How to use this tool
Use this tool when you have identified an unwanted hidden separator in text and want a reviewable result. Paste a working copy, inspect the detected code points, and start with the conservative rules. The tool applies explicit character transformations in your browser; it does not decide which characters are meaningful in a language or document. Keep the original until you have tested the output in the application where you intend to use it.
Scroll the table sideways to see every column.
| Character | Conservative | Custom: Remove invisible |
|---|---|---|
| U+200B / U+FEFF / U+00AD | Remove | Remove |
| U+200C / U+200D / U+2060 | Preserve | Remove |
| Selected special spaces | Preserve | Separate Normalize spaces option |
| Control and directional marks | Report; preserve | Report; preserve |
Choose the right starting point
A useful removal task has an expected result. An accidental U+200B inside a product code might need deletion, while a deliberate boundary between words might need replacement with an ordinary space. This tool deletes supported invisible characters when the selected rule allows it. It does not infer missing word boundaries, rebuild formatting, or choose punctuation from the meaning of a sentence.
If you do not yet know what is wrong, use the inspection report before pressing Clean text. Look for the code point and the surrounding text. The count tells you how many supported findings exist, not how many mistakes the document contains. A reported joiner can be intentional. Do not aim for zero findings when that would remove characters the text needs.
What conservative cleanup changes
The default Conservative mode removes U+200B ZERO WIDTH SPACE, U+FEFF BOM / ZERO WIDTH NO-BREAK SPACE, and U+00AD SOFT HYPHEN. It preserves U+200C, U+200D, and U+2060, along with special spaces and dashes. Ordinary spaces, tabs, and line breaks remain. Directional and control marks are reported for manual review but are not deleted by the cleaner.
The word conservative describes the scope of these rules. It does not promise that every removed character is unnecessary. A soft hyphen can be intentional in a publication, and a zero-width space can express an intended break. Review the diff against the source and the destination's requirements. For a strict U+200B-only edit, use a targeted editor after inspecting the other findings, because this mode removes a small group of characters.
Work through the provided example
Try an example replaces the current input with a short demonstration. Save your own sample elsewhere before pressing it. The demonstration places an actual U+200B between Zero and width. After conservative cleaning, that character is removed, so the visible beginning becomes Zerowidth. The tool does not insert a regular space between those words. This makes the consequence of deletion easy to see.
For an identifier, the same behavior may be exactly right: AB[U+200B]12 becomes AB12. In that notation, the bracketed label represents a single real character. Typing the bracketed label literally will not reproduce the test. Use the example button to see a real finding, then inspect your own sample. The diff labels invisible removals with code points so you can review them.
Use custom rules deliberately
Changing a checkbox or dash option selects Custom mode. With Remove invisible enabled, custom cleanup can also delete joiners that conservative mode preserves. That may alter joined emoji or connected-script text. If your input contains either, compare a short representative sample before applying the rule to a larger passage. Retain the original until the language and display have been reviewed.
Normalize spaces is a separate rule that replaces supported special spaces with ordinary spaces. It does not collapse repeated ordinary spaces, trim leading gaps, or join lines. Dash handling is also independent. Keep dashes for a narrow hidden-character task unless there is a separate reason to alter punctuation. The visible rule note describes the active behavior, so check it before generating the result.
Read the report and diff correctly
The character counter uses Unicode code points. The finding locations are displayed starting at one, so a symbol represented by several code points can affect later positions even when it looks like one character. The list displays the first thirty findings; its length can be smaller than the total counter. For many findings, work in smaller sections that preserve enough context to judge each edit.
Show spaces and Show tabs only change the preview. Their dots and arrows are visual aids, not part of the copied result. The diff is also a preview and may be shortened for long input. The result textarea contains the generated output. Changing input or rules clears an old result, so run Clean text again after any adjustment before using Copy result.
Validate the change in its destination
Copy the result and repeat the operation that motivated the cleanup. For a reference number, rerun the lookup. For a paragraph, inspect word boundaries and wrapping in the editor or publishing preview. For multilingual text, check the actual words and emoji rather than approving a smaller character count. A successful button click means an output was generated, not that your document is ready.
If the character returns after saving, compare direct typing, copying from the original source, and copying the cleaned result. That experiment helps identify the stage that changes the string. Repeated removal without checking the path may conceal a source editor setting or an import transformation. Keep a short before-and-after example to make the issue reproducible.
What this remover does not promise
This utility does not enumerate all Unicode characters, sanitize source code, verify external links, or remove all invisible content from every possible string. It does not perform general Unicode normalization. A remaining comparison problem may involve ordinary whitespace, different accents, letter case, or visually similar characters outside the scanner's supported set.
Hidden-character cleanup also does not establish authorship or guarantee an AI detector result. Use it to make an explained text edit while preserving attribution and any required disclosure. The application processes pasted text locally, while loading the website still involves ordinary page requests described in the privacy policy. For support, share a short invented sample and the selected rule rather than a confidential document.