No sign-up · No watermark · Nothing uploaded
Free ChatGPT Image Metadata Remover
ChatGPT signs every image with OpenAI’s Content Credentials. Strip them here before you post, with no upload and no quality loss.
Keeps camera info, date, orientation and motion photos.
Privacy mode: also clears camera, GPS, dates and comments. Orientation and colours stay.
After cleaning, writes generic placeholder camera info (make, model, date). Real camera info already in the file is kept.
- 0 bytes uploaded
- Lossless: pixels untouched
- C2PA · IPTC · XMP · EXIF
What ChatGPT Writes into Your Images
We opened PNG images downloaded from ChatGPT in October and November 2026. Each one carried the same block of Content Credentials (C2PA), about 23 KB, stored in a PNG chunk called caBX. This is what it says:
| Field | What we found |
|---|---|
| Created with | ChatGPT, gpt-image |
| Source type | trainedAlgorithmicMedia, the IPTC code for AI-generated |
| Written by | OpenAI Media Service API |
| Other actions | Converted, and an invisible watermark applied (c2pa.watermarked) |
| Signed by | OpenAI OpCo, LLC, with a time stamp from OpenAI |
| EXIF and XMP | None |
So the metadata of a ChatGPT image is simple: one signed record that says “AI-generated by ChatGPT”. That record is what Instagram, Facebook, Threads, TikTok and LinkedIn look for, and it is the only thing our remover needs to delete.
How to Remove ChatGPT Image Metadata
- Download the image from ChatGPTUse the download button in ChatGPT, so you have the original file rather than a screenshot.
- Drop it into the toolWe show the ChatGPT credentials we find and remove them automatically. You can add a whole batch.
- Save the clean copyDownload it, or save it to Photos on your phone, and post that file.
After cleaning, the tool checks the file again and confirms that no Content Credentials are left. The image data is untouched, so the picture looks exactly the same.
Why Instagram Labels ChatGPT Images
Meta reads Content Credentials and the IPTC source type when you upload. A ChatGPT image carries both signals in its signed record, so it gets the “AI info” label. TikTok labels content with Content Credentials as AI-generated, and LinkedIn shows a small “CR” badge you can click to read the record. Without the credentials, there is nothing in the file for those checks to read.
What Removing It Does Not Change
OpenAI also applies an invisible watermark, and its credentials say so. That watermark is part of the pixels, not the metadata.
No metadata tool can remove the invisible watermark, and we don’t try, because that would mean changing the picture. If a platform checks for OpenAI’s watermark, a cleaned ChatGPT image can still be recognised. The tool shows a “Watermark” tag on files like this so you know what to expect.
Edited ChatGPT Images
If you open a ChatGPT image in Photoshop, Canva or another editor before posting, that app may add its own Content Credentials or keep OpenAI’s. Clean the file after your last edit, and use the AI metadata checker if you want to see what an exported file contains. Images from other generators work the same way; seeGemini images or the full list on the AI metadata remover page.
Frequently Asked Questions
Does ChatGPT add metadata to images?
Yes. Every image we downloaded from ChatGPT carried Content Credentials (C2PA) signed by OpenAI, saying it was created with ChatGPT and is AI-generated. OpenAI has added Content Credentials to its images since the DALL·E 3 days in early 2024.
Why does Instagram label my ChatGPT image as AI?
Instagram and Facebook read Content Credentials when you upload. Because the ChatGPT file says it is AI-generated, Meta adds its “AI info” label. Removing the credentials takes away that signal. See the Instagram guide for more.
Can I remove the ChatGPT watermark?
The ChatGPT images we checked have no visible logo. OpenAI does record that it applied an invisible watermark, which lives in the pixels. No metadata tool, including this one, can remove it, so a platform that checks for it may still recognise the image.
Does this work for DALL·E and Sora images too?
Yes. DALL·E images use the same kind of Content Credentials, and Sora videos carry them in the MP4 file. Drop a Sora clip into the tool above or use the video page.
Will the image look different afterwards?
No. Only the metadata block is removed. The image data is copied byte for byte, so the picture is identical and the file just gets about 23 KB smaller.
Is a screenshot just as good?
A screenshot drops the credentials, but it also lowers the resolution and adds your phone’s own data. Removing the metadata keeps the original image exactly as ChatGPT made it.
Back to the AI label remover.