Hidden Character Finder

Hidden Character Finder shows what your text really contains. Paste a paragraph and it is redrawn beside the box with every invisible or unusual character turned into a small labelled badge: ZWSP for a zero-width space, NBSP for a no-break space, RLO for a right-to-left override, TAG for an invisible tag letter. Curly quotes and dashes get a soft underline, and a Cyrillic “а” hiding inside an English word gets a red dashed outline.

It is for anyone whose text behaves strangely: a password that won’t match, a CSV header that breaks an import, code that compiles differently from how it reads, a username that looks blank, or a phishing link that looks right but isn’t. A table lists each character with its code point, name, count and purpose.

Then tick the clean-ups you want. Emoji built from several pieces, such as 👨‍👩‍👧 and the Scotland flag, are kept whole by default so cleaning never breaks them. Your text never leaves your device.

1. Paste into “Your text”, drop in a text file, press “Open”, or press “Try a sample” to see a demo full of tricks.
2. Read the labelled view under “What’s hiding”. The status line counts the characters that matter, and coloured chips group them.
3. Use the “Every character found” table to learn what each one is. Click a row to highlight it; click again to jump to the next.
4. Under “Clean it up”, tick the fixes you want: invisible characters, control codes, odd spaces, quotes, dashes, the ellipsis, lookalike letters, or Unicode normalisation.
5. Check the change count, then press “Copy”, “Download”, or “Use as my text” to scan the clean version again.

• Finds zero-width spaces, joiners and non-joiners, word joiners, byte order marks, soft hyphens, Hangul and Braille blank fillers.
• Flags all nine direction controls used in “Trojan Source” attacks and right-to-left file name tricks (LRE, RLE, PDF, LRO, RLO, LRI, RLI, FSI, PDI) plus LRM, RLM and ALM.
• Decodes invisible tag characters into the hidden message they spell, a trick used to smuggle instructions to AI chatbots.
• Spots Cyrillic and Greek letters that look like Latin ones, but only inside words that mix scripts, so real Russian or Greek text is not flagged.
• Lists 16 kinds of unusual space, from no-break and narrow no-break to em, thin, hair and ideographic, plus the line separator, paragraph separator and next-line characters.
• Shows control codes, variation selectors, curly quotes, primes, en and em dashes, minus signs and the ellipsis character, with tabs and line endings on request.
• Clean-up options with a live change count, NFC and NFKC normalisation, and emoji and flags left whole unless you choose otherwise.
• Opens UTF-8, UTF-16 and Windows-1252 text files without hiding a leading byte order mark.

• Hidden characters are not proof that text was written by AI. Phones, Word and web pages add curly quotes, em dashes and no-break spaces every day, and AI text can be perfectly plain.
• For code, JSON and CSV, tick “Smart quotes to straight” and “Dashes to hyphens” as well as the defaults. For prose, leave them off: curly quotes are correct typography.
• NFKC is strong: it turns ² into 2 and fi into fi. Use NFC unless you specifically want those forms flattened.
• A lone RLO or LRI in source code is a red flag. Remove it and read the code again before you run it.