ChatGPT Image C2PA Remover and Its Limits

OpenAI's help center says images generated with ChatGPT, Codex or the OpenAI API carry two provenance signals: C2PA Content Credentials in the metadata and a SynthID watermark in the image itself. ScrubPhoto can strip the C2PA part, which is the part Meta and TikTok read, but SynthID is out of its reach. A scrubbed ChatGPT image is still an AI image, and tools that look for SynthID can still find it.

Check a file for Instagram, TikTok and Facebook label triggers or scrub it now. Both run in your browser and nothing is uploaded.

What OpenAI puts into a ChatGPT image

OpenAI's help article on provenance signals, updated in October 2026, lists Content Credentials and SynthID watermarks for images. It says coverage can vary by product, model, export path, file type and when the image was made. SynthID is a Google DeepMind technology, and OpenAI now uses it alongside C2PA.

OpenAI explains why it uses both. Content Credentials record details like which tool created a file and when, and OpenAI says they can be lost through everyday actions like a screenshot or a file conversion. SynthID sits in the content and is designed to survive light edits, cropping and screenshots, though heavy changes can make it undetectable. OpenAI also says its provenance signals do not include details about the user, the organization or the prompt.

Which platforms label ChatGPT images

Start with metadata, since that is what the big social apps say they read. Meta detects C2PA and IPTC AI indicators in uploads to Instagram, Facebook and Threads and labels those posts. Instagram's label is AI info, which it has also shown as AI content and originally named Made with AI. TikTok's May 2024 announcement said content with Content Credentials is labeled automatically, and OpenAI says ChatGPT images carry them. Pinterest, LinkedIn, YouTube and X explain their own rules in their help centers.

Detection does not stop at metadata. OpenAI runs a checker at openai.com/verify that looks for SynthID or a trusted C2PA manifest associated with OpenAI. Google DeepMind's SynthID page says its detector can check content made with AI from Google or partners, and it names OpenAI among them.

ChatGPT image metadata remover: what it changes

Add the image to the tool and it removes the C2PA manifest, plus XMP, EXIF and any PNG text chunks. It does this without decoding the image, so the pixels in the output match the download exactly. That also means the SynthID watermark comes through untouched. ScrubPhoto neither detects nor removes SynthID, and no metadata tool can. See can you remove SynthID for the reason, and the C2PA remover page for the metadata side.

The most useful step here is often the first one. The checker shows whether your file still holds C2PA, which app signed it, and how each platform will likely treat it, before you change anything.

Why remove C2PA from a ChatGPT image at all

Removing the metadata changes what the file declares, not what the image is. It does not turn an AI image into a photograph, and posting it as one would mislead people. What it gives you is control over the file: you can say it is AI in your own words, in your caption or with the platform's own disclosure option, instead of leaving it to an automatic label. Meta applies its label when the person posting discloses AI, so that route stays open.

Some disclosure is mandatory. Meta's rules cover photorealistic video and realistic audio made or altered with digital tools, and Meta says it may penalize people who skip that disclosure. A clean file does not change those rules, and a platform may still label the post through other checks.

Questions about ChatGPT images and C2PA

Does removing C2PA from a ChatGPT image remove the AI label?

It removes the metadata signal that Meta and TikTok read. The SynthID watermark stays in the pixels, a self-disclosure stays, and platforms may use other checks, so the image can still be identified as AI.

Is there a SynthID remover for ChatGPT images?

Not here, and we do not claim otherwise. The watermark is part of the picture, and ScrubPhoto never changes pixels, so it can neither find nor strip it.

Does ChatGPT image metadata include my prompt or account?

OpenAI says its provenance signals do not include details about the user, the organization or the prompt. Run your file through the checker to see every field it actually carries.

How can I check a ChatGPT image for C2PA?

Add it to the checker on this site, which reads the file in your browser and shows the C2PA signer. OpenAI's own checker at openai.com/verify also looks for its watermark and Content Credentials.

Sources

Official pages this page was checked against in October 2026: