Remove AI metadata

Remove AI metadata from images — locally, then verify it.

AI-created and AI-edited images can carry machine-readable provenance, generator details and ordinary metadata that travels with the file. MetaTaken shows you what is there, removes supported metadata from a separate copy, and checks the result on your Windows PC.

Local Windows processing No image uploads No account Automatic verification

“AI metadata” is not one single tag.

If an image was created or edited with ChatGPT, DALL·E, Gemini, Midjourney, Adobe Firefly, Stable Diffusion or another AI tool, the useful question is not just which brand made it. The useful question is what the exported file actually contains.

Depending on the file and workflow, that can include C2PA / Content Credentials, JUMBF provenance structures, XMP or IPTC fields, software and generator information, comments, descriptions, timestamps and other metadata. Not every tool writes the same data, and not every exported image carries all of it.

What MetaTaken can inspect and clean

C2PA / Content CredentialsSupported embedded provenance structures are inspected and can be targeted by Full Metadata Clean.
Generator & software informationSee tool, software and creator-related fields that are actually present in the image.
XMP, IPTC & EXIFInspect ordinary metadata families that can also contain editing, source, creator or privacy information.
Other removable metadataMetaTaken uses ExifTool plus format-aware checks and reports what was removed, retained, unsupported or uncertain.

The workflow is deliberately boring: inspect, clean, verify.

  1. Drop the final image you plan to share. MetaTaken inspects the file before changing anything.
  2. Review the findings. Plain-language categories are shown first; raw metadata is available in Advanced view.
  3. Choose Full Metadata Clean. A separate cleaned copy is created. Your original is not overwritten.
  4. Let MetaTaken verify the output. The cleaned copy is inspected again, and decoded pixel content is compared where supported.
Why the verification matters: a metadata tool should not call a file clean merely because a command returned successfully. MetaTaken reports the post-clean state it can actually verify.

Will this remove an “AI content” label?

It can remove supported metadata-based signals that may be used to apply a label, including embedded C2PA / Content Credentials and generator-related metadata. If a platform's label comes from those file-level signals, removing them can stop that metadata from travelling with the cleaned copy.

That is not the same as making an image universally “undetectable”. A platform may use its own classifier, account context, upload history or other signals that are outside the image metadata MetaTaken can inspect.

Read the full explanation: what “remove AI detection” can and cannot mean →

What about SynthID?

SynthID is a pixel-level watermarking system, not conventional metadata. MetaTaken does not attempt to remove it and does not treat “metadata cleaned” as evidence that SynthID is absent.

SynthIDNot tested / detection unavailableThis is not evidence that SynthID is absent.

Inspect. Clean. Verify.

See what the file is actually carrying.

MetaTaken is a free local Windows utility. It inspects supported image metadata, creates a separate cleaned copy, then inspects the result again instead of assuming the job worked.

Download MetaTaken v0.3.0-rc2 Release notes and checksum on GitHub