C2PA, EXIF, XMP: What Your AI Photos Are Quietly Telling People

Every photo file carries more than pixels. Tucked inside the file, invisible on screen, is a layer of metadata that can reveal your camera model, your GPS location, the exact second the photo was taken — or, if the image was made by AI, the model that generated it and when.

Most people never see it. Platforms, forensics tools and nosy strangers do.

The three layers hiding in your files

EXIF is the oldest and most familiar one. Cameras and phones write it automatically: device model, lens, exposure settings, and — if location services were on — GPS coordinates accurate to a few meters.

XMP/IPTC is the layer editing software adds: author name, copyright notice, keywords, sometimes the software version used to edit the file. Photoshop, Lightroom and most AI generators write to this layer.

C2PA (Content Credentials) is the newest and the one causing the most confusion in 2026. It's a cryptographically signed manifest that records how an image was produced — camera capture, software edit, or AI generation — and which tool did it. It's what makes Instagram, TikTok and Google add a "Made with AI" badge to a photo, and it survives a simple screenshot or resave in some cases.

Why AI-generated images are the most exposed

A photo from your phone typically carries EXIF and not much else. A photo from ChatGPT, Gemini, Midjourney or another AI generator usually carries all three layers at once — camera-style EXIF placeholders, XMP authorship data, and a C2PA manifest naming the exact model and generation timestamp. That combination is precisely what automated detectors look for, and it's why AI images get flagged even when they look completely convincing.

What this actually means for you

Platform labeling. Content Credentials are the main reason your AI images get an automatic "AI-generated" tag on social platforms, even after you've edited them.
Client and marketplace trust. Some print shops, stock platforms and freelance clients reject files with visible AI provenance markers.
Privacy. If you're re-sharing a personal photo, EXIF can leak your location down to the building.
Context stripped, not restored. Removing metadata doesn't change the pixels — it only removes the paper trail around them.

How to check what's hiding in your file

Drag a photo into any EXIF viewer and you'll usually see it in seconds — camera fields, software tags, and (for AI images) a C2PA or "jumbf" block near the end of the metadata list.

How to remove it

You don't need Photoshop or command-line tools. Azaiscale's free metadata cleaner strips EXIF, XMP, IPTC and C2PA/AI-provenance data in your browser — no upload, nothing stored, works on JPEG, PNG, WebP, HEIC and TIFF. Drop in a batch of photos and download them clean in seconds.

One caveat worth knowing upfront: metadata removal doesn't touch invisible pixel watermarks like Google's SynthID. We cover that distinction in our SynthID vs C2PA breakdown.

Skip the cleanup step entirely.

Try the free metadata cleaner

Skip the cleanup step entirely

If you're generating images regularly, cleaning metadata after the fact is a chore that doesn't scale. Photos generated with Azaiscale come out already clean — no EXIF, no AI-provenance markers — with a consistent persona across every batch, so you never have to think about it again.

Frequently asked questions

Does removing metadata reduce image quality?

No. For JPEG, PNG and WebP, only the metadata segments are removed — the image data itself is untouched.

Is it illegal to remove C2PA metadata?

No. Removing metadata from your own images is legal in virtually every jurisdiction. What matters is what you claim about the image afterward, not whether metadata is present.

Will resaving or screenshotting an image remove C2PA on its own?

Sometimes, but not reliably — some platforms and export paths preserve or re-embed it. A dedicated cleaner is the only consistent way to guarantee it's gone.