Invisible Character Remover
Text can carry characters that draw nothing at all. They sit inside a word, survive a copy and paste, and break exact matches in a search. This tool finds them, names them, and shows you where each one sits before it takes them out.
What an invisible character is
Every character in a text file has a code point, a number that says which character it is. A letter A is U+0041. An em dash is U+2014. Some code points stand for characters that draw nothing: they take up no width, they borrow the width of a space, or they change how the text around them is read while staying invisible themselves.
They are not broken text. A document with a zero width space in the middle of it looks normal, reads normally, and behaves normally in every program that understands Unicode. That is what makes them annoying. You cannot see them, so you cannot tell that they are there, but they change what you get when you compare two strings or paste into a search box.
Most of them come from somewhere specific. A web page separates keywords with a zero width space so that search engines see a different phrase. A word processor marks an optional hyphen with a soft hyphen. A translation tool leaves directional marks behind. A database dumps a byte order mark at the start of a field. Once one of those characters is in your clipboard, it stays there until this tool takes it out.
The characters this tool looks for
The tool checks every character in your text against a list of known invisible code points, then groups the matches by code point so you get one row per character with a count.
Format characters that draw nothing
These are the characters most people meet without knowing. The zero width space (U+200B) is the classic one, used by search engines to split phrases. The zero width non-joiner (U+200C) and the zero width joiner (U+200D) control how letters join in scripts such as Persian, and the joiner also holds emoji sequences together. The word joiner (U+2060) keeps two words on the same line. The zero width no-break space (U+FEFF), also known as a byte order mark, sneaks in from older systems and from exports.
Then there are the marks that change direction. The left-to-right mark (U+200E) and the right-to-left mark (U+200F) are single characters. The block from U+202A to U+202E holds the embedding and override controls, and U+2066 to U+2069 holds the isolate controls. They are normal in a document written in Arabic, Hebrew or Persian, and out of place in an English sentence. They are a common reason a phrase copied from somewhere else will not match a search.
Three more round out the list. The soft hyphen (U+00AD) appears only when a word breaks across a line, so it usually survives long after the document was pasted somewhere else. The Mongolian vowel separator (U+180E) is a leftover from a character that used to be a space. The Hangul filler (U+3164) and the braille pattern blank (U+2800) are the two characters people use to build blank looking names, and they are the reason the invisible name tool exists.
Spaces that are not quite spaces
A second group looks like a space but behaves differently. The no-break space (U+00A0) and the narrow no-break space (U+202F) stop a line from breaking at that point, so a page laid out in a design tool can hand you a phrase that will not wrap on a phone. The figure space (U+2007), the thin space (U+2009) and the hair space (U+200A) are typographic spaces that carry a fixed width. The ideographic space (U+3000) is the full width space used in Chinese and Japanese.
The tool can convert all of these into a normal space with one switch. That is usually the right move for text going into a form, a search box, a spreadsheet cell, or a code file. Leave them alone when the spacing is deliberate, for example in a table set with figure spaces.
How to use it
Paste your text into the box at the top. The counts under the box update as you type, so you can tell straight away whether there is anything to clean. Every character that was found gets a row with its name, its code point, and how many times it appears.
The checkbox next to a row decides what happens to that character. Leave it ticked and the character is removed, untick it and the character stays exactly where it is. Two switches sit under the list: one keeps the zero width joiner when it holds an emoji sequence together, and one turns the unusual spaces into normal spaces.
The panel below the list shows your text with every hidden character highlighted, so you can see where each one sits. Hover a highlight to read its name and code point. Then copy the cleaned text from the button at the bottom.
A worked example
Here is a sentence with a zero width space in the middle of the first word, a no-break space instead of the space after the comma, and a soft hyphen at the end of the last word.
The tool reports three characters: the zero width space once, the no-break space once, and the soft hyphen once. With the default settings all three are removed. The result reads as plain text, and a search for the exact phrase now matches.
Before: Th e<U+200B> quick brown fox<U+00A0> jumped<U+00AD> After: The quick brown fox jumped
Keeping a character when it matters
Take the same text and untick the no-break space row. The cleaned text keeps that space exactly as it was, and everything else is still cleaned. The same trick works in reverse: if you know a soft hyphen belongs to a word that should break across lines, untick it and leave the hyphen in.
What this tool does not do
Removing hidden characters does not change how an AI detector judges a piece of writing. Detectors look at word choice, sentence structure and phrasing. A character that draws nothing is not part of that judgement.
It is also worth being clear about the narrow no-break space (U+202F). Some people noticed it in output from certain OpenAI models and asked whether it was a hidden marker. At the time of writing, OpenAI said in its community forum that the narrow no-break spaces seen in some o3 and o4-mini output were not a watermark. Removing the character does not make the text more human or less detectable, and it does not change the score an AI detector gives.
The tool runs entirely in your browser. Nothing is uploaded and nothing is stored. That matters most when you are checking text you are not allowed to send anywhere.
Tips
Look at the highlighted view before you copy. The position of a hidden character tells you where it came from. A zero width space in the middle of every other word often points at a search engine snippet or a keyword variant, not at a mistake you made.
When you find a directional mark such as a right-to-left override, check the text around it. These characters reverse the way a line reads, so a line with one of them can look scrambled while still working fine in a script that expects it.
Run the tool a second time on the result if you want to be sure. If you are pasting into a database, a spreadsheet, or a code file, convert the spaces as well, because fixed width spaces break column alignment and can change how a string is compared.
Troubleshooting
If the tool reports nothing but the text still behaves strangely, the culprit is usually a character outside the checked list, most often a variation selector or a combining accent that belongs to a letter. Copy the text into an editor that shows code points and look at the numbers.
If removing the zero width joiner breaks an emoji into pieces, tick the switch that keeps emoji joiners together. The tool looks for a joiner with an emoji on each side, including the rainbow flag and other sequences that use a variation selector.
If the cleaned text still shows a strange gap, the gap came from a fixed width space that you left untouched. Turning the space conversion on handles that. If your paste came from a rich text editor, empty paragraphs can arrive as zero width characters rather than real line breaks, so remove the characters first and then check the line breaks.
Explore more text tools on Tolz.me.