A label is not always stored in the file

An AI label shown by a website may come from a portable record in the file, an upload or editing history held by that service, a creator disclosure, or the service's own analysis. These are different sources of information. Removing embedded metadata changes the local file; it does not erase records held elsewhere.

The C2PA specification explainer describes Content Credentials as signed provenance attached to an asset. It also distinguishes provenance from a general claim that content is true. A service can make its own labeling decision even when the file has no readable Content Credential.

Diagnose the situation in order

  1. Inspect the original supported file with the cleaner. Note exactly which portable traces were found.
  2. If you choose to clean it, check the output report. It verifies only the selected supported traces.
  3. If a label remains on a platform, check that platform's explanation and your own disclosure or publishing settings. Do not assume the file still contains the same metadata.
  4. If you need to preserve attribution or provenance for a client, newsroom or archive, keep the original file and its credentials.

What cleaning can and cannot promise

Free AI Wiper can remove selected supported portable records from a supported file. It cannot change a platform database, remove a visible mark from pixels or frames, or guarantee a different moderation or labeling decision. A clean file is not proof of human authorship.

For a file-specific process, use the image, PDF or video format page.

If you are publishing in the EU, Article 50's marking and disclosure rules are a separate question from what this cleaner can find in a local file.