Paste any text. The cleaner removes invisible characters, odd spaces and formatting leftovers, then shows you exactly what it took out. It runs in your browser: your text is not uploaded, and there is no word limit.
Paste text and press Clean text.
These are the characters people usually mean when they talk about a watermark in AI text. None of them are visible on screen.
| Character | Code | Why it matters |
|---|---|---|
| Zero-width space, joiner, non-joiner | U+200B, U+200D, U+200C | Invisible. Splits words for search, word counts and spell check. |
| Word joiner, byte order mark | U+2060, U+FEFF | Invisible. Often left behind by copy and paste between apps. |
| Soft hyphen | U+00AD | Invisible until a line breaks, then a stray hyphen appears. |
| Direction marks and overrides | U+200E, U+200F, U+202A to U+202E, U+2066 to U+2069 | Invisible. Can reorder how text is displayed. |
| Tag characters | U+E0000 to U+E007F | Invisible text that can carry hidden instructions. |
| Non-breaking and narrow spaces | U+00A0, U+202F, U+2000 to U+200A | Look like spaces but behave differently in editors and forms. |
| Markdown symbols | **bold**, ## headings, --- | Left over when chat output is pasted into a document. |
There are two different things called an AI watermark, and only one of them can be removed by a cleaner like this.
Hidden characters. Invisible Unicode characters and unusual spaces end up in text copied from chat windows, PDFs, web pages and some writing apps. Whether a given AI tool adds them on purpose is rarely documented, but they are easy to find and easy to remove, and that is what this tool does.
Statistical watermarks. Researchers have shown that a model can be made to prefer certain words in a pattern only a matching detector can see (Kirchenbauer et al., 2023). That kind of watermark lives in the word choice itself. Deleting characters does not touch it; only rewriting the text does.
Three steps, a few seconds.
It does not change your words. Every sentence comes out exactly as it went in, minus the characters you cannot see.
It also does not make AI writing read as human. AI detectors look at how the writing flows: sentence length, word choice, how predictable each sentence is. Hidden characters play little or no part in that. To see how your text scores, use the free AI detector. To change how it reads while keeping what it says, use the AI humanizer.
For the full picture, read how AI detectors work and why writing gets flagged as AI.
Invisible characters break word counts, search, and character limits in application forms.
A zero-width space in a variable name or a CSV value causes errors that are hard to spot.
Non-breaking spaces stop lines from wrapping and leave odd gaps on mobile.
Markdown symbols turn into literal asterisks and hash marks in Word or Google Docs.
Yes. It is free, needs no account and has no word limit.
No. The cleaning runs in your browser. The text never leaves your device.
It removes the hidden characters and formatting leftovers that people usually mean by a watermark. A statistical watermark in the word choice cannot be removed by deleting characters; it needs a rewrite.
Not by itself. Detectors judge the writing, not invisible characters. Check the text with the AI detector, and rewrite it with the AI humanizer if it is flagged.
No. Only invisible characters and the formatting options you select are changed. Every word stays the same.
Only if you want to. They are normal punctuation, so those options are off by default. Turn them on if your document or form needs plain characters.