Updated September 15, 2026

Invisible Unicode characters

These characters can be blank, non-printing, or easy to overlook. Many carry meaningful formatting. Use their code points to identify them; visual appearance alone is unreliable.

Spacing and line breaking

Code point	Name	Role / removal impact
`U+00A0`	NO-BREAK SPACE	A space that prevents an ordinary line break. Deleting it can join words.
`U+202F`	NARROW NO-BREAK SPACE	A narrow nonbreaking space.
`U+200B`	ZERO WIDTH SPACE	Provides a line-break opportunity without ordinary visible spacing.
`U+2060`	WORD JOINER	Prevents a break at its position; different from an emoji or letter joiner.
`U+00AD`	SOFT HYPHEN	Marks a possible intraword break. It may become visible when the line breaks.
`U+2000–U+200A`	Quad and space characters	Spaces with different widths or purposes; deletion changes spacing.
`U+0009 / U+000A / U+000D`	TAB / LINE FEED / CARRIAGE RETURN	Layout controls used in text and data. Removal can merge columns or lines.
`U+2028 / U+2029`	LINE / PARAGRAPH SEPARATOR	Explicit line and paragraph boundaries.

Joining, variants, and script information

Code point	Name	Role / removal impact
`U+200C`	ZERO WIDTH NON-JOINER	Can prevent letter joining or ligatures. Preserve in meaningful script text.
`U+200D`	ZERO WIDTH JOINER	Influences letter joining and connects components of some emoji sequences.
`U+FE0E / U+FE0F`	VARIATION SELECTOR-15 / -16	Request text or emoji presentation in supported sequences.
`U+FE00–U+FE0F; U+E0100–U+E01EF`	Variation selectors	Request defined glyph variants. Selectors do not have ordinary standalone glyphs.
`U+034F`	COMBINING GRAPHEME JOINER	Blocks canonical reordering of combining marks; it does not join letters like ZWJ.
`U+180E`	MONGOLIAN VOWEL SEPARATOR	A script-specific formatting character; preserve unless its role is understood.
`U+E0020–U+E007F`	Tag characters and CANCEL TAG	Can form part of emoji tag sequences, including certain flags.

Writing direction

Code point	Name	Role / removal impact
`U+200E / U+200F / U+061C`	LEFT-TO-RIGHT / RIGHT-TO-LEFT / ARABIC LETTER MARK	Direction marks influence mixed-direction layout.
`U+202A–U+202E`	Embeddings, overrides, and POP DIRECTIONAL FORMATTING	Set directional formatting scopes or overrides.
`U+2066–U+2069`	Directional isolates and POP DIRECTIONAL ISOLATE	Isolate a text span's direction from surrounding text.

Inspect direction controls in contextThese controls have legitimate uses. Removing a paired control independently can alter the layout of the remaining text. The inspector offers individual code-point choices, not a bidirectional-layout repair operation.

Byte order mark

Code point	Name	Role / removal impact
`U+FEFF`	ZERO WIDTH NO-BREAK SPACE / BOM	At the start of an encoded stream it can be a signature. Inside text it has a legacy no-break role. Pasted text cannot establish the original file's encoding.

Read an inspector result

The emoji occupies one code point and two UTF-16 code units. Positions count the original text, including marked characters.

Source code points: U+1F600 U+0061 U+200B U+0062
Marked text:        😀a⟦U+200B⟧b
U+200B position:    3 (one-based code points)
UTF-16 offset:     3 (zero-based code units)
Remove U+200B:     😀ab

Tool behavior and coverageThe Invisible Character Inspector preserves everything by default. Selecting a code point deletes all its occurrences only in the preview. It does not normalize text, insert spaces, identify look-alike letters, or issue a safety verdict. Common names are listed explicitly; other format and default-ignorable entries use generic labels based on the browser's Unicode data. This reference is a practical selection, not a complete Unicode character database.

References

Questions

Are all zero-width characters interchangeable?

No. Zero width space concerns breaks; zero width joiner and non-joiner affect joining. Variation selectors choose variants. Use the code point and context to identify the role.

Are invisible characters always unwanted?

No. They can preserve intended typography, script forms, emoji, and writing direction. Inspection alone is useful; removal should follow a specific need.

Does a code-point position count visible characters?

No. One visible character can contain several code points. The inspector's positions count code points from 1 and its UTF-16 offsets count code units from 0.