# ChatGPT still does not watermark text

Originally published at https://gifi.ai/blog/chatgpt-does-not-watermark-text

OpenAI marks ChatGPT images and supported audio. It has not shipped a text watermark. openai.com/verify reads file signals, not paragraphs.

## The category error

Most “remove ChatGPT watermark” pages are selling a detector score or a style rewrite. ChatGPT does not watermark the words. As of August 2026 OpenAI has not published a text method.

What is live: C2PA and SynthID on supported images, SynthID on supported audio, and a verification tool that reads those file signals. That tool will not tell you whether a paragraph came from ChatGPT.

## What we will not sell

A rewrite as “removing ChatGPT’s text watermark.” There is no published mark to remove. Hidden Unicode, if present, shows up on the home page for free. Pixel SynthID is out of scope.

## The longer note

The evergreen guide is /guides/does-chatgpt-watermark-text. Claude and Gemini are the labs that watermark text today.

## Primary sources

- [OpenAI on provenance signals](https://help.openai.com/en/articles/8912793-provenance-signals-content-credentials-synthid-in-openai-generated-content)
- [OpenAI on expanding provenance](https://openai.com/index/advancing-content-provenance/)