In post-ChatGPT world, em dash — which has carried the rhythm of human thought for long — has become the target of linguistic witch-hunt. Here’s why it’s not an AI quirk, but a human fingerprint
In late 2022, when ChatGPT entered public life, something unexpected happened: work emails became better written, research papers flowed more tightly, and college assignments took fewer hours to complete — and with that, a quiet paranoia began to spread. Professors, editors, office-goers — everyone — started squinting at prose. “Did ChatGPT write this?” became a standard question, if only in their heads.
To sniff out AI-generated text, the sceptics began looking for telltale signs — certain words, tone shifts, punctuation choices. Some indicators were useful; others, frankly, absurd. But perhaps the most unexpected casualty in this linguistic witch-hunt? The em dash.
Yes, the em dash — or as some now call it, the ChatGPT hyphen — became a supposed giveaway of machine-written text. Never mind that writers have used it for centuries. Suddenly, this punctuation mark was being treated like an AI watermark. As if the dash had just been invented by OpenAI.
A marker of linguistic elitism
But the em dash has been around far longer than large language models — and has survived more than one technological upheaval. In Practical Printing: A Handbook of the Art of Typography (1892, fifth edition 1900), John Southward writes that “an ‘em’ is always equal to the depth or body of the type of which the fount is composed. These names were given to the measures because in book fonts the type for the letter ‘m’ was usually exactly as wide as it was deep, and the type for the letter ‘n’ half as wide as it was deep.” Thus, an em dash is an em wide; an en dash, half that length.
For centuries, printers took the dash seriously. But the typewriter changed everything. Keith Houston writes in Shady Characters: The Secret Life of Punctuation that the arrival of Christopher Latham Sholes’s typewriter made a hash of the dashes. The 1868 machine had only one dash: a hyphen-minus, a compromised avatar of a dash. The salad days of nuanced punctuation — of em dashes, en dashes, colons, commash, colash and more — were over by the arrival of the typewriter. Did someone say technology makes life simpler?
Also read: Why films drop, humans tweet, mice guide and clouds hug information to themselves
QWERTY typewriter’s entry in 1878 didn’t help much either. Though later typewriters added marks like the colon and apostrophe, the dashes remained elusive. Writers made do. Isaac Pitman suggested using a spaced hyphen - , or better, two unspaced hyphens -- to represent the em dash. The second method caught on, creating a divide: the printed world still had proper em and en dashes; the typewritten world survived on hacks.
Why do we even need the em dash? Because, as Ben Yagoda once wrote in the New York Times, it's “the mark that — unlike commas, periods, semicolons and all the others — doesn’t seem to be subject to any rules.” Britannica.com agrees: “While the hyphen’s forerunners sought to connect, the dash’s forerunners sought to separate.” The em dash — especially — does this best.
So the literary world kept the dash alive. Writers dashed with style; printers followed suit. Meanwhile, the masses — journalists, typists, office workers — either ignored it or replaced it with double hyphens. Over time, the dash itself became a marker of linguistic elitism.
A human fingerprint, not an AI quirk
Even the rise of computers didn’t immediately restore the dash. The original ASCII standard in 1963 made no room for it. It wasn’t until Unicode arrived in 1991 — succeeding ASCII — that the em and en dashes were officially added to our digital repertoire. But there was a catch: they weren’t on the keyboard. To type them, you had to know arcane shortcuts, do a bit of finger-gymnastics, or copy and paste from somewhere else. Worth it? Most people didn’t think so.
Even those who knew the difference often didn’t bother. The dash remained obscure. Then came Microsoft Word 97, which finally gave the em dash a break. Type two hyphens and Word would autocorrect them to an em dash. It also handled numeric ranges with en dashes. Google Docs went further: it turns hyphens between numbers into en dashes and three hyphens into an em dash. Mobile keyboards followed. On Android and iPhone, you can long-press the hyphen key and choose the dash you want. But not many know this — and even fewer care.
Now, here’s the irony: Large Language Models (LLMs) like ChatGPT are bringing the em dash back. Once you discover it — and learn to use it — your writing takes off. It lets you pause, interrupt, pivot. It sounds modern, clear, confident. And because it’s showing up more often, the critics assume it must be artificial. Stop blaming LLMs just because you’ve only now started noticing the dashes.
To confirm, I asked ChatGPT where it learned em dash usage from. ChatGPT learned em dash usage from us. The answer: “From writers, editors, marketers, authors, poets, professors, and newsletter scribblers. The LLM didn't invent the em dash — it absorbed it by observing how humans used it across genres and decades.” Exactly. The em dash is not an AI quirk — it’s a human fingerprint.
Carriers of the rhythm of thought
The newest trick in em-dash detection is to check whether there are spaces before and after it. If there are, some say, it’s “LLM-written.” Not true. It’s a style thing. The Chicago Manual of Style uses unspaced em dashes. The Associated Press recommends spacing. I asked Gemini to weigh in — and it returned an unspaced em dash. When I pressed it further, Gemini replied that neither it nor ChatGPT is locked into a particular style guide. The dash spacing simply mirrors their training data.
Also read: What Banu Mushtaq’s International Booker win means for Indian literature in translation
There’s no right answer — though I personally prefer the unspaced Chicago style. I’m using spaced dashes in this piece to match the Ben Yagoda quote I cited earlier. Love the dashes? Use them. Here’s a cheat sheet:
Hyphen (-) joins words (e.g., LLM-written).
En dash (–) connects ranges (e.g., US–China trade deal; purely hypothetical).
Em dash (—) — the longest and most elegant — can function like a comma, colon, or parentheses.
Em dash haters aren’t new. Back in 2011, Noreen Malone wrote a takedown of the em dash in Slate. Today’s critics just have a new excuse — LLMs. But here’s the beauty of it: the same dashes that were exiled from popular use by 19th-century technology are now being revived by another piece of technology — LLM-powered apps. Poetic justice? Absolutely. What better proof could there be that LLMs are human too?
If you accuse the em dash of artificiality, you’re not just lazy — you’re turning your back on style itself. Ignore the naysayers. Learn your dashes — and use them freely. They’ve been around far longer than any language model. From Emily Dickinson to Herman Melville, Virginia Woolf to David Foster Wallace, em dashes have carried the rhythm of thought — even inserted the unsaid. They reflect how we think: digressions, hesitations, clarifications.
