Guide
C2PA content credentials, explained
A signed record of how a file was made, attached to the file itself. Not a watermark, not DRM, and more consequential than either.
Updated
What they are
C2PA — the Coalition for Content Provenance and Authenticity — is a standard backed by Adobe, Microsoft, Google, the BBC and others. It defines a manifest that records who made a file, with what tool, when, and what happened to it since. The manifest is cryptographically signed, so alteration is detectable.
In marketing they are called Content Credentials. Functionally, they are a chain of custody.
Where they live
A manifest is stored in a JUMBF box, and the container decides where that box sits:
- JPEG — an
APP11marker segment - PNG — a
caBXchunk - HEIC, AVIF and MP4 — a
jumbbox, or a top-leveluuidbox
This is why generic EXIF tools miss them entirely. C2PA is not an EXIF field, so a tool that only knows how to clear an EXIF block leaves the manifest exactly where it was.
What removing them does
Removing a manifest deletes the signed history. A verifier then reports the file as carrying no credentials — the same state as any ordinary photograph that never had them.
It is worth being precise about what that is not:
- It does not forge anything. The signature cannot be edited to say something false; it can only be present or absent.
- It does not make an image undetectable as AI. Metadata is one signal among several. Pixel-level classifiers work on the picture. Invisible watermarks such as SynthID are embedded in the pixels and survive metadata removal completely.
- It does not remove credentials from copies. Anything already published elsewhere keeps its manifest.
Legitimate reasons to remove them
Provenance data is identifying data. A manifest can name the account, the application, the device and the timeline behind a file. Reasonable cases include:
- Delivering work to a client without exposing your internal tooling and workflow.
- Publishing under a pseudonym when the manifest carries your real identity.
- Removing an employer's or a licensed application's identifiers from personal work.
- The same reason you strip EXIF: reducing what a file says about you by default.
The rules worth knowing first
The EU AI Act's Article 50 sets transparency obligations for AI-generated content, applying from 2 August 2026. Most large platforms separately ask that AI content be disclosed, and some apply labels automatically based on what they find in the file.
Cleaning metadata for privacy is an ordinary thing to do. Removing a disclosure so that a person, a client, a publication or a competition judge believes something false is a different act, and it is judged differently. This tool will show you what is in your file and remove it if you ask. What it will not do is help with the second thing — which is also why it does not implement pixel modification aimed at defeating detectors.
You can inspect what your file carries with the EXIF viewer, or go straight to the AI image metadata remover. If your images come from ChatGPT specifically, there is adedicated page for that.