What does the ChatGPT text cleaner remove?
It fixes the formatting marks that make pasted AI text look machine-made, without touching your wording. Every option can be switched on or off.
| What it finds | Looks like | What the cleaner does |
|---|---|---|
| Em dashes and en dashes | — and – | Replaces them with a comma or a spaced hyphen. Number ranges like 20–30 become 20-30 |
| Invisible characters | Nothing at all | Removes zero-width spaces, joiners, soft hyphens, byte order marks and direction marks |
| Look-alike spaces | A normal space | Swaps no-break, narrow no-break, thin and hair spaces for a regular space |
| Curly quotes and ellipses | “ ” ‘ ’ … | Converts them to straight quotes and three periods |
| Markdown | bold, ## headings, code, links | Strips the symbols and keeps the text. Bullets become simple hyphens |
| Extra whitespace | Double spaces, stacked blank lines | Collapses them to single spaces and single blank lines |
The cleaner runs in your browser, so the text never leaves your device.
Why does ChatGPT text contain hidden characters?
AI chat apps format answers for the screen. When you copy that text, some of the formatting can come along: non-breaking or narrow no-break spaces that look exactly like normal spaces, zero-width characters, and markdown symbols that only render as bold or headings inside the chat window.
These characters are not a reliable watermark. They're easy to remove, they also appear in text copied from ordinary web pages and documents, and AI detectors such as GPTZero and Turnitin analyze the wording itself rather than invisible characters. They can still cause problems: odd line breaks, words that won't search or spell-check properly, and spacing that looks wrong in a document. Cleaning them out is good hygiene either way.
Does removing em dashes make AI text undetectable?
No. Em dashes have become a running joke about ChatGPT's style, but detectors don't count punctuation. They measure how predictable your word choices are and how evenly your sentences are built, signals often called perplexity and burstiness. A paragraph with every em dash removed still scores the same if the sentences underneath are uniform and generic.
Cleaning is the first step. To change the patterns detectors actually read:
- Check the rhythm with the free sentence length checker. Long runs of similar-length sentences are a classic AI tell.
- Find stock vocabulary with the AI word checker, words like "delve", "crucial" and "seamless".
- Rewrite the structure. The Humanizer rewrites sentence rhythm, structure and word choice in one click while keeping your meaning, and it's designed to pass detectors like GPTZero, Turnitin and ZeroGPT.
Read more in our guide to why ChatGPT uses so many em dashes.
Should you replace em dashes with commas or hyphens?
Use a comma when the dash sets off a short aside or joins two parts of one thought ("Time management isn't just a skill, it's a habit"). Use a spaced hyphen if you want to keep a visible break, for example in headings or lists. Sometimes the best fix is a period: split the sentence in two. After cleaning, give the text a quick read and adjust any spot where a comma feels wrong.
Human writers use em dashes too, so there's nothing wrong with keeping a few in work you wrote yourself. The problem with AI drafts is volume: three or four dashes in one paragraph is a strong signal of an unedited ChatGPT answer.
How do you remove em dashes in Google Docs or Word?
You can do it by hand with find and replace:
- Google Docs: press Ctrl+H (Cmd+Shift+H on Mac), type — in the "Find" box, type a comma and a space in "Replace with", then click Replace all.
- Microsoft Word: press Ctrl+H, enter ^+ in "Find what" (Word's code for an em dash) and ", " in "Replace with", then Replace All.
Hidden characters are harder to spot by hand because they're invisible. Pasting your text into this cleaner and copying it back is the fastest way to catch them all at once.