ChatGPT text watermarks and textGrain
OpenAI describes textGrain as a statistical pattern in generated wording. It does not add invisible characters or extra tokens. Availability depends on the product, eligible region, model and rollout; you should not assume every ChatGPT answer carries the same watermark.
AIWatermark.dev focuses on the text you actually paste. It removes supported hidden characters and can normalize non-breaking spaces while keeping visible wording and paragraph breaks. This cleanup is useful for documents, email, CMS editors and code-like identifiers, regardless of where the text came from.
A ChatGPT draft⟦U+2060⟧ with ordinary words⟦U+FEFF⟧.After local cleanupA ChatGPT draft with ordinary words.How to clean copied ChatGPT text
- Paste your ChatGPT response or upload a plain UTF-8 text file.
- Inspect character code points, choose space and line-ending settings, then clean the text.
- Copy or download the result. To recast sentence structure, try an optional rewrite and check its facts.
This page is for plain text. It does not remove a visible watermark or provenance metadata from ChatGPT-generated images, videos or files. For statistical text signals, rewriting is an unverified transformation, not a guarantee that a detector will return a particular result.
Common questions
Does removing zero-width spaces remove textGrain?+
No. OpenAI’s textGrain is a pattern in word choices, not a zero-width-space marker. Removing clipboard artifacts does not establish the status of that signal.
Can I use this ChatGPT text cleaner without an account?+
Yes. Local character cleanup requires no login or key. The optional external rewrite requires your own OpenRouter account and explicit consent to send text.
Can the tool clean ChatGPT images or videos?+
No. AIWatermark.dev handles plain text only. Image and video watermark queries have different needs and different tools.
Source: OpenAI’s text provenance announcement. Reviewed October 11, 2026.