Does ChatGPT watermark its text?
By the Remove AI Watermarking team · Updated
Short answer: OpenAI has said it does not add a hidden watermark to ChatGPT text. What people find in copied answers are invisible Unicode characters, such as zero-width spaces, non-breaking spaces and narrow no-break spaces, that come from formatting rather than tracking.
Key facts
- OpenAI researched text watermarking in 2024 but chose not to ship it in ChatGPT.
- Copied AI text often contains non-breaking spaces (U+00A0) and narrow no-break spaces (U+202F).
- Zero-width characters (U+200B, U+200D, U+FEFF) are invisible but break code, search and word counts.
- AI detectors judge writing style, not hidden characters, so cleaning characters does not change a detector score.
Where the characters come from
Chat apps render answers as formatted text. Copying that formatting carries typographic spaces, joiners and direction marks into your clipboard. Some appear around punctuation and numbers; others sit at line ends.
Why they cause problems
- Code fails to run because a non-breaking space looks like a normal one.
- Find and replace misses words that contain a zero-width character.
- Word and Google Docs show odd line breaks and spacing.
- Forms and CMS editors reject or mangle the text.
How to remove them
Paste the text into our free AI Text Cleaner. Hidden characters are highlighted in red and removed in one click. It runs in your browser, so the text never leaves your device.
Try it free, right in your browser
Clean AI text now