AI Text Cleaner — Remove Em Dashes & Invisible Characters

How to use

Paste text into the input pane — the detection report immediately lists every AI-trace character found, with its Unicode code point and count: em dashes, curly quotes, zero-width spaces, narrow no-break spaces, soft hyphens, and more. Choose how to handle em dashes (replace with a comma, a spaced hyphen, or keep them), then toggle the cleanups you want: straighten smart quotes, strip invisible characters, normalize unusual spaces, convert the ellipsis character to three dots, or strip markdown bold/heading markers. The cleaned result updates live — copy it or download as .txt. Everything runs locally in your browser.

Why use this tool

Text from ChatGPT, Claude, and Gemini carries typographic fingerprints: the famous em dash, curly quotes, the ellipsis character, and sometimes genuinely invisible Unicode like zero-width spaces or narrow no-break spaces. These cause real problems beyond style — invisible characters break JSON parsing, corrupt code snippets, produce phantom diffs in git, and break structured data like JSON-LD. Curly quotes break shell commands and config files. This tool shows you exactly which hidden characters your text contains (most editors render them as nothing) and removes them in one pass, without the paid 'humanizer' upsell — it is a transparent character-level cleanup, not a detection-evasion service.

FAQ

How do I remove em dashes from ChatGPT text?

Paste the text, pick a replacement style — comma + space (reads most naturally), a spaced hyphen, or plain space — and copy the result. The tool collapses surrounding spaces so you never get double spaces where the dash was.

What invisible characters does it detect?

Zero-width space (U+200B), zero-width non-joiner (U+200C), zero-width joiner (U+200D), word joiner (U+2060), byte order mark (U+FEFF), soft hyphen (U+00AD), directional marks (U+200E/200F), and invisible math operators (U+2061–U+2064). Each is listed with its code point and count before you clean.

Are invisible characters proof that text is AI-generated?

No. Word processors, web pages, and copy-paste chains also introduce them, and research suggests they are not deliberate watermarks. Treat them as a hygiene issue, not evidence — this tool removes them but makes no claim about detecting AI authorship.

Why do invisible characters break my code or JSON?

Parsers see them as real bytes. A zero-width space inside a JSON key makes it a different key; a BOM at the start of a file can break scripts; a narrow no-break space in a shell command causes 'command not found'. If code pasted from a chatbot mysteriously fails, hidden characters are a common culprit.

What unusual spaces get normalized?

No-break space (U+00A0), narrow no-break space (U+202F), en/em spaces (U+2002/2003), thin and hair spaces (U+2009/200A), figure space, and similar — all replaced with a regular space. The CJK ideographic space (U+3000) is deliberately left alone because it is meaningful in Chinese and Japanese text.

Does it change Chinese or Japanese punctuation?

No. Full-width CJK punctuation (，。！？「」) is correct in East Asian text and is never touched. Only Western typographic characters and invisible Unicode are cleaned.

Can it make AI text undetectable?

No, and it does not try to. AI detectors rely on statistical patterns in wording, not on em dashes or hidden characters. This is a formatting cleanup tool — for correctness and style, not detection evasion.

Is my text uploaded anywhere?

No. Detection and cleaning are pure client-side JavaScript — your text never leaves the browser. No sign-up, no upload, no logging.