Invisible Unicode character list

The complete reference for what this site detects. Every row is a code point or range the engine classifies, with the action it takes and the reasoning behind it. This table is generated from the same data file the checker runs on, so it cannot drift out of date relative to the tool.

How to read the table

• Risk: high means the character can change how text is displayed or parsed, bidirectional controls, tag characters and control characters. These are removed and flagged in red.
• Risk: medium means the character is invisible and disruptive but does not alter display order. Removed or replaced, flagged in yellow.
• Kept or removed depending on context means the engine examines the neighbouring characters before deciding. The same code point is legitimate in one place and junk in another.
• Replaced with a normal space is used for every space-like character. They are never deleted, because deleting a space merges the words on either side of it.

57 individually named code points plus 7 ranges. This table is generated from the same data file the checker itself uses, so it always matches what the tool does.

Test it on your own text

Paste anything into the checker below to see which of these characters it contains and what the cleaner does with each one.

Paste your text/code

Clean Watermark Clear All

Cleaned Text/Code

Copy

More: what was found, x-ray view and options

Try an example

Normalise unusual spaces Straighten smart quotes Em dashes (—) to hyphens

X-ray view

Your text with every hidden character exposed as a labelled chip. Hover a chip for its Unicode name.

What is deliberately not in this list

Three categories are out of scope, and each for a reason:

• Homoglyphs. Cyrillic а, Greek ο and their Latin lookalikes are visible characters and legitimate letters in their own scripts. Deciding that one is an impostor requires knowing the intended script of the surrounding text, which is a linter's job. See invisible characters in code for how they are misused.
• Mathematical alphanumeric symbols. The fake bold and italic letters common on social platforms are visible, so they fall outside a cleaner that removes invisible characters. They do break search and screen readers.
• Combining marks. Stacked diacritics can be used to distort layout, but they are the normal mechanism for writing accented text in many languages. Removing them would break far more than it fixed.

Notes on individual entries

U+00A0, the no-break space

The single most common find in real-world text. It looks exactly like a space, and it is produced by autocorrect in word processors, by   in HTML, and by almost every PDF extractor. Most whitespace-trimming code does not trim it, so a value can look trimmed and still carry it.

U+FEFF, the byte order mark

Metadata at the start of a file, and an invisible character everywhere else. A BOM at the top of a JSON file makes parsing fail with a message that points at the wrong thing entirely; at the top of a shell script it breaks the shebang line.

U+200D, the zero-width joiner

Structural in two very different worlds. It builds emoji sequences, a family emoji is three people and two joiners, and it is required in Devanagari and Arabic-script typing. The engine keeps it when both neighbours are emoji, and when both neighbours belong to a script that uses it.

U+E0000 to U+E007F, the tag block

An invisible copy of ASCII. Any sentence can be encoded in it and rendered as absolutely nothing, which is why this block is the standard vehicle for hiding instructions inside text that looks harmless. The one legitimate use is regional flag emoji, where a tag sequence closed by U+E007F encodes the region code, and those are kept.

U+FE0F, variation selector 16

The character that turns a monochrome symbol into a colour emoji, and part of every keycap sequence. Removing it unconditionally is the most common way a naive cleaner mangles emoji. The engine keeps it after any emoji-capable character and before a keycap combining mark, and removes it only where there is nothing it could apply to.

Questions about the list

How many invisible Unicode characters are there in total?

There is no single official count, because "invisible" is a description of rendering rather than a Unicode property. This list covers the code points that appear in real text and cause real failures: the zero-width and format characters, the bidirectional controls, the space characters, the tag block, the variation selectors and the control characters.

Why are ordinary space characters in a list of invisible characters?

Because they are invisible in the sense that matters: a no-break space and an ordinary space are indistinguishable on screen and are different characters. That difference is what breaks lookups, numeric parsing and form validation.

Is this list the same as the one the tool uses?

It is literally the same data. The table on this page and the detection engine are generated from one file at build time, so a change to the engine's behaviour appears here automatically and the two cannot disagree.

How do I detect these characters in my own code?

Iterate over code points rather than UTF-16 units, and test each one against the ranges in this table. The important detail is the ordering: check for characters that must be kept in context before removing anything, or you will destroy emoji and non-Latin text on the way through.

Related pages

• Hidden characters in text - where each class comes from and what it breaks.
• Invisible characters in code - the practical consequences of the bidirectional controls above.
• Clean hidden characters out of text - the cleaner with its scope stated plainly.