Firefly watermark detector
What the Adobe Firefly image watermark detector does
Firefly does not stamp a logo into the pixels of what it makes. The mark people are looking for is Content Credentials: a cryptographically signed C2PA manifest riding in the file's metadata, declaring which tools touched the file and whether a model generated it.
This detector reads that record. Drop an image into the upload control and it goes into our media workflow, where the file is parsed, any manifest is located, and what it declares is reported back. The file is genuinely uploaded, because reading a manifest requires reading the file.
How it works
How to detect a Firefly watermark in three steps
One pass through the upload control on this page.
- Step 1
- Use the file itself, not a screenshot of it. A screenshot is a new image and carries none of the original file's metadata.
- Step 2
- Drop it into the box above. The file is read and any C2PA manifest is parsed and checked.
- Step 3
- Read the verdict against the three outcomes below, because found and not found are not the only two answers.
Reading the result
The three answers a Content Credentials check can give
A detector that only reports yes or no cannot explain its own output. There are three distinct states and they mean very different things.
- A manifest present and valid
- The signature checks out against a certificate. This is strong evidence about what the file declares, because a signed record is hard to forge and easy to verify.
- A manifest present but invalid
- A record exists but the signature does not verify, or the content no longer matches what was signed. That usually means the file was edited after signing.
- No manifest at all
- The weakest result of the three. It is consistent with an image that never had credentials, and equally consistent with a Firefly image whose credentials were stripped.
The important part
Why a clean result does not mean a real photo
Content Credentials are metadata, and metadata is fragile. Re-encoding a file, taking a screenshot, or uploading through a pipeline that recompresses images all strip the manifest as an ordinary side effect. Nobody has to be trying to remove it.
So the two results are not symmetrical. Finding a manifest tells you a great deal, because it is signed. Not finding one tells you almost nothing, because the most common reason a file has no manifest is that it passed through something that removed it. Treat a blank result as no information rather than as an all clear.
Inside the manifest
What a Firefly Content Credentials record actually says
A C2PA manifest is a signed structure holding assertions about the file. The one that matters for this question is the digital source type. When it is set to trainedAlgorithmicMedia, the record is declaring that a trained model produced the content, which is the assertion that separates a generated image from a photograph that happens to carry credentials.
Alongside that sit the claim generator, meaning the tool that wrote the record, the edit history as a chain of ingredients, and the certificate the signature was made with. A manifest naming Firefly is one possibility among many: cameras, editors and other generators write manifests too, so a found record does not automatically mean Firefly.
Mechanics
Why screenshots lose the watermark
The manifest is stored inside the file format itself, in structures built to carry arbitrary data: an APP11 segment in a JPEG, a metadata chunk in a PNG, a uuid box in an MP4. None of that is part of the picture. It is attached to the picture.
A screenshot rebuilds the image from what was on screen, so nothing attached to the original file comes with it. The same logic explains most of the other ways credentials vanish: anything that writes a new file writes it without the old record unless it deliberately carries it forward.
Straight answer
Does a valid credential prove the image is genuine?
No, and this is the part most tool pages skip. A Content Credentials manifest attests to origin and integrity: who signed it, and whether the file has changed since. It does not attest to truth. A misleading or staged image can carry a perfectly valid credential, because the credential is about the file's history rather than about what the picture claims.
The standard is also not invulnerable. Certificate issues have been found and revoked in the wild, and signing keys living on a device can be abused when that device is compromised. A signed manifest is good evidence and it is not a guarantee, which is exactly how this page reports it.
Who it's for
Who checks images for a Firefly watermark and when
Editors and publishers doing provenance checks on supplied artwork before it runs, where a declared generation record changes how the image gets captioned. Agencies auditing what a supplier delivered against what the brief allowed.
Anyone about to reuse an image and wanting to know what history is still attached to the file, and anyone who ran a removal and wants to confirm the manifest is actually gone before publishing.