Find invisible characters

Paste text and every hidden or unusual character shows up, with what it is and where it is.

Your text, with hidden characters shown

  

What this tool shows

Some characters cannot be seen: a zero-width space takes up no room at all, a no-break space looks exactly like a space, and a letter from the Cyrillic alphabet can look exactly like a Latin one. They are still part of the text, so a search, a comparison or a program treats them as different characters.

Paste text in the box and the tool shows it again with each of those characters drawn as a small label: NBSP for a no-break space, ZWSP for a zero-width space, TAB for a tab. Below, a table lists every kind it found with its Unicode code point and name, how many there are, what it is, and the line and column of the first few, so you can find them in your own file.

Pasted text has one blind spot: a text box turns Windows line endings (CR LF) into plain LF. To check a file’s line endings, or whether it starts with a byte order mark, open the file itself with “Open file” or drop it on the box; it is read exactly as saved, and the cleaned version can be downloaded.

Not everything it lists is a problem. Tabs are normal in code, and an emoji like 👨‍👩‍👧 is three emoji held together by two zero-width joiners. Those are marked as normal and left alone. The labels are colored by risk: blue for harmless, amber for characters that cause bugs, red for the ones used to deceive people.

Where they come from

  • Web pages and word processors: the no-break space (U+00A0), which keeps two words on one line, comes along when you copy.
  • A Mac keyboard: Option+Space types a no-break space. On keyboards where | needs Option, one often ends up after a pipe in a terminal command, which then fails with “command not found”.
  • Dates and times: newer versions of the Unicode locale data (CLDR 42, 2022) write English times with a narrow no-break space (U+202F) before AM or PM. Browsers, Node.js and macOS that use it produce that character, and code that reads the time back expecting a space breaks.
  • Websites and chat apps: zero-width spaces (U+200B) mark places where a long word may wrap. They are also used on purpose to mark copies of a text so that a leak can be traced.
  • PDFs and e-books: soft hyphens (U+00AD) that only show when a word breaks at the end of a line.
  • Spreadsheets: Excel’s “CSV UTF-8” format starts the file with a byte order mark (U+FEFF), which then sticks to the first column name.

The ones you will meet most

LabelCode pointWhat it doesWhat to do
NBSPU+00A0A space that does not break a lineReplace with a space
NNBSPU+202FNarrow no-break space, before AM/PMReplace with a space before parsing
ZWSPU+200BAn invisible break pointRemove
ZWJU+200DJoins emoji into oneKeep inside emoji, remove elsewhere
BOMU+FEFFByte order mark at the start of a fileRemove, or save as UTF-8 without BOM
SHYU+00ADSoft hyphenRemove
CRU+000DCarriage return (Windows line endings with LF)Convert line endings to LF
RLOU+202EDisplays the following text backwardsRemove, and ask why it was there

Why they cause trouble

  • Code that will not compile: a no-break space or a zero-width space inside a line of code is not whitespace to a compiler, so the error points at something that looks fine.
  • Searches and comparisons that fail: “10:30 PM” with a narrow no-break space is not equal to “10:30 PM” with a space, and a lookup by name finds nothing.
  • Passwords and codes that do not match: a code copied from an e-mail with a zero-width character in it is rejected even though every visible character is right.
  • Data files: a byte order mark puts an invisible character in front of the first column name of a CSV file, so a program looking for the column “id” does not find it, and a soft hyphen splits a word in a database.

Hidden text and look-alike letters

Direction controls such as the right-to-left override (U+202E) change the order characters are displayed in. In 2021 researchers showed that they can make source code look different from what the compiler reads (the “Trojan Source” vulnerability, CVE-2021-42574), and they have long been used to disguise file names. Ordinary text rarely needs them, so the tool marks them red.

Unicode tag characters (U+E0000 to U+E007F) are invisible copies of the ASCII letters. In 2024 security researchers showed that some AI chatbots read instructions written with them, which a person reviewing the text cannot see (a technique called ASCII smuggling). The tool decodes them and shows the hidden text. Their one legitimate use, the flags of England, Scotland and Wales, is recognized and left alone.

A word that mixes alphabets, like “pаypal” with a Cyrillic а, is how fake web addresses and spam get past filters. The tool flags any word with Latin letters and Cyrillic or Greek letters in it, and says which Latin letter each one imitates. It cannot flag a word written entirely in another alphabet.

Questions

Two lines look identical but my program says they are different. How do I find out why?

Paste each one here and compare the tables: one of them usually has a no-break space or a zero-width character. To see exactly where two texts differ, use the diff checker.

Is it safe to remove everything the tool finds?

For ordinary text in English or Spanish, yes. The clean-up keeps emoji whole and leaves tabs alone. Be careful with Persian, Arabic or Indian scripts, where the zero-width non-joiner and direction marks can be needed; untick those options there.

Why is a joiner in my emoji listed as normal?

Many emoji are built from several: 👨‍👩‍👧 is a man, a woman and a girl joined by two zero-width joiners, and ❤️ is a heart followed by a selector that asks for the colored style. Removing those characters would break the emoji, so the tool keeps them.

Does my text leave my computer?

No. The tool runs in your browser: the text is checked and cleaned on your device and is not sent anywhere.

More text tools

See all tools