Example
Before
Paste the text and inspect it before removing anything.
After
Known controls appear by name, code point, count, and position.
What to check
- Unicode artifacts are not proof of AI authorship.
- Statistical watermarks are not removable characters.
- Direction controls and joiners can be meaningful.
How this cleaner works
Some copied text contains zero-width characters, direction controls, variation selectors, tag characters, special spaces, or Markdown marks. TextClean can report those characters and remove selected categories. Their presence alone does not prove where the text came from.
A statistical text watermark is different. It can be encoded through patterns in token choice rather than a removable hidden character. Unicode cleanup cannot reliably remove that signal, prove human authorship, or guarantee that a detector will classify text differently.
Useful for
- Inspecting copied assistant output for hidden controls
- Removing visible Markdown artifacts
- Understanding the limits of AI watermark claims
Common questions
Can this make AI text undetectable?
No. Detector results are not reliable proof of authorship, and this tool does not promise evasion.
Can it remove hidden Unicode?
Yes, by category. Inspect first because joiners and direction controls can be required for language and emoji.
Does a zero-width character prove AI use?
No. Websites, editors, document exporters, keyboards, and legitimate multilingual text can all contain such characters.