The em dash is a long typographic dash, roughly the width of the letter "M" in the current font. It has been a standard part of English punctuation for centuries, used for strong breaks in sentences, parenthetical statements, and dramatic pauses.

But in the age of AI, the em dash has taken on a new identity: it is the signature character of ChatGPT and other large language models. AI-generated text uses em dashes at a rate that far exceeds typical human writing, making it one of the most recognizable tells of machine-generated content.

Our tool detects em dashes not because they are invisible, but because they are one of the most common special characters that AI injects into text, and they can cause real problems when that text is used in code, data files, or technical contexts.

Why AI Models Overuse Em Dashes

The reason is straightforward: AI language models like ChatGPT, Claude, Gemini, and others learn their writing style from their training data. That training data includes millions of professionally edited texts: published books, academic papers, newspaper articles, and magazine features. In professionally edited text, em dashes are standard.

Human writers, by contrast, default to whatever is on their keyboard. A hyphen requires one keystroke. An em dash requires a special key combination or auto-correct. So most casual human writing uses hyphens or double hyphens where an em dash would be "correct."

AI models have no keyboard. Generating an em dash costs them nothing extra. So they produce the typographically "correct" version every time, resulting in text with dramatically more em dashes than a human would naturally write.

Other AI-Favored Characters

Em dashes are not the only character AI models overuse. The full pattern includes:

Common Uses (Legitimate)

The em dash is perfectly valid punctuation with several standard uses:

• Strong break in a sentence. Creates a stronger pause than a comma or semicolon. Example: The results were clear—the experiment was a success.
• Parenthetical statements. Sets off a phrase mid-sentence, like parentheses but with more emphasis. Example: The team—all five members—agreed to the plan.
• Interrupted dialogue in fiction. Shows a speaker being cut off abruptly. Example: "I was just about to—" He never finished the sentence.
• Replacing colons or semicolons. Introduces an explanation with more dramatic weight. Example: There was only one problem—nobody showed up.
• Attribution in quotes. Placed before the author name at the end of a block quote.

How to Type

Em Dash vs En Dash vs Hyphen

This is one of the most common points of confusion in typography. Here is the definitive comparison:

Quick Rules

• If you are joining two words, use a hyphen: self-aware, well-being, twenty-one.
• If you are showing a range or connection, use an en dash: pages 1–10, Monday–Friday, the London–Paris train.
• If you are creating a pause, interruption, or aside in a sentence, use an em dash: He arrived—finally—at midnight.

When Em Dashes Cause Problems

Em dashes in AI-generated content can cause real technical issues:

• CSV and TSV files. If AI-generated text containing em dashes is placed in a CSV column, some parsers will misinterpret the character or produce encoding errors.
• Command-line arguments. Pasting text with em dashes into a terminal can break commands, since the em dash is not a valid flag prefix (that requires a regular hyphen).
• JSON and XML. While technically valid in JSON strings, some older parsers or systems that expect ASCII may choke on em dashes.
• Search and matching. Searching for a hyphen will not find an em dash, and vice versa. This can cause search results to miss relevant content.
• Plain text contexts. Emails, code comments, commit messages, and other plain-text contexts may not render em dashes correctly in all fonts and terminals.

Detecting Em Dashes

Use our Invisible Character Viewer to paste any text and instantly see all em dashes and other special characters highlighted. The tool shows the Unicode code point and character name on hover.

In code, you can detect em dashes with:

// JavaScripttext.includes('\u2014')
// Replace em dashes with regular hyphenstext.replace(/\u2014/g, '-')
// Replace em dashes with double hyphens (common plain-text convention)text.replace(/\u2014/g, '--')

Style Guides and Em Dashes

Different style guides have different rules about em dash spacing:

• AP Style: Em dash with no spaces on either side: word—word
• Chicago Manual of Style: Em dash with no spaces: word—word
• British style: Often uses en dash with spaces instead: word – word
• AI default: ChatGPT typically uses em dashes with no spaces, following American conventions.

There is no single "correct" style. Pick one and be consistent.