AI image metadata
What Is C2PA? Content Credentials Explained
Published October 21, 2026 · 8 min read
C2PA Content Credentials are signed provenance records attached to photos and PDFs. Here is how the standard works, where it is stored, and what removing it does and does not mean.
The problem C2PA tries to solve
AI-generated images are now indistinguishable from photographs by eye. The C2PA standard (Coalition for Content Provenance and Authenticity), backed by Adobe, Microsoft, Intel and others, attacks the problem from a different direction: instead of guessing from pixels, it attaches a signed, tamper-evident record of where the content came from — which camera or tool created it, what edits were applied, and by whom.
These records are called Content Credentials. When you see a small "CR" pin or a credential popup on an image, you are looking at a C2PA manifest.
Where the manifest is stored
The manifest contains a cryptographic signature over a hash of the file content, which is what makes it tamper-evident rather than merely present.
- JPEG: an APP11 marker segment containing a JUMBF (JPEG Universal Metadata Box Format) container.
- PNG: a caBX auxiliary chunk.
- WebP: an XMP-less dedicated C2PA chunk in the RIFF container.
- HEIC/AVIF: a UUID box in the ISO-BMFF container.
- PDF: a C2PA manifest in the document's XMP and embedded structures.
What happens when you re-encode an image
Cleaning or re-encoding an image with normal tools — including canvas re-encoding in the browser — strips the C2PA manifest, because the manifest is just another metadata structure and re-encoding does not preserve it. That has two consequences.
First, it removes the AI-generation record from an AI image: a cleaned file no longer carries its Content Credential. Second, and more importantly for trust, it also destroys the provenance of a real photograph. A news photo that survives verification only until its first edit is a standard with a real adoption problem. For the practical difference between EXIF fields and these signed manifests, see does removing EXIF data remove AI labels?.
What C2PA does and does not prove
A valid manifest tells you the file matches a signed history. It does not tell you the history is true — a tool can record false claims about its input. And the absence of a manifest proves nothing either way: it may never have existed, or it may have been stripped in transit. Platforms that re-compress uploads routinely remove credentials without any privacy intent.
For a practical privacy workflow, the takeaway is: if you want to remove provenance and AI-generation records from your own files, cleaning the metadata does it. If you want to verify someone else's image, a missing credential should make you more skeptical, not more trusting.
Check your own files
Inspect and clean metadata from images and PDFs directly in your browser. Files are never uploaded to a server.
Check an image's metadata now →Related guides
- Does Removing EXIF Data Remove AI Labels?Short answer: it depends where the AI label is stored. A field-by-field breakdown of EXIF, XMP, PNG text chunks, C2PA records, and what a full cleaning pass actually removes.
- SynthID and Invisible Watermarks: What Metadata Cleaning Cannot RemovePixel-level watermarks like Google's SynthID survive metadata removal, re-encoding and screenshots. Understand what they are, what survives, and what that means for AI images.
- Stable Diffusion Metadata: Parameters, Prompts and Workflows in PNG FilesPNGs from AUTOMATIC1111, ComfyUI, NovelAI and InvokeAI carry the full generation recipe in text chunks. Here is what each tool writes and how to read or remove it.