NewOctober 5, 2026: OpenAI announces text watermarking for ChatGPT models. Learn more →
Clean AI WatermarkChatGPT output can carry hidden Unicode characters and markdown leftovers, and OpenAI has committed to marking text. Paste the text below to clean it in one pass.
Paste on the left and hit Remove watermark. Takes about ten seconds.Paste above and hit Remove watermark. Takes about ten seconds.
ChatGPT output has been found to contain hidden Unicode characters such as narrow no-break spaces and soft hyphens, plus markdown leftovers like stray asterisks and headings. OpenAI has signed the EU code of practice on AI-generated content, which requires text marking, and says it will extend its provenance signals to text.
We strip every hidden character and the markdown leftovers, and the word pass runs on every text we clean, whatever its source. If a word-pattern watermark is present, it breaks in the same pass. Sentence structure, meaning, length and tone stay the same.
They come from the model's formatting and from copy-paste between apps. Narrow no-break spaces, soft hyphens and zero-width characters are invisible but easy to find with a script.
Yes. It runs on every text, so a word-pattern watermark breaks whether it comes from ChatGPT, Claude or Gemini. And the same pass removes the characters that mark ChatGPT text today.
No, and we don't claim it. Detectors guess from writing style. We remove hidden characters and break watermark patterns; the style is yours to handle.