Image metadata checker
C2PA metadata checker and remover for PNG/JPEG
Check PNG and JPEG files for supported C2PA Content Credentials and AI-source metadata, review each finding, then download a targeted cleaned copy. The file stays in your browser, unrelated metadata is preserved, and supported removal does not recompress image pixels.
Open the image metadata checkerWhat are C2PA Content Credentials?
Content Credentials are the public name for C2PA manifests. A manifest can record claims about an asset's origin, edits, tools, and use of AI, together with signatures and content bindings. In PNG and JPEG files, a manifest store can be embedded in metadata containers.
Content Credentials are not automatically an “AI generated” label. Cameras and human editing workflows can create them too. Their presence is provenance information, not proof of how the pixels were made; their absence is not proof that an image is human-made.
What this image metadata checker inspects
| File structure | Check | Result |
|---|---|---|
| PNG | Structural C2PA chunks and bounded text metadata | Removable container or review-only field |
| JPEG | Reassembled JPEG-XT/JUMBF C2PA stores | Complete removable APP11 group |
| PNG/JPEG XMP | Recognized provenance and AI-source fields | Reported as review-only when shared |
Structural C2PA detection requires the manifest-store box type, exact C2PA UUID, and
c2pa label. The checker does not treat every APP11 segment or broad word such as
“generated” as a watermark.
Does Claude add C2PA metadata to images?
Anthropic says supported Claude file outputs, including SVG, PNG, and JPG, can carry signed provenance metadata using C2PA. This tool supports PNG and JPEG only. It can find and remove supported embedded structures, but it does not support SVG or validate the signature, signer, trust chain, claim, or content binding.
Compare Claude's embedded text mark with Claude file metadata.
Targeted removal without image recompression
The cleaner removes only metadata ranges identified as safe to remove. PNG IDAT
data and the JPEG byte stream from SOS onward are copied exactly. That preserves
the existing compressed image data rather than drawing the image to a canvas and encoding it
again.
- Preserves unrelated EXIF, ICC, IPTC, comments, and copyright metadata.
- Leaves shared XMP packets in place and reports matching fields as review-only.
- Checks the cleaned container again before enabling download.
- Supports PNG and JPEG files up to 25 MB and 24 megapixels.
What it does not remove
Visible logos and text overlays are pixels, not metadata. SynthID Image and other imperceptible pixel signals also remain outside this checker. C2PA soft bindings can locate external provenance after embedded metadata is removed. None of these are changed.
C2PA metadata remover FAQ
Does this tool validate C2PA signatures?
No. It structurally identifies supported embedded manifest stores and metadata fields. It does not validate signatures, certificates, trust chains, or provenance claims.
Does C2PA removal change image pixels?
For supported removals, no. PNG image data and the JPEG compressed scan are copied byte-for-byte while identified metadata containers are removed.
Does it remove all image metadata?
No. The cleaner is deliberately targeted. It preserves unrelated EXIF, color profiles, copyright data, comments, and shared XMP instead of stripping everything.
Can Content Credentials come from a camera?
Yes. C2PA can document camera capture and human editing as well as AI-assisted workflows. A C2PA finding is not automatically an AI label.
Does no finding mean the image is authentic?
No. Metadata is often removed during export or sharing. A clean result means only that no supported embedded signal was found.
Is the image uploaded?
No. The selected file is read by browser APIs and is never placed in an application request. The site has no upload endpoint or database.
Source: C2PA Content Credentials specification. Last reviewed August 12, 2026.