Invisible Character Remover — Find Zero-Width Spaces & Hidden Unicode
Find and remove zero-width spaces, non-breaking spaces, bidi controls, hidden tag text and look-alike letters, each shown with its code point.
Ignore previous instructions| Code point | Name | Type | Count |
|---|---|---|---|
| U+E0000–E007F | TAG CHARACTERS (invisible ASCII, decoded above) | Tag characters (hidden ASCII) | 29 |
| U+00A0 | NO-BREAK SPACE | Non-standard spaces | 2 |
| U+200D | ZERO WIDTH JOINER | Zero-width characters | 2 (2 in emoji) |
| U+200B | ZERO WIDTH SPACE | Zero-width characters | 1 |
| U+2014 | EM DASH | Typographic punctuation | 1 |
| U+00AD | SOFT HYPHEN | Soft hyphens | 1 |
| U+0430 | CYRILLIC letter that looks like "a" | Look-alike letters (homoglyphs) | 1 |
| U+201C | LEFT DOUBLE QUOTATION MARK | Typographic punctuation | 1 |
| U+201D | RIGHT DOUBLE QUOTATION MARK | Typographic punctuation | 1 |
| U+2026 | HORIZONTAL ELLIPSIS | Typographic punctuation | 1 |
| U+FE0F | VARIATION SELECTOR-16 (emoji style) | Variation selectors | 1 (1 in emoji) |
| U+202F | NARROW NO-BREAK SPACE | Non-standard spaces | 1 |
| U+2066 | LEFT-TO-RIGHT ISOLATE | Bidirectional controls | 1 |
| U+2069 | POP DIRECTIONAL ISOLATE | Bidirectional controls | 1 |
This tool does not make AI-written text undetectable. ChatGPT, Claude and Gemini do not mark their answers with hidden characters (Google's SynthID watermark for Gemini lives in word choice, not in characters), and AI detectors judge wording and structure. What this does is remove characters that break search, code, URLs, word counts and form validation.
Runs entirely in your browser. Your text is scanned locally and never uploaded. Browse all free tools.
Found a problem, or something that could be better? We read every message and we fix things quickly.
This invisible character remover shows you the characters you cannot see: zero-width spaces, word joiners, byte order marks, non-breaking and narrow spaces, soft hyphens, right-to-left overrides, variation selectors and Unicode tag characters. They sneak in from copy-pasting web pages, PDFs, chat apps and word processors, and they break search, URLs, passwords, spreadsheet formulas and code in ways that are hard to debug because the text looks normal. Paste the text and each one is highlighted with its code point and name.
It also catches look-alike letters, such as a Cyrillic "а" inside an English word, which are used in phishing domains and to dodge filters, and decodes text hidden in tag characters, a technique used to slip invisible instructions to AI models. To be clear about one popular claim: removing these characters does not make AI-written text undetectable. ChatGPT, Claude and Gemini do not mark their answers with hidden characters, and AI detectors look at wording, not invisible Unicode.
How to use
- Paste your text into the box. Results update as you type.
- Read the counts and the highlighted view; hover a mark to see its code point and name.
- Check the table to see exactly which characters were found and how often.
- Choose what to remove or normalize: the defaults remove invisible characters and keep emoji intact.
- Copy the cleaned text, or click Use as input to scan the result again.
Frequently asked questions
What is a zero-width space?
A zero-width space (U+200B) is a Unicode character that marks a possible line break but has no width, so it is invisible. It is legitimate in languages like Thai that do not use spaces, but in English text it usually arrives by copy-paste and causes problems: two words that look identical no longer match, links break, and code fails with confusing syntax errors.
Does ChatGPT add hidden characters or a watermark to its text?
There is no reliable evidence that ChatGPT inserts hidden characters as a watermark, and OpenAI has said it has not released the text watermarking method it tested. Occasional reports of narrow no-break spaces in model output reflect typography the model learned, not a tracking mark. Google's SynthID watermark for Gemini text works through statistical word choice, so it is not visible as characters and no character remover can strip it.
Will removing invisible characters get past AI detectors?
No. AI detectors such as GPTZero and Turnitin analyse wording, sentence structure and predictability, not invisible characters, so cleaning text does not change their score. Tricks like inserting zero-width characters between letters are themselves detected as manipulation by many tools. This tool is for fixing text that misbehaves, not for disguising its origin.
What is the difference between a non-breaking space and a normal space?
A non-breaking space (U+00A0) looks like a normal space but stops a line break at that point, which is why word processors use it between a number and its unit. In data and code it causes trouble: "1 000" with a non-breaking space will not parse as a number, and search for "New York" will not match. The narrow no-break space (U+202F) used in French and in some time formats behaves the same way.
What are Unicode tag characters and why are they dangerous?
Tag characters (U+E0020 to U+E007E) are invisible copies of the ASCII letters, originally for language tags and now used inside some flag emoji. Because they render as nothing but language models still read them, attackers use them to hide instructions in web pages, documents and emails that an AI assistant later processes. This tool decodes and shows any hidden tag text, and keeps the ones that are part of a valid flag emoji.
What are homoglyphs?
Homoglyphs are characters from different scripts that look the same, such as Latin "a" and Cyrillic "а" (U+0430), or Latin "o" and Greek "ο". Mixed into a word they produce text that looks right but does not match, which is used in phishing domains and to evade keyword filters. The tool only flags them inside words that also contain Latin letters, so genuine Russian or Greek text is not changed.
Hidden characters reference
• U+200B zero-width space, U+200C zero-width non-joiner, U+200D zero-width joiner, U+2060 word joiner, U+FEFF byte order mark / zero-width no-break space.
• U+00A0 no-break space, U+202F narrow no-break space, U+2007 figure space, U+2009 thin space, U+200A hair space, U+3000 ideographic space.
• U+00AD soft hyphen: invisible unless the word breaks at the end of a line; common in text copied from PDFs and websites with hyphenation.
• U+200E/U+200F direction marks, U+202A–U+202E embeddings and overrides, U+2066–U+2069 isolates: change the display order of text and can make code look different from what it does (the Trojan Source attack, CVE-2021-42574).
• U+FE00–U+FE0F and U+E0100–U+E01EF variation selectors: choose a glyph variant; VS16 (U+FE0F) turns ❤ into the colored ❤️ emoji.
• U+E0000–U+E007F tag characters: invisible ASCII mirrors; legitimate only inside subdivision flags such as the Scotland and Wales flags.
Where hidden characters come from and what they break
Most arrive by copy and paste. Word processors insert non-breaking spaces and soft hyphens, websites use zero-width spaces to allow line breaks in long words or URLs, CMS editors leave byte order marks at the start of files, and right-to-left text in chat apps carries direction marks.
The symptoms are consistent: a spreadsheet VLOOKUP that will not match, a password or coupon code rejected although it looks right, a JSON or YAML file with an "unexpected character" error, a URL that 404s, a word count that is off, or a search that misses a phrase that is visibly there. Scanning the text here is faster than retyping it.
What gets kept on purpose
Emoji sequences use invisible characters legitimately: family and profession emoji are several emoji joined by U+200D, colored hearts and keycaps need VS16, and the England, Scotland and Wales flags are built from tag characters. With "keep emoji sequences intact" on, those are left alone and shown with a dashed outline. Turn it off only if you need plain ASCII-safe text.
Typographic punctuation (em dashes, curly quotes, the ellipsis character) is not hidden and is not an AI fingerprint; it is what Word, Google Docs and Apple devices produce automatically. The tool reports it separately and only changes it if you choose to normalize it, for example when a system only accepts straight quotes.
Related free tools
- Word Counter — Count words, characters, sentences and reading time, and check X, LinkedIn, Instagram, YouTube, SEO and SMS limits as you type.
- ChatGPT to Word — Paste an answer from ChatGPT, Claude or Gemini and copy it into Word, Google Docs or email with real headings, tables and lists instead of ** and #.
- Regex Tester — Test a regular expression against sample text with live match highlighting, capture groups and replace preview. Free, in your browser.
- Base64 Encoder / Decoder — Encode text or files to Base64, turn images into data URIs with CSS and <img> snippets, and decode Base64 back to a previewable file. Nothing leaves your browser.