ChatGPT · OpenAI · SynthID
ChatGPT’s mark is not one trail. The file, the sound, then the words.
A ChatGPT picture already carries two marks: a signed C2PA note in the file, and SynthID in the pixels. Supported audio carries SynthID as well. The words are the unfinished part. OpenAI signed the same EU code as Claude. They have not published a text method.
Reviewed by Can Balkaya, CTO · Published
or type below, or drop a file
Inspected in your browser. A plan is required to finish.
Choose the surface, then the tool
| Surface | Gifi action | What can be verified |
|---|---|---|
| Ordinary words | Rewrite the prose | The words changed; no private detector certificate |
| C2PA, EXIF, XMP, properties | Strip named metadata | Action list plus an independent reinspection |
| Pixels, audio, or video | Not supported | No removal claim |
The file already carries two marks
Images from ChatGPT, Codex or the API take C2PA and SynthID with them. openai.com/verify reads those signals and nothing else. It will not see Claude. It will not see a rewrite.
C2PA lives in metadata. We can strip the local note. Soft binding can still re-link the file. SynthID in the pixels is out of scope.
The words are a different job
When a text mark lands, it lives in which ordinary word was chosen. Light edits usually keep that trail. New wording does not. We cannot see OpenAI’s key.
What this cannot promise
- A rewrite does not prove the text was never AI-assisted.
- No one outside the lab can certify their detector. The key is private.
- Pixel watermarks, audio and video are out of scope.
Questions this usually raises
- Does every ChatGPT reply have a watermark?
- The pictures and the voice from supported tools, yes. The words are promised. They are not documented as live on every message.
- Is openai.com/verify the same as this?
- No. They look for their marks. We change the words, or we strip a file.
- Will rewritten ChatGPT text fail their detector?
- We cannot say. The key is private.
Primary sources
ChatGPT
Statistical signal? Rewrite the prose.
A ChatGPT mark, if it is in the words, needs different words. Hidden characters and C2PA need the file cleaner.