Glossary

Metadata stripping: how provenance disappears

Metadata stripping is the removal of a file’s embedded records, EXIF, XMP, IPTC, PNG text chunks and C2PA manifests, either deliberately by a privacy tool or incidentally by re-encoding in an editor, converter, messaging app or social platform, leaving the pixels but none of the provenance.

Controlled example PNG file that carries this record: No record in this copy

without-records.pngPNG file · controlled example

  1. Signature + IHDR
  2. iTXt XML:com.adobe.xmp
  3. tEXt parameters
  4. tEXt prompt / workflow
  5. tEXt invokeai_metadata
  6. eXIf
  7. caBX C2PA manifest
  8. IDAT pixels
  9. IEND

No record in this copy

Screenshots and re-exports often drop records.

A clean copy is not proof of a human author, so the report says what is missing and asks for the original.

Check an image →
Records like these are often removed when a picture passes through Metadata stripping, so the report says what is missing instead of guessing. The picture is a code-drawn controlled example with no records; download it and check it yourself.

What strips metadata

Social platforms re-encode uploads: Instagram, Facebook, Threads, X, Reddit and Pinterest serve copies without records. Messaging apps compress photos unless sent as documents or files. Browser converters and most editors write fresh files. Screenshots create new images with nothing. Privacy tools strip on purpose. OpenAI, Black Forest Labs and Meta all state in their documentation that platforms and edits remove their marks.

Hashes change too: a stripped file has a new SHA-256 even though it looks identical.

How isGenAI handles stripped files

The report says “no evidence found” and explains that absence proves nothing. The perceptual fingerprint may still match a known original, which the report lists as a matching copy. The image metadata compare tool shows exactly which records an original has that a copy lacks, and the image copy versus export guide walks through an example. The metadata remover is isGenAI’s own deliberate stripper, which lists every chunk it removed.

Avoiding accidental stripping

Send files as documents, share zips, keep originals and avoid re-saving before checking. If you must convert, convert a copy.

Stripping is not always bad. Removing GPS coordinates before posting protects privacy, and a creator may reasonably keep a prompt private. The point is to know which copy you hold: strip the copy you share, and keep and check the original.

How to use this term when checking a file

  1. Upload the original file to the AI image detector or the tool named on the page.
  2. Find the finding that uses this term under “How we checked” and read the source field it names.
  3. Weigh it by kind: a signed record, a declared field, a saved prompt and a matching copy are different strengths of evidence.
  4. Remember that absence of a record is not evidence of a human author.

Compare an original with a copy

Frequently asked questions

Which platforms strip metadata?

Most do: Instagram, Facebook, Threads, X, Reddit, Pinterest and compressed messaging sends. LinkedIn and TikTok display information from C2PA; TikTok says it attaches credentials to downloads.

Does stripping change the pixels?

Not by itself, but the re-encode that usually causes it does. The hash changes either way.

Can stripped evidence be recovered?

Not from the file. A perceptual match to a known original, or a vendor’s cloud copy of a credential, are the only routes.

How do I strip metadata on purpose?

Use the metadata remover; it rebuilds a clean PNG in your browser and lists the removed chunks.

Sources

Definitions follow the published standards linked above; the isGenAI behaviour described is the deployed checker as of October 2026.