How ChatGPT / DALL-E marks images
OpenAI embeds C2PA Content Credentials into all images generated by ChatGPT and DALL-E 3. In JPEG files, this is written as an APP11 JUMBF segment with claim generator information. In PNG files, it is written as a caBX chunk.
What gets embedded in the file
The manifest contains digital signatures, the generator name ('OpenAI DALL-E 3'), creation timestamps, and cryptographic hashes verifying content provenance.
Lossless cleaning method
Clean strips the APP11 JUMBF segments (JPEG) or caBX chunks (PNG) directly at the byte level. The actual pixel data and color profile remain 100% bit-identical.
Technical limits and policy
This tool removes file metadata only. Where platforms or legal frameworks mandate AI disclosure, you must disclose AI origin.
How to Clean ChatGPT / DALL-E Images
Add your ChatGPT / DALL-E file
Drag and drop your JPEG, PNG, or WebP image into Clean on your desktop or mobile browser.
Select your preset
Choose "Remove AI markers only" to drop generation manifests while keeping color profiles and EXIF.
Download clean image
Download your cleaned image with bit-identical pixel data and no embedded generator flags.
ChatGPT / DALL-E Metadata FAQs
What metadata does ChatGPT and DALL-E 3 attach to images?
ChatGPT embeds C2PA Content Credentials into JPEG (APP11 JUMBF segment) and PNG (caBX chunk) files. This manifest contains a digital signature declaring OpenAI DALL-E 3 as the generator along with creation timestamps.
Does Clean re-compress ChatGPT images?
No. Clean strips the C2PA JUMBF segment or caBX chunk losslessly at the binary level. Compressed image bytes and color profiles remain 100% bit-identical.
Can social platforms still tell my image came from ChatGPT?
Removing C2PA manifests prevents automated platform bots from reading digital source credentials. However, if platforms analyze visual pixel patterns or prompt text, manual disclosure remains your responsibility.