ChatGPT Image Metadata: Check and Remove It

What OpenAI adds to ChatGPT images

OpenAI's Help Center (2026) says that images made with ChatGPT, Codex or the OpenAI API include two provenance signals:

The history is short. OpenAI began adding Content Credentials to DALL·E 3 images in 2024 and later to its image generator in ChatGPT. On May 19, 2026, it announced that it was adding SynthID, starting with images from ChatGPT, Codex and the API (OpenAI, May 2026).

OpenAI also says coverage may vary by product, model, export path, file type and creation date. API outputs are covered, except for models that OpenAI plans to retire soon, including gpt-image-1 and gpt-image-1.5.

What the Content Credentials say

OpenAI's developer guide (2026) shows what its check reads from the C2PA data of an image:

• the issuer, for example "OpenAI OpCo, LLC";
• the model, for example "gpt-image";
• the time when the image was generated;
• whether the signature is valid and trusted;
• an action that says an AI model generated the image.

OpenAI states that provenance signals do not include details about the user, the organization or the prompt.

What else is in the file

OpenAI does not publish a list of other metadata fields in these files. The safest way to know is to look at your own file. The AI metadata checker in WipePic shows the metadata that it finds, grouped by kind.

The checker reads the file in your browser. WipePic does not upload your photos.

How to check a ChatGPT image

1. Use the original file that you downloaded. A screenshot or a converted copy has lost its Content Credentials.
2. Open the AI metadata checker and add the file.
3. To check the signals of OpenAI itself, use OpenAI's Verify page at openai.com/verify. It looks for a trusted OpenAI C2PA manifest and a SynthID watermark. OpenAI says it does not store the files that you check, unless the law requires it.

OpenAI also says a result of "not detected" does not prove that a person made the image. The file may be older than the signals, or the metadata may be gone, or the watermark may be damaged.

How to remove the metadata from a ChatGPT image

1. Open Remove Content Credentials.
2. Add your images. You can add up to 300 files, with up to 50 MB each.
3. Select Remove Metadata. WipePic removes the Content Credentials and the other hidden data. Your photos do not go to a server.
4. Download the cleaned files. Several files come as one ZIP.

WipePic does not re-encode the image, so the pixels do not change. It also re-reads the cleaned file and saves it only if no removable metadata is left. For all AI labels at once, use Remove AI Metadata.

What removal does not do

• It does not remove SynthID. The watermark is in the pixels. Read Does Removing Metadata Remove SynthID? for the details.
• It does not change what the image shows. WipePic cannot remove visible content, faces or pixel watermarks.
• It does not undo your own posts. If you shared the original file before, copies of it may still hold the Content Credentials.
• It does not stop platform labels that rely on other signals. Meta, for example, also labels content when a person says it was made with AI (Meta, 2024).

Sources

OpenAI Help Center, Provenance signals in OpenAI-generated content (2026); OpenAI, Advancing content provenance (May 2026); OpenAI API, Content provenance (2026); Meta, Our approach to labeling AI-generated content (2024).