# Content Credentials Viewer

> Free C2PA viewer: check image provenance in a JPEG, PNG or WebP. See who signed it, when, and whether it changed since. In your browser, nothing uploaded.

- URL: https://bytescope.dev/en/content-credentials/
- Locale: en · Also: [ja](https://bytescope.dev/ja/content-credentials.md) · [zh-Hant](https://bytescope.dev/zh-Hant/content-credentials.md)
- Runs: Your files never leave your browser — all processing is local.
- Requires: Any modern browser (Chrome, Edge, Firefox, Safari), desktop and mobile.
- Updated: 2026-09-08

## What it does

This is a C2PA viewer and content credentials checker for one image at a time. Drop a JPEG, PNG or WebP and the page reads the C2PA manifest embedded in it, verifies the signature, and gives exactly one of four answers: credentials found and verified against the C2PA trust list; credentials found but verification failed; no credentials found; or cannot tell, for a file that is damaged or in a format the page does not read. The work is done by the open-source C2PA library compiled to WebAssembly — about 3 MB, downloaded on first use — so the image itself never leaves the tab.

### What a content credential proves, and what it does not

A verified credential means one narrow thing: this signer signed these bytes, and the bytes have not changed since. It says nothing about whether the picture shows something that happened. A signer can sign anything — a photograph, a composite, a render — and the signature is equally valid on all of them. The opposite reading is just as wrong: a file with no credentials is not suspicious. Most cameras, phones and editors do not write them, and social sites strip them on upload, so 'none found' is the normal result for an ordinary photo. This page does not tell you whether an image is AI-generated or real, and nothing that reads metadata can. If the manifest itself declares that generative AI was used, that is shown as 'the signer declares: …' — their statement, not our finding.

### Check image provenance: what the report lists

When credentials are present the page lays out the signer, the signing time, the software that produced the file, the action history the manifest records, and the ingredients — the source files it was made from — with the thumbnails embedded for them. When verification fails, the reason is kept separate rather than folded into one red badge, because the four cases mean different things: the file changed after signing, the signature itself is broken, the signer is not on the trust list, or the certificate has expired. A signer missing from the list is not a forgery, and a changed file is not a bad signer.

### What your own file reveals before you send it

Credentials are written to be read by strangers, so the same report doubles as a checklist for the sender: does this file carry your name, the device, the software, the edit history, thumbnails of the picture as it looked before you edited it? A cropped image whose ingredient thumbnail still shows the uncropped frame is the case worth catching. This page only reads. To remove content credentials, use the Remove Photo Metadata tool on this site — and treat that as final, because the recipient cannot get back what you stripped, and neither can anyone who wants to verify the file later.

### Limits worth knowing

The trust list is a dated snapshot bundled with the site — 2026-11-04 — and it can go stale: a signer added after that date shows as not on the list until the snapshot is refreshed. The page does not fetch remote manifests and makes no OCSP request. It does not sign files, and it does not detect pixel watermarks such as SynthID; those live in the image data, not in metadata. Signed JPEG is tested. PNG and WebP are supported by the library, but we have only tested them without credentials. Video and PDF are not supported yet.

## Limits

Does not tell whether an image is AI-generated or authentic, does not remove or add content credentials (use remove-exif to remove), does not fetch remote manifests or check certificate revocation online, and does not read video or PDF yet.

## FAQ

### Can this tell whether an image is AI-generated?

No — and no tool can decide that from metadata. Content credentials are a signed statement by whoever made or edited the file; they are not a detector. If the manifest declares generative AI use, the page shows it as 'the signer declares: …', which is that signer's claim and nothing more. A missing declaration proves nothing either, because the credentials may never have been written or may have been stripped along the way. This page also does not look for pixel watermarks such as SynthID.

### Does this image have C2PA credentials — and what if it has none?

Drop it in and the first line of the report answers that. If the answer is none found, read nothing into it: most cameras, phones and editors do not write content credentials, and social sites strip them on upload. An image with no credentials is neither fake nor real on that evidence; it is simply unsigned. 'Cannot tell' is a separate result, used when the file is damaged or not a format the page reads, so that a file we could not parse is never reported as having none.

### It says verified. Does that mean the picture is genuine?

No. Verified means the signature checks out, the signer leads back to the bundled C2PA trust list, and the bytes have not changed since signing. That tells you who signed the file and that nobody altered it afterwards. It does not tell you the content is true, because the signer can sign anything. Whether to believe the picture is still a judgement about that signer, not something the signature settles.

### Why did verification fail?

The page names one of four reasons and keeps them apart. File changed after signing: the bytes no longer match what was signed, for example after an edit or a re-save. Signature broken: the signature itself does not validate. Signer not on the trust list: the signature may be fine, but the certificate does not lead to an entry in our 2026-11-04 snapshot of the C2PA trust list, and that snapshot can be out of date. Certificate expired: the signing certificate is past its validity period. None of these, on its own, says the image is fake.

### Is my image uploaded anywhere?

No. The file is opened locally and checked by the open-source C2PA library running as WebAssembly in the tab. The only downloads are that library, about 3 MB on first use, and the trust list snapshot bundled with the site. No network request carries your file, remote manifests are not fetched, and no OCSP request is made.

### How do I remove content credentials before sharing a file?

Not here — this page only reads. The Remove Photo Metadata tool on this site keeps content credentials by default and clears them when you ask it to. Before you do, look at the “What this file reveals to the recipient” checklist to see what you would be sending: name, device, software, edit history, pre-edit thumbnails. Then decide knowing it is one-way for the recipient: once the credentials are gone they cannot recover them, and cannot verify where the file came from.

## Related tools

- [EXIF Viewer](https://bytescope.dev/en/exif-viewer.md) — Read every EXIF tag in a photo, including camera, GPS and software.
- [Remove EXIF](https://bytescope.dev/en/remove-exif.md) — Strip this photo's metadata losslessly — pixels are never re-encoded.
- [Image Inspector](https://bytescope.dev/en/image-inspector.md) — Pick colors and read pixel coordinates with a magnifying loupe.
- [Office Document Inspector](https://bytescope.dev/en/office-inspector.md) — List what a Word, Excel or PowerPoint file carries inside it.
