What are invisible characters?

Invisible characters are Unicode code points that render with no visible glyph — zero-width spaces (U+200B), zero-width joiners, soft hyphens, directional marks, variation selectors and special space variants. They appear in text through copy-paste, word processors, AI tools, or deliberate obfuscation.

How do I remove invisible characters from text?

Paste your text into the detector above. It highlights every hidden character with its Unicode code point, and the Remove button strips them all and copies the cleaned text to your clipboard. Everything runs in your browser — the text is never uploaded.

Are zero-width characters dangerous?

They can be. Zero-width characters are used for fingerprinting and tracking leaked documents, Unicode tag characters can smuggle hidden instructions into AI prompts, and directional overrides (U+202E) enable the Trojan Source attack in source code. In normal prose they usually just break search, deduplication and character counts.

Invisible Character Detector

Paste any text to reveal hidden Unicode: zero-width spaces, soft hyphens, directional marks, no-break spaces and other characters you can't see but computers can. Remove them all in one click.

🔒 Your text never leaves your device — all analysis happens in your browser.

Ignore joiners inside emoji sequences (recommended)0/10000 characters

Every invisible character this tool detects

Code point(s)	Name	Severity	Why it appears
U+200B	Zero Width Space	suspicious	Invisible word-break hint; abused for tracking and steganography.
U+200C	Zero Width Non-Joiner	contextual	Legitimate in Persian and Indic scripts; suspicious in Latin prose.
U+200D	Zero Width Joiner	contextual	Required inside emoji sequences (👨‍👩‍👧) and Indic scripts.
U+2060	Word Joiner	suspicious	Invisible no-break glue; rare in normal prose.
U+FEFF	Zero Width No-Break Space (BOM)	suspicious	Only legitimate as a byte-order mark at file start; mid-text is a red flag.
U+2061–U+2064	Invisible Math Operator	suspicious	Function application / invisible times / separator / plus — math notation only.
U+180E	Mongolian Vowel Separator	suspicious	Reclassified as a format character; renders invisibly.
U+034F	Combining Grapheme Joiner	suspicious	Rare collation edge cases only.
U+00A0	No-Break Space	contextual	Very common (HTML  , typography); also a mild LLM tell.
U+202F	Narrow No-Break Space	contextual	The headline ChatGPT artifact; also legitimate French typography.
U+2000–U+2002	En Quad / Em Quad / En Space	contextual	Typographic spacing.
U+2003	Em Space	contextual	Typographic spacing; statistically associated with LLM output.
U+2004–U+200A	Typographic Space (Three-Per-Em … Hair Space)	contextual	Fine typographic spacing widths.
U+205F	Medium Mathematical Space	suspicious	Math layout only.
U+3000	Ideographic Space	contextual	Standard full-width space in CJK text; suspect in Latin text.
U+1680	Ogham Space Mark	contextual	Ogham script only.
U+200E–U+200F	Left-to-Right / Right-to-Left Mark	contextual	Legitimate bidirectional text layout.
U+061C	Arabic Letter Mark	contextual	Arabic text layout.
U+202A–U+202C	Directional Embedding / Pop	contextual	Legacy bidirectional embedding controls.
U+202D–U+202E	Directional Override (LRO/RLO)	suspicious	Trojan Source attack vector (CVE-2021-42574) — almost always malicious in code.
U+2066–U+2069	Directional Isolate	contextual	Modern bidirectional isolation controls.
U+00AD	Soft Hyphen	suspicious	Invisible unless the line wraps at it.
U+2028	Line Separator	suspicious	Breaks JavaScript string literals; invisible in most editors.
U+2029	Paragraph Separator	suspicious	Same risks as Line Separator.
U+FE00–U+FE0F	Variation Selector	contextual	VS15/VS16 pick text vs emoji presentation; also a known steganography vector.
U+E0100–U+E01EF	Variation Selector Supplement	contextual	CJK ideographic variation; steganography vector.
U+E0000–U+E007F	Unicode Tag Character	suspicious	Legitimate only in emoji subdivision flags (🏴󠁧󠁢󠁳󠁣󠁴󠁿); arbitrary tags hide invisible ASCII payloads used in prompt-injection attacks.
U+115F–U+1160	Hangul Choseong/Jungseong Filler	suspicious	Korean jamo composition; abused for blank text.
U+3164	Hangul Filler	suspicious	Commonly abused for invisible usernames.
U+FFA0	Halfwidth Hangul Filler	suspicious	Same abuse pattern as Hangul Filler.
U+2800	Braille Pattern Blank	contextual	Legitimate Braille; abused as invisible text elsewhere.
U+FFFC	Object Replacement Character	suspicious	Placeholder for embedded objects; abused for blank nicknames.
U+FFF9–U+FFFB	Interlinear Annotation Control	suspicious	Ruby-text plumbing; should not appear in plain text.
U+206A–U+206F	Deprecated Format Control	suspicious	Deprecated shaping/digit-shape controls.
U+0000	Null	suspicious	Should never appear in text.
U+000B–U+000C	Vertical Tab / Form Feed	suspicious	Legacy control characters.

Checking whether text came from an AI tool? Try the AI watermark checker, or identify any visible symbol with the symbol analyzer.