AI Text Cleaner · Guide

Remove Em Dashes from AI Text

Last updated: August 28, 2026

The em dash is a common stylistic pattern in AI-generated and heavily edited text — and a genuine problem in plain-text contexts: code, JSON, CSVs and CMS fields that expect a plain hyphen. Here's how to remove them in one click.

Why em dashes are everywhere in AI text

AI models love the em dash — it's a stylistic habit baked into their training data. ChatGPT, Claude and Gemini all produce them liberally, and every copied block carries a few. In a document that's fine. In a codebase, a config file, or a database field, an em dash (U+2014) can be the difference between working and broken. It's one of several AI formatting habits — see the full list →

What ClearPaste does

ClearPaste normalizes em dashes (U+2014) and en dashes (U+2013) to plain hyphens — everywhere in the text, automatically, along with the rest of the AI formatting cleanup. One paste, one click, done.

And yes — removing em dashes removes one of the most common stylistic tells people use to spot AI text. We're not in the business of "making text undetectable," but normalizing typography is a legitimate part of text hygiene. We're honest about what that does and doesn't mean →

How to remove em dashes

1. Paste your text into the cleaner above.
2. Every em dash and en dash becomes a plain hyphen.
3. Copy — your text is now safe for code, JSON, CSVs and CMS fields.

Everything runs in your browser. Your text never leaves your device.

The em dash family: which characters are we talking about?

ClearPaste normalizes the first three to the plain hyphen (U+002D). The minus sign is left alone by default — it's semantically different and mostly shows up in math content, not prose.

Where em dashes actually break things

An em dash looks harmless in a document, then fails silently in the systems you paste into:

• Code — a stray U+2014 inside a string literal or a comment can fail linting, break a minifier, or produce an encoding mismatch in older toolchains.
• JSON and CSVs — parsers expecting ASCII punctuation can reject or mangle fields containing typographic dashes, especially when the file is later processed by a script.
• Search and matching — an em dash is a different character from a hyphen, so "long-form — writing" won't match "long-form - writing" in a database or a CMS search.
• Regex — character classes like [a-z-] don't include U+2014, so patterns that work on clean text silently miss lines with em dashes.
• Markdown renderers — some renderers interpret " — " differently depending on surrounding whitespace, producing inconsistent output.

Here's a before/after of what the tool does:

Before: "The team shipped — quietly, without fanfare — the new API."

After: "The team shipped - quietly, without fanfare - the new API."

Who needs this

• Developers pasting AI-generated code or docs into repos, linters and build pipelines.
• Writers moving drafts between a chat tool, a CMS and an email client, where each hop can add or keep typographic dashes.
• Anyone feeding text into scripts — CSV cleanups, keyword matching, translation pipelines — where punctuation must match exactly.

Em dashes are only one member of the formatting family AI text carries. See the full cleanup →

Frequently asked questions