For ChatGPT images

Remove metadata from ChatGPT images.

Check a generated image for Content Credentials, then download a copy without them.
Your files are processed on your device and never uploaded.

Drop your images here, or choose files

JPG, PNG, WebP, HEIC, AVIF · Max 20MB each · Up to 30 images

Files never leave your device. No account. No uploads.

1Choose images2Remove AI & hidden metadataClean3Download & share

Share the picture, not your location or camera details.
Prepare cleaner files for social posts, product listings, and everyday sharing.

Free to useNo sign-upBrowser-only processing

Metadata cleanup doesn't remove visible or pixel-based watermarks, or guarantee that a platform won't apply an AI label. Read the limits

What a ChatGPT image contains

OpenAI says supported images from ChatGPT, Codex, and its API carry two signals: C2PA metadata in the file, and a SynthID watermark in the pixels. It also says metadata can be lost through platforms, editing tools, or file conversions.

So a file you saved may already have lost its record. Check it above before assuming anything is there.

What gets removed

Removed: Content Credentials, XMP, IPTC, EXIF, and PNG text blocks.
Not removed: the SynthID watermark. OpenAI describes it as a signal that can remain when metadata is gone.
Not removed: anything visible in the picture.

A cleaned ChatGPT image is still an AI image, and detection tools may still identify it.

Good reasons to clean a copy

You want a smaller file with no extra blocks for a website or app.
You're combining images into one set and want consistent files.
You'd rather not share account or tool history in a public file.

Please stay open about AI images. Follow the rules wherever you publish, and never present a generated image as a real photo to mislead anyone. Read what metadata AI tools add for how other generators compare.

How to inspect a ChatGPT image

Save the image from ChatGPT, and avoid taking a screenshot, which creates a new file with no record.
Open the Metadata checker tab above and select the file.
Look for Content Credentials, and note any XMP data.
To read the full record and validate its signature, use Content Credentials Verify.

OpenAI has also announced a checker for its C2PA and SynthID signals. If a check finds nothing, that doesn't prove much, because metadata can be removed by platforms, editing tools, and conversions.

What the metadata does and doesn't prove

A valid record shows that an OpenAI tool signed the file, and it can show the steps recorded after that. It doesn't show that the image is accurate, or who wrote the prompt. A missing record proves nothing, because even an ordinary re-save can drop it. Treat the metadata as one piece of information about the file's history.

Sharing AI images honestly

Plain wording works best and takes very little space. Try "Image created with ChatGPT," or "AI-generated illustration" for a blog post. For client work, say in writing which images are AI-generated, because rights, stock rules, and contract terms can depend on it. If a platform asks whether content is AI-generated, answer truthfully, whatever the file's metadata says.

News, health, finance, and anything showing real people deserve extra care. In those areas, a realistic generated image can mislead readers even when no one intends it to.

Sources

A few things to know

Questions, answered.

Do ChatGPT images always contain C2PA metadata?

OpenAI says supported images include it, but the record can be lost along the way. The checker shows whether your copy still has it.

Can this remove the SynthID watermark?

No. SynthID is part of the pixels, and this tool only leaves file metadata behind.

Will the image stop being detected as AI?

It may not. A watermark, a visual detector, or the image itself can still show that AI was used.

Does the file store my prompt?

Check your own file, since this can vary. The checker lists any Content Credentials, XMP, and PNG text that it finds.