ScrubPic

What metadata is in ChatGPT images?

Images made with ChatGPT carry two signs that they were made with AI: a signed Content Credentials (C2PA) label in the file’s hidden data, and an invisible SynthID watermark in the picture itself. Removing the label doesn’t remove the watermark.

1. Content Credentials (C2PA)

OpenAI adds a signed Content Credentials label to images from ChatGPT and its API. In ScrubPic, it shows up as lines like these:

LineWhat it shows
TypeAI-generated
Made withThe app or service, such as ChatGPT
Signed byOpenAI
Signed onThe date and time the label was signed

2. SynthID watermark

OpenAI says images from ChatGPT, Codex and its API also include SynthID, an invisible watermark that is part of the image itself. You can’t see it, and it isn’t stored in the hidden data.

How to check an image

Add it to the Content Credentials checker. The top line shows “AI-generated” and the app, and the AI label group lists every line in plain words. Nothing is uploaded.

When the label disappears

  • Screenshots and re-saving in many apps drop the hidden data, but the watermark usually survives.
  • Some apps strip hidden data when you send or upload an image.
  • ScrubPic removes the Content Credentials without re-saving the picture, so quality stays the same. It doesn’t remove the SynthID watermark.

Some platforms read these signals to label AI images, and many ask you to label realistic AI content yourself. Removing the hidden data doesn’t change those rules.

Check your own files. See what they reveal and remove it, free and without uploading.

Open ScrubPic

Related guides