What's inside a ChatGPT image file, and how to remove it
Images you create in ChatGPT or through OpenAI's image API ship with C2PA Content Credentials. The manifest records that an OpenAI model generated the image. Drop a file below to read the metadata and strip it without re-encoding the pixels.
Drop PNG, JPEG, WebP, HEIC, AVIF, MP4, or MOV files here
Runs 100% locally in your browser · Original quality & colors kept
0
0
0
0
Queue (0)
Select a file to view or clean
Choose a photo or video from your queue — or click Try a Sample Image — to see what hidden tags are inside and clean them cleanly.
What OpenAI puts in the file
C2PA manifest
A signed record stored in a caBX chunk in PNG files, or the equivalent block in JPEG and WebP. It names OpenAI as the generator and logs the creation step.
AI source type
Inside the manifest, the digital source type is set to trainedAlgorithmicMedia, the standard value for fully AI-generated media.
No visible watermark
OpenAI doesn't draw a visible mark on ChatGPT images, so the metadata is what other apps and checkers read from the file itself.
How to check and clean your file
- Download the image from ChatGPT and drop it into the cleaner above. It scans the file on your device right away without uploading it.
- Open View raw metadata details to see the C2PA manifest and any other metadata the file held.
- Click Clean Image (or Clean All for a batch), then check the After Clean tab to confirm no AI tags remain.
- Download the cleaned PNG, JPEG, or WebP. The pixels match the original exactly.
What cleaning can't change
- OpenAI has said it keeps internal tools that recognize its own images. Those don't rely on file metadata.
- Social platforms can still apply AI labels based on their own detection or your disclosure.
- Many platforms ask you to disclose realistic AI-generated images. Removing metadata doesn't change that.