- It can't force anyone to use it. Absence of a credential means nothing on its own.
- It doesn't survive every copy. Without a soft binding, a stripped file has no link back to its history.
- It isn't a detector. It records what the signer declared; it doesn't analyze pixels.
That's why it's best used with other checks, as described in how to check if an image is AI-generated. And if you're wondering what stripping metadata does to an AI label, see can you remove an AI watermark from an image.
Privacy: check before you share
Content Credentials are designed to be shared, and an app might include details you'd rather not publish, such as account names or edit history. Before sending a file, look at what it carries. To remove EXIF and other metadata for privacy, the EXIF remover re-encodes the image locally.
sigil. works on the same principle for your private photos: it strips EXIF and location metadata when you import photos into its encrypted vault and when you export them, and the vault opens only with a sign you draw.
Bottom line
C2PA is the open standard behind Content Credentials: a signed, tamper-evident history of how a file was made and edited, including AI use. It's strong evidence when present and valid, silent when stripped, and never a verdict on whether the content is true.
Sources
- C2PA Technical Specification 2.2 and Explainer
- C2PA founding press release (c2pa.org)
- About C2PA (c2pa.org)
- Content Credentials (contentcredentials.org)
- c2patool (Content Authenticity Initiative, GitHub)
- Image metadata in Google Images (Google Search Central)
- C2PA in ChatGPT images (OpenAI Help Center)
- Labeling AI-generated images (Meta)
Frequently asked questions
What does C2PA stand for?
Coalition for Content Provenance and Authenticity. It's the standards group, founded in 2021 by Adobe, Arm, the BBC, Intel, Microsoft and Truepic, that publishes the open specification behind Content Credentials.
Are Content Credentials the same as a watermark?
No. Content Credentials are signed metadata attached to the file and can be removed with it. A watermark is hidden in the pixels. The C2PA specification lets the two work together so a removed credential can be found again.
Does C2PA prove an image is real?
No. It proves who signed a record about the file and that the record hasn't been altered since. Whether the content itself is true is a separate question.
How can I see an image's Content Credentials?
Upload it to the Verify tool from the Content Authenticity Initiative, use the open-source c2patool on a computer, or click the CR pin on sites that display it.
Do Content Credentials include my name or location?
Only if the app that created them adds that. Identity is optional in the standard, and some apps let you choose what to include. Check a file before sharing it.