Whitespace Character Reference

Published and reviewed: August 27, 2026

Whitespace is a category, not a single character

In everyday use the word space covers dozens of distinct code points. The Unicode Character Database marks a set of them with the White_Space property, and text specifications such as the WHATWG Infra Standard define a much narrower "ASCII whitespace" set of five characters. A tool, a programming language, and a layout engine can each use a different definition, which is why text can still look wrong after you believe every space has been removed.

Two properties matter for cleanup. The first is visible width: does the character draw a gap, and how wide. The second is line-break behavior: does the character allow, forbid, or force a wrap at that point. The line-breaking algorithm is specified in Unicode Standard Annex #14. A non-breaking space has the width of a normal space but forbids a wrap; a zero-width space has no width but permits one. Knowing both properties tells you whether to convert a character to a plain space or delete it.

Horizontal space characters

These characters separate words or align text within a line. They never start a new line.

Tab, line-ending, and separator characters

These characters control vertical structure. Removing or converting them changes how many lines the text has.

Zero-width and invisible format characters

These characters usually draw nothing. They are grouped with whitespace during diagnosis because they make two identical-looking strings behave differently, but several of them carry meaning and should not be deleted blindly.

Because of the last two rows, a careful invisible-character cleanup separates U+200B and a stray U+FEFF, which are usually residue, from ZWNJ and ZWJ, which are usually intentional. The zero-width character remover offers a cautious mode that keeps the joining controls.

How these characters reach your text

• PDF extraction preserves typographic spaces (thin, figure, narrow no-break) and inserts a line ending at every visual row, plus occasional form feeds between pages.
• HTML and the web turn   into U+00A0, and copying rendered text keeps the Unicode character rather than the entity, as covered in the WHATWG named character reference.
• Word processors auto-insert protected spacing around initials, units, and punctuation, and add soft hyphens during hyphenation.
• Spreadsheets export CRLF line endings and can leave a leading BOM in the first cell of a CSV.
• Chat and design tools sometimes emit U+2028 or zero-width characters that survive a copy and paste.
• Input methods for CJK languages can produce full-width U+3000 instead of an ASCII space.

A detection checklist

1. A short phrase overflows the right edge and never wraps: suspect U+00A0 or U+202F.
2. An exact-match search or a database join fails on text that looks identical: suspect U+200B, U+FEFF, or a trailing regular space.
3. Columns line up in one program and collapse in another: suspect U+0009 tabs.
4. Every sentence sits on its own line: suspect hard LF or CRLF from a PDF or wrapped email.
5. A script or parser throws an "unexpected token" or "unterminated string": suspect U+2028 or U+2029.
6. To confirm, paste a small sample into a tool here and read the live audit, or in a console run JSON.stringify(text) and inspect the \uXXXX escapes.

Which tool removes which character

For behavior-focused explanations rather than a character list, see the Unicode whitespace guide and Non Breaking Space. To compare this approach with writing code, read Browser Tool Versus Code Methods.