Free · Private · No sign-up

AI Dash Remover — clean the dashes out of machine-written text

ChatGPT, Claude and Gemini love the em dash — and type “--” when they aim for plain text. This tool removes both, plus the double spaces they leave behind, in one click.

The em dash has become the most cited tell of machine-written prose. It is not that the character is wrong — the em dash is a legitimate mark that good writers use — it is that models reach for it at a rate no human editor would, and readers have learned to notice. If a draft is going out under your name, the quickest way to make it sound like you wrote it is to take the dash out. This page is built for exactly that job: paste the reply, press one button, copy the clean version.

The subtlety is that AI answers contain two different kinds of dash. When the model is writing prose it emits the real em dash (U+2014), and sometimes the horizontal bar (U+2015) that its output pipeline substitutes. When it is trying to look plain-text-friendly it writes “--” or “---” instead, which are two or three ordinary hyphens with no em dash in sight — which is why a plain find-and-replace for the em dash character often reports zero results on text that clearly looks AI-written. This tool matches both forms by default, treating typed dashes as em dashes, so a mixed reply comes back clean in a single pass.

It also cleans what a dash leaves behind. A dash removed from “the plan — which nobody liked — was shelved” can leave a double space that is just as much of a giveaway as the dash itself, and typed “--” dashes often arrive surrounded by stray spaces. The remover collapses those runs and trims trailing blanks, so the paragraphs come back tidy. What it deliberately does not do is rewrite your sentences: every comma, colon and capital letter stays exactly where the model put it, which keeps the text faithful to the draft you reviewed.

Why the em dash became the AI tell

Language models are trained on text where the em dash appears constantly, because it is the most flexible mark in English: it can replace a comma, a colon, a semicolon or parentheses, and it never produces a grammatical error. A model that wants a fluid, confident-sounding sentence reaches for the safest tool, and the safest tool is the dash. Human editors, by contrast, are constrained by style guides — AP, Chicago and most newsroom desks restrict the dash precisely because it is dramatic — so a human sentence is far more likely to contain a comma. That gap between model preference and house style is what readers are picking up on, and removing the character closes most of it in one click.

The “--” problem: dashes a search cannot find

A lot of AI output substitutes two or three hyphens for the em dash, and text pasted through a plain-text pipeline can convert a real em dash into “---” on the way. Those characters are ordinary hyphens, so a find-and-replace for — finds nothing, and text that is visibly dash-heavy survives the cleanup. This remover treats “---” and “--” as em dashes when it runs: the counter shows how many it caught, and each one is replaced with a single space so the sentence stays readable. Always check that “---” is matched before “--” — a naive replace order turns three hyphens into a leftover hyphen. This tool gets that order right.

Can I just ask the model to stop using em dashes?

You can, and you should — an instruction such as “never use em dashes; use commas, colons or parentheses instead” improves the first draft noticeably. Treat it as a first pass rather than a guarantee: models reintroduce the character within a paragraph or two, especially in longer answers, and a system prompt does not survive a copy-paste through another tool. Instructions also cannot fix the replies that already exist, and they do nothing about your own archive of drafts. A deterministic remover is the only step that gives you a checkable answer — zero em dashes — because it looks at characters rather than intentions.

What else to check in a machine-written draft

The em dash is the loudest tell, not the only one. Typographic quotes and apostrophes are a second: pasted model text often carries curly quotes into systems that want straight ASCII, which is visible in code and email templates. Sentence rhythm is a third — nearly every sentence landing at the same length reads as generated. Non-breaking spaces and invisible soft hyphens are a fourth: they survive a copy-paste, then break a search index or a database import. The dash is the one you can fix mechanically in a second, so it is the right place to start; the rest is editing. The remove-all-dashes tool covers the invisible characters when you need a fully ASCII result.

Why AI drafts look the way they do

The evidence behind this tool, from the blog that measures it.

Stop ChatGPT Using Em Dashes

The prompt wording that survives, and why the mechanical pass is the step that cannot fail.

Why ChatGPT Overuses Dashes

Training data, token economics and preference tuning — the mechanical reason.

Is My Writing AI?

Count the patterns before and after the cleanup, with the checker.

More dash cleanups

Each tool does one job. Start with the one that matches the text in front of you.

Em Dash Tool

The plain version: remove, replace or count, one page.

Remove All Dashes

Em dashes, en dashes, hyphens and the invisible variants, for an ASCII-only result.

En Dash Tool

Fix the ranges and scores a model wrote with an en dash before they break an import.

Hyphen Tool

Open compound words and count the four hyphen characters.

Em Dash Copy

The opposite job — copy the em dash, the horizontal bar and its relatives in one click.

All Dash Symbols

Every dash-like character on one page, with code points and HTML entities.

Questions

What makes this an “AI” dash remover rather than a plain one?

It matches both forms of the character a model produces: the real em dash (U+2014) and its relatives, plus the typed “--” and “---” stand-ins that an ordinary find-and-replace cannot see. It also collapses the double spaces those removals leave behind, which is the second half of the giveaway.

Does it work on Claude, Gemini and Copilot output too?

Yes. Any assistant that writes prose produces the same two problems, so the same click fixes them. Copy the reply into the box above, press Remove AI dashes, and paste the clean text back wherever it belongs.

Will it remove dashes that are legitimately part of my text?

That is the trade-off of a focused tool — it takes out every em dash it finds, including one a human editor wrote on purpose. If a document needs its punctuation judged rather than stripped, use the Replace Em Dash page instead, which shows what each substitute does to a sentence.

Does removing em dashes make text undetectable as AI-written?

No, and no tool can promise that. What it does is remove the single most recognisable surface signal, which is the part a reader notices. Everything else — sentence rhythm, vocabulary, structure — is still yours to edit.

Is there a prompt that removes em dashes reliably?

Try: “Rewrite this without any em dashes or double hyphens. Use commas, colons or parentheses where the sentence needs a break.” It helps, but models drift back to the dash within a few paragraphs, so run the result through this tool as the final step before you publish.

Is the text I paste stored anywhere?

No. The whole cleanup happens in JavaScript in your browser tab — nothing is uploaded, logged or kept. That matters more than usual here, because AI drafts are often unpublished work.

Ready to clean your text?

No sign-up, no upload, no word limit. Paste, pick, and copy the result.

Remove dashes now