What is textGrain — and what can a character cleaner do?

A pattern in wording

OpenAI announced textGrain on 5 October 2026. It adjusts the statistical pattern of word choices during generation. It does not insert invisible spaces, hidden characters or special punctuation.

The announced rollout covers eligible ChatGPT and Codex text outputs in the EU over the following weeks. API customers can opt in for supported models; the API setting is off by default.

What this tool actually checks

This tool checks a published list of Unicode characters. Finding one is not evidence of a textGrain watermark. Finding none does not prove that the text was written by a human.

The cleaner changes only the removable character types you select. It does not rewrite wording, access an OpenAI detector or issue a watermark-removal certificate.

Detection has limits

OpenAI currently restricts text-detector access to approved organizations. Editing can weaken detection, but that is not a guarantee that any specific text is watermark-free. A general AI-text classifier is also not a textGrain detector.