TL;DR: Adobe Content Credentials are Adobe's implementation of the open C2PA provenance standard. When you export from Photoshop, Lightroom, Camera Raw, Firefly, or Adobe Express with the feature on, Adobe writes a cryptographically signed manifest into the file — a JUMBF box carried in JPEG APP11 marker segments or a PNG caBX chunk, holding assertions, a claim, and a COSE signature. Opening the file in Metadata Cleaner and clicking Clean rewrites the file and removes that embedded manifest along with EXIF and XMP, without re-compressing the picture. What removal does not reach: a durable Content Credential, which also stores an invisible TrustMark watermark in the pixels and a fingerprint in Adobe's cloud, both designed to recover the record even after the embedded manifest is stripped.
You exported a JPEG, looked at the picture, and assumed the picture was the whole file. Adobe had quietly added a signed footnote saying which of its tools touched the image and whether Firefly generated any of it. We think that's worth understanding before you decide whether the footnote travels with every copy you send.
This post covers what Adobe Content Credentials are, which apps write them, what sits inside the file, how to remove the embedded manifest, how to confirm it worked, and what removal does not reach.
What Are Adobe Content Credentials?
Content Credentials is Adobe's brand name for provenance metadata built on C2PA — the Coalition for Content Provenance and Authenticity standard — under the Adobe-led Content Authenticity Initiative. The standard defines the format; Content Credentials is the product layer Adobe ships on top of it, including the little "Cr" pin you see on a marked image and the human-readable history that Adobe's Verify tool displays.
The two terms are close but not identical. C2PA is the open technical specification — it defines the manifest format, the JUMBF container, and how the box is embedded per file type. Content Credentials is Adobe's implementation of that standard, so when an Adobe app writes a credential it is writing a C2PA manifest. For removal that matters because it tells you what you are deleting: stripping "Content Credentials" from a file means removing the embedded C2PA manifest store, the same removable layer we cover for generators in Midjourney, DALL-E, and Stable Diffusion EXIF and explain in full in what C2PA metadata is.
The credential answers one question for whoever opens the file later: where did this come from, and was a generative tool involved. That is reasonable for a newsroom to want attached, and a disclosure a creator may not want on every casual upload — different situations that deserve different defaults.
Which Adobe Apps Attach Content Credentials?
More than people expect, and the behaviour differs by app. Adobe has rolled Content Credentials support across its creative line: Photoshop, Lightroom, Camera Raw, Adobe Express, and Firefly all can attach them, and Adobe Stock requires provenance on generative submissions.
The split that matters is automatic versus optional. For anything generated by Firefly — whether in the Firefly web app, Generative Fill in Photoshop, or generative features in Express — Adobe attaches Content Credentials automatically, and for fully Firefly-generated assets you cannot turn that off. For ordinary edits in Photoshop, Lightroom, and Camera Raw, Content Credentials are opt-in: there is a toggle in the export settings that controls whether the credential is written to the file. Turn it on and your exports carry a manifest; leave it off and they don't.
Photo by Stefan Lobont on Pexels.
The practical consequence is that you often cannot tell whether a file is marked by looking at your workflow. A designer who used Generative Fill once to heal a power line gets the same "Firefly was used" assertion as someone who generated an entire synthetic scene — the manifest records that a generative tool ran, not how much of the frame it changed. If a credential is going to affect how a platform treats your upload, you want to check for it and remove it deliberately, not discover it after the fact — the same pattern we walked through for Photoshop Content Credentials.
What's Actually Inside the Credential?
The credential is not loose text in an EXIF field. It is a structured object called a manifest, built to be tamper-evident. A manifest carries three things: a set of assertions (statements such as "created with Adobe Photoshop" and "Adobe Firefly generative AI was used"), a claim that binds those assertions to the specific bytes of the file, and a digital signature over the claim. The assertions are encoded in JSON or CBOR, and the signature uses COSE — typically an ECDSA P-256 key — so any later reader can confirm the credential hasn't been altered.
Where does that physically live in the file? In a JUMBF box — the JPEG Universal Metadata Box Format the C2PA spec defines per format. In a JPEG, the manifest store is written into APP11 marker segments; because a single JPEG marker segment maxes out at 65,535 bytes, a larger manifest is split across several contiguous APP11 segments. In a PNG it goes in an ancillary caBX chunk placed before the IDAT image data; in WebP it rides in its own RIFF chunk, the structure we picked apart in WebP metadata. Either way it is a distinct block sitting alongside the pixels, not baked into them — which is exactly why a clean rewrite of the file can remove it.
Photo by energepic.com on Pexels.
To be precise about what the manifest asserts: it records the tools and the generative step. It does not grade the size of your edit, and on its own it is a record in a file you control — not a tracking beacon. The complication, covered below, is that Adobe also offers a version that does not stay confined to the file.
How to Remove Adobe Content Credentials
The removal itself takes a few seconds. Metadata Cleaner runs entirely inside your browser tab — the file's bytes are read into memory and written back to disk locally, never uploaded to a server, which is the architecture behind everything we build.
- Open Metadata Cleaner in any browser. Desktop or mobile both work; there is nothing to install and no account to create.
- Drag your exported JPEG, PNG, TIFF, or WebP into the drop zone, or tap to pick it on a phone. The file loads into the tab's memory.
- Click Clean. The tool rewrites the file and removes the C2PA manifest store — for a JPEG it drops the
APP11segments holding the JUMBF box; for a PNG it removes thecaBXchunk — along with the EXIF, XMP, and IPTC metadata blocks. The pixel data is copied through unchanged. - Click Download. The cleaned image lands back in your downloads, visually identical to the file Adobe gave you and a few kilobytes lighter.
The picture is byte-for-byte the same after the clean. Removing the manifest does not re-encode the JPEG or re-filter the PNG; it edits the metadata blocks beside the image stream, so there is no second round of compression and no quality loss. A typical export cleans well under a second.
If you would rather not write the embedded credential in the first place, the toggle in your Adobe app's export settings controls that going forward. It does not retroactively clean files you already exported, and for fully Firefly-generated assets it is not offered — which is why a file-level strip is still the reliable path for anything already on disk.
How to Verify the Content Credentials Are Gone
Trusting a tool is fine; checking it is better. Three ways to confirm, in rising order of authority:
- Adobe's Verify tool. Upload the cleaned file to contentauthenticity.org. If the embedded manifest is gone, Verify reports no Content Credentials found in the file. Read the next section before treating that as final, though — Verify can also recover credentials that were never in the file.
- ExifTool. Phil Harvey's
exiftoolreports JUMBF and APP11 content. Runningexiftool -G1 -a -u cleaned.jpgdumps every group; after a clean, the JUMBF / C2PA group and theAPP11segments should be absent, along with EXIF and XMP. - A hex check on the markers. A JPEG with a manifest contains
FFEB(APP11) marker bytes carrying aJPsignature and ajumbbox label; a PNG carries acaBXchunk name. After a clean, those markers are gone.
Photo by Vlad Deep on Pexels.
One distinction to keep straight in Verify's output: "no credentials in this file" means the embedded manifest is removed — the part the clean controls. "Recovered from the Content Credentials cloud," if it appears, is a different mechanism, and it is the subject of the next section.
What Removing Content Credentials Doesn't Do (Honest Limits)
Removing the embedded manifest is a real, verifiable change — but it is one layer, not a cloak, and anyone claiming otherwise hasn't read Adobe's own documentation. The reason is durable Content Credentials, which Adobe built on what the Initiative calls three pillars of provenance: the cryptographically signed metadata, an invisible watermark, and a digital fingerprint. The three are designed to back each other up so the provenance survives the loss of any one of them.
The watermark survives the strip. Adobe's invisible watermark, called TrustMark, is encoded into the pixels themselves and is built to persist through processing that destroys metadata — re-compression, resizing, even a screenshot. A metadata clean edits the file's metadata blocks; it does not alter pixel values, so it does not touch a pixel-domain watermark. If the watermark carries an identifier, a verifier can use it to look up and recover the provenance record from Adobe's cloud even though the embedded manifest is gone.
The cloud fingerprint survives too. Fingerprinting computes a perceptual hash from the pixels and matches your image against copies Adobe holds. Strip the file and the fingerprint of the visible picture is unchanged, so a match can still recover the record. Between the watermark and the fingerprint, a durable credential is specifically engineered so that removing the embedded copy does not make the image anonymous.
Server-side records and platform behaviour are also out of reach. Whatever Adobe logged when Firefly generated the pixels lives on Adobe's side, independent of the file you hold. And platforms like Instagram and TikTok run their own AI detection and read the IPTC Digital Source Type property in addition to C2PA — removing the embedded manifest takes one input away from how they classify your upload, but it is not a guarantee of the label, as we covered in platform-side detection.
The honest framing: cleaning an Adobe export reliably removes the embedded C2PA manifest — the APP11 / caBX JUMBF box that travels inside the file to every recipient, and the thing most readers mean when they say "the Content Credentials." If the credential was made durable, the TrustMark watermark in the pixels and the cloud fingerprint are a separate matter no file-level metadata tool can clear. Know which one you are dealing with before you assume an image is unmarked. This is the same point we make about audio in metadata versus watermarks: the privacy you control is the metadata layer, and it is worth controlling — just not by pretending the other layers don't exist.
FAQ
Does removing Content Credentials change the image quality?
No. The clean rewrites the metadata blocks only — the JUMBF manifest store, EXIF, XMP, and IPTC — and copies the pixel data through unchanged. There is no re-compression, so resolution, color, and detail match the Adobe export exactly.
Will the cleaned image still open everywhere?
Yes. A JPEG, PNG, TIFF, or WebP with its C2PA manifest removed is a standard image file. Every viewer, editor, browser, and platform opens it normally — it is just a few kilobytes smaller.
If I remove the manifest, can anyone still tell the image used Firefly?
Possibly. If the credential was a durable Content Credential, the invisible TrustMark watermark in the pixels and the fingerprint match in Adobe's cloud can recover the provenance even after the embedded manifest is gone. Removing the manifest clears the embedded layer, not those.
Can I stop Adobe from writing Content Credentials in the first place?
Partly. Photoshop, Lightroom, and Camera Raw have an export toggle that controls whether the credential attaches to the file going forward. It does not clean files you already exported, and for fully Firefly-generated assets Adobe attaches credentials automatically with no opt-out.
Can I remove Content Credentials on my phone?
Yes. Metadata Cleaner is browser-only and runs on mobile Safari, Chrome on Android, and Firefox mobile. Drag-and-drop becomes tap-to-pick, and the cleaned file saves to Files or your camera roll.
Is removing Content Credentials legal?
Yes. A C2PA manifest is metadata in a file you own, and no general law requires you to keep it — a point consistent with how the EFF frames control over your own file metadata. Narrow professional contexts, like newsroom or evidentiary chains of custody, may expect provenance to be preserved; outside those, whether the credential travels with your image is your call.
Exported something from an Adobe app and would rather the AI notice didn't ride along inside the file? Strip the embedded manifest before it leaves your machine. Try Metadata Cleaner free — drop the image, hit Clean, download. The pixels are yours; the signed paper trail in the metadata doesn't have to come with them.