TL;DR: EXIF and C2PA are two different metadata layers that can sit in the same image file. EXIF (Exchangeable Image File Format, maintained by CIPA) records how a photo was captured — camera make and model, lens, exposure, timestamp, and often GPS coordinates — and lives in a JPEG APP1 segment as an embedded TIFF structure. It is unsigned and anyone can edit it silently. C2PA (Coalition for Content Provenance and Authenticity) records a photo's provenance — which tools touched it and whether a generative AI was involved — as a cryptographically signed manifest in a JUMBF box, carried in APP11 segments. EXIF answers "what camera made this"; C2PA answers "what happened to this and can you trust the record." Both are separate blocks alongside the pixels, and Metadata Cleaner removes either one without re-compressing the image.
You opened the file properties, saw a date and a camera model, and figured that was the metadata. Then a platform flagged your upload as AI-assisted and you realized there was a second, invisible layer you had never looked at. EXIF and C2PA are both "metadata," but they were built for different jobs by different people in different decades, and treating them as the same thing is how creators get surprised.
This post covers what EXIF is, what C2PA is, the concrete differences between them, why those differences matter for privacy and for AI labeling, and how to remove each.
What Is EXIF Data?
EXIF is the old layer. It is the Exchangeable Image File Format, originally defined in the 1990s by a Japanese industry group and now maintained by CIPA, the Camera and Imaging Products Association, with the current revision being Exif 3.0 (released 2023, which added UTF-8 text support). Almost every camera and phone writes it.
EXIF describes the moment of capture. When your phone takes a photo, it records the camera make and model, the lens, the focal length, the aperture and shutter speed and ISO, the orientation, the date and time down to the second, and — if location services were on — the GPS latitude and longitude. That is genuinely useful: it is how your photo app sorts shots by date and plots them on a map.
Mechanically, EXIF borrows its structure from TIFF. In a JPEG, the EXIF block is written into the APP1 marker segment, which opens with the bytes Exif\0\0 followed by a complete little TIFF stream: a byte-order marker, an offset to the first image file directory (IFD), and a chain of tagged fields. Because it has to fit inside a single JPEG application segment, EXIF in a JPEG is capped at roughly 64 KB. The same tag dictionary travels in TIFF, PNG, and HEIC files in slightly different containers.
The defining property of EXIF, for our purposes, is that it is unsigned and freely editable. There is no signature, no integrity check, nothing that proves the values are real. Any editor can rewrite the timestamp, blank the GPS, or change the camera model, and nothing in the file records that it happened. That makes EXIF easy to strip — which is exactly what you want for privacy — but it also means EXIF can never prove anything about a file. We walk through the privacy side of this in EXIF data privacy risks and the location-specific case in how to remove GPS data from photos.
Photo by Wallace Chuck on Pexels.
What Is C2PA?
C2PA is the new layer, and it was built to do the one thing EXIF cannot: prove something. C2PA is the standard from the Coalition for Content Provenance and Authenticity — backed by Adobe, Microsoft, the BBC, camera makers, and others — and it defines a tamper-evident record of where a file came from and what was done to it. Adobe's "Content Credentials" are the best-known consumer implementation of it, and the same standard now reaches audio, which we cover in C2PA in your audio files.
Where EXIF records the camera, C2PA records the history. A C2PA manifest holds a set of assertions — statements like "created with Adobe Photoshop" or "Adobe Firefly generative AI was used" — plus a claim that binds those assertions to the exact bytes of the file, plus 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 verify two things: that the record has not been altered, and which certificate authority vouched for the signer. That cryptography is the whole point. A C2PA credential is not just data about the file; it is a claim you can check.
Physically, the manifest lives in a JUMBF box (JPEG Universal Metadata Box Format). In a JPEG it rides in one or more APP11 marker segments — a different segment from EXIF's APP1 — and because a single segment maxes out at 65,535 bytes, a large manifest is split across several contiguous APP11 segments. In a PNG it sits in a caBX chunk; in WebP, its own RIFF chunk. We picked this apart in detail in what C2PA metadata is.
Photo by Nemuel Sereti on Pexels.
C2PA vs EXIF: The Core Differences
Put side by side, the two standards differ on five things that actually matter.
Purpose. EXIF answers "how was this captured" — a technical record for cameras and photo software. C2PA answers "where did this come from and what happened to it" — a provenance record for a world where AI generation is common and platforms want to label it.
Who writes it. EXIF is written by the capture device: your camera, your phone. C2PA is written by editing and generation tools: Photoshop, Lightroom, Firefly, and a growing list of AI generators that we documented in Midjourney, DALL-E, and Stable Diffusion EXIF.
Trust model. This is the big one. EXIF is unsigned, so it proves nothing and can be edited without a trace. C2PA is signed, so it is tamper-evident: change the file or the manifest and the signature breaks, which is detectable. EXIF is a note; C2PA is a notarized note.
Where it lives. Both are separate blocks alongside the pixels, not baked into them — but in different places. EXIF sits in APP1; C2PA sits in APP11 inside a JUMBF box. A file can carry both at once, neither, or either alone.
What removal means. Because both are discrete blocks, a clean rewrite of the file can drop either or both while copying the pixels through untouched. Stripping EXIF removes the APP1 block; stripping C2PA removes the APP11/JUMBF manifest store. Removing one does not touch the other, which is why a tool that only clears EXIF can leave a C2PA credential sitting in the file, and vice versa.
The practical upshot: "I removed the metadata" is ambiguous. Removing EXIF protects your privacy but leaves any AI-provenance credential in place. Removing C2PA drops the AI label but leaves your GPS coordinates in APP1. If you care about both, you need a tool that clears both layers.
Why Does the Difference Matter?
It matters because the two layers create two different problems, and most people only know about one of them.
The EXIF problem is privacy. EXIF can carry the exact coordinates of where a photo was taken, the serial number of your camera, and a precise timestamp. Posting a photo with EXIF intact can publish your home address without you realizing it — the classic risk the EFF has documented for years. For most people, stripping EXIF before sharing is the entire goal.
The C2PA problem is labeling and reach. Platforms increasingly read provenance signals at upload. They look at C2PA manifests and at the related IPTC "Digital Source Type" property to decide whether to attach an "AI info" label or a "made with AI" tag. A credential that says Firefly was used can trigger that label even if a generative tool only touched a small part of the frame. For a creator whose reach depends on not being filed under "synthetic," that embedded credential is a business problem, not a privacy one.
Photo by Malte Luk on Pexels.
Here is the honest-limits part, because it would be easy to oversell removal. Stripping the embedded C2PA manifest removes the in-file record, but it does not reach everything provenance systems can do. Adobe's "durable" Content Credentials also place an invisible watermark in the pixels and store a fingerprint in the cloud, both designed to recover the record even after the embedded manifest is gone — so a determined verifier may still match a stripped file back to its history. Platforms also run their own AI-detection classifiers that never look at metadata at all. And removing C2PA does nothing about audible or visible watermarks in other media. Removing metadata cleans the layer you can see in the file; it is not a cloak of invisibility. We are clear about that so you can decide what is worth doing.
How to Remove EXIF and C2PA
Because both are discrete blocks, removal is the same basic operation for either: rewrite the file and drop the block, leaving the pixels alone. Metadata Cleaner does both layers in one pass. Open the site in any browser, drag in your JPEG, PNG, or WebP, and click Clean. The tool rewrites the file and removes the EXIF APP1 block, the XMP and IPTC blocks, and the C2PA manifest store — for a JPEG that means dropping both the APP1 EXIF segment and the APP11 JUMBF segments; for a PNG it removes the caBX chunk. Nothing uploads to a server; the file is processed in the browser tab and the pixels are copied through unchanged.
To confirm it worked, you can run exiftool -G1 -a on the cleaned file and check that no EXIF group and no C2PA/JUMBF group is reported, or drop the file into a C2PA verifier and confirm no credential is found.
FAQ
Are C2PA and EXIF the same thing?
No. They are separate metadata layers with different jobs. EXIF records capture details (camera, lens, GPS, timestamp) and is unsigned. C2PA records provenance (edit history, AI involvement) as a cryptographically signed manifest. A single image can contain both, and they live in different parts of the file — EXIF in APP1, C2PA in APP11.
Does removing EXIF also remove C2PA?
Not necessarily. They are distinct blocks, so a tool that only clears EXIF can leave a C2PA credential in the file, and a tool that only clears C2PA can leave your GPS coordinates in EXIF. To clear both you need a tool that targets both layers. Metadata Cleaner removes EXIF, XMP, IPTC, and the C2PA manifest in a single pass.
Which one carries my location?
EXIF. GPS coordinates are stored as EXIF tags in the APP1 segment. C2PA manifests record provenance, not capture location, so the privacy-critical location data lives in the EXIF layer.
Which one triggers an AI label on platforms?
C2PA, alongside the IPTC Digital Source Type property. Platforms read these provenance signals at upload to decide whether to attach an AI label. EXIF has no field that platforms use for AI labeling.
Can I tell which layers a file has without special tools?
Partly. File properties in your OS usually surface a slice of EXIF, but they rarely show C2PA. Running exiftool -G1 -a lists both groups, and a C2PA verifier will confirm whether a signed credential is present. If you are unsure, assume a file from a modern editor may carry both.
Try Metadata Cleaner free
If you want both layers gone — the EXIF that exposes where and when, and the C2PA credential that flags how — clean the file before you share it. Try Metadata Cleaner free: it runs in your browser, nothing uploads, and it clears EXIF and C2PA in one pass.