Does an Em Dash Really Mean AI Wrote It?

Hyphens join. En dashes span. Em dashes interrupt. An ellipsis trails off. None of them were born in a language model, and Emily Dickinson did not invent the dash either.

A grammar post making the rounds gets the first part exactly right. The hyphen, the en dash, and the em dash are three different marks. They look related. They are not interchangeable. Most people press the same key for all three. Most editors, as the post says, quietly weep.

The second part of the current argument is newer, and it is wrong. Somewhere in the last two years the em dash acquired a rumor: that the long mark is a tell, a "ChatGPT hyphen," proof the sentence was assembled by a model. If you come to Quoteiac you will find em dashes. You will also find ellipses. We used both long before the rumor, and we will keep using them. The company we keep on that point is older than any chatbot.

We have already written about Dickinson's dash as a choice with weight, not as a glitch, in The Dickinson Dash: What AI Can't Counterfeit. This piece is the companion: what the three marks actually are, where the typewriter flattened them, and why presence of a dash is not proof of a machine.

Three marks, three jobs

The names come from type metal, not from mood. An em is a unit equal to the point size of the type, roughly the width of a capital M in many faces. An en is half that. The marks are named for those widths.

Hyphen (-). The shortest. It joins. Compound adjectives: well-known, print-on-demand, nineteenth-century. It is the only one of the three that lives on a standard keyboard as its own key. That is why it gets asked to do work it cannot do.

En dash (–). Longer than a hyphen, shorter than an em dash. It spans. Pages 10–20. The years 1830–1886. The London–Edinburgh train. It can also join an open compound: the New York–based editor. It is not a hyphen with ambition. It means to or through.

Em dash (—). The longest of the three common marks. It interrupts, sets off, or amplifies. It can stand in for a comma, a colon, or a parenthesis when the break needs more air than those marks give. It can cut a sentence off. It can swing a thought around. Used once, it clarifies. Used in every clause, it preens.

That last distinction matters more than the AI rumor does. The mark is not the problem. The habit of interrupting every sentence for effect is the problem, whether a person does it or a model does.

Ellipsis or ellipses?

One set of three dots is an ellipsis (… ). Two or more of those marks are ellipses. The singular names the mark. The plural names the instances like on this Whitman phone case.

The job is omission or an unfinished thought: something left out, or a sentence that does not close. It is not a dash. It is not a hyphen with extra periods. On the page it can be the single character (…) or three separate periods (...). Either way, it means there is more, or that the more is being withheld.

We use an ellipsis, like on this Thoreau quote shirt, when the silence belongs to the line. We do not use it as a mood. That is the same rule as the dash: the mark has to do work.

How one key swallowed three marks

Printers once kept separate pieces of metal for each width. The typewriter, from the 1860s onward, did not. Keyboard space was scarce, so designers issued a single compromise character: the hyphen-minus. It had to stand in for a hyphen, a minus sign, an en dash, and an em dash. Early computers inherited the compromise. Most of us still type on that inheritance.

So the popular habit the reel describes is not laziness in a moral sense. It is infrastructure. Word processors will turn two hyphens into an em dash if you let them. Many phones will not. The result is a century and a half of people using the shortest mark for jobs that belong to longer ones, and a smaller group of writers who went looking for the real character and kept it.

On a Mac, Option-Shift-Hyphen produces an em dash. Option-Hyphen produces an en dash. On iPhone, hold the hyphen key. The marks are not hidden. They are simply not the default.

What Dickinson actually did

Emily Dickinson did not invent the dash. Printed English already had one. Anne Toner, in Ellipsis in English Literature, traces an early dash in a letter from 1580 and an early dramatic use to a 1588 translation of Terence. Nineteenth-century prose is full of them. What Dickinson did was refuse to treat the mark as furniture. In the early 1860s she made it her primary pause: more often than the comma, more often than the period, more often than the exclamation she had favored earlier.

The manuscripts do not show a single tidy em dash. Scholars who have sat with the fascicles describe a family of strokes: long and short, sometimes tilted, sometimes closer to an elongated period, sometimes almost vertical, as if they were phrasing marks as much as punctuation. Print has to choose. Thomas H. Johnson's 1955 edition rendered most of them as short dashes. Earlier editors did something rougher. They replaced them.

When "Because I could not stop for Death" first appeared in 1890 as "The Chariot," the second line ended in a semicolon:

Because I could not stop for Death,
He kindly stopped for me;

Dickinson's manuscript ends those lines with dashes. The last word of the poem, in her hand, is not closed with a period. It is left open:

Since then - 'tis Centuries - and yet
Feels shorter than the Day
I first surmised the Horses' Heads
Were toward Eternity -

A period would have shut the door. The dash does not. That is not decoration. It is the thought.

She noticed when print took the choice away from her. In 1866 the Springfield Daily Republican printed "A narrow Fellow in the Grass" (as "The Snake") with a question mark she had not written. She wrote to Thomas Wentworth Higginson that the poem had been "defeated too of the third line by the punctuation." The third and fourth lines, she said, "were one." The mark was not a nervous tic. It was structure. She had already told him she did not print.

So the useful Dickinson story is not "she started the dash." The useful story is that she trusted an informal mark more than the tidy ones, that her first editors "corrected" her into conventional punctuation, and that the poems go flat when you do that. If you want company for keeping the long mark, she is good company. She is not the origin myth. For the argument that her dash is a design choice AI cannot counterfeit by copying the glyph, read The Dickinson Dash.

The rumor that a dash means a machine

By 2025 the em dash had been nicknamed the "ChatGPT hyphen." Commenters treated density of the mark as a detector. Writers and editors pushed back. The Washington Post asked the people who build the models. An OpenAI model-behavior lead said ChatGPT may favor the mark. The company has also said style depends on the prompt and is not a reliable authorship test. Later that year Sam Altman announced a smaller victory: if you put "no em dashes" in custom instructions, the model would finally comply.

None of that is a theory of punctuation. It is a theory of training data. Models learned from newspapers, essays, and marketing pages, the same places human writers already used the em dash because it is useful. A model that has read a century of magazine prose will reach for magazine furniture. Overuse can look synthetic. Presence cannot.

The test people actually want is not "does this sentence contain an em dash?" It is whether the sentence needed the break. Dickinson's dashes change what the line can mean. A model's dashes often add a second clause that sounds finished and says nothing the first clause did not already say. That is a writing problem. It is not a character problem.

What we do here

Quoteiac prints verified lines. We also write about them. In both places we use the marks the line requires. Hyphens when words join. En dashes when a span is a span. Em dashes when a thought breaks or widens. An ellipsis when a silence is part of the sentence. We do not swap them to look less like a machine, and we do not multiply them to look more like a magazine.

If a reader arrives convinced that an em dash is a confession, they will find a great deal of unrepentant punctuation on this site. That is intentional. The alternative is to let a rumor about software edit the English sentence backward, past Dickinson, past the newspaper column, past the type case, until the only "human" mark left is the one key the typewriter happened to keep.

Use the right mark. Use it on purpose. If someone accuses the sentence of being artificial, point them at Amherst, 1862, and at a letter that complains about a question mark. The dash was a human habit long before it was a statistical one.

From the Journal

The Dickinson lines we print, and the other Journal pieces that sit with them, live on the Emily Dickinson collection.

• The Dickinson Dash: What AI Can't Counterfeit — why her punctuation looks the way it does, and why copying the glyph is not the same as using it.
• Emily Dickinson Wasn't Fragile. She Was Volcanic. — the life behind the quiet room.
• Emily Dickinson's Poems Almost Didn't Survive — how the fascicles reached print.
• Why Emily Dickinson Valued Private Journals Over Public Fame — the argument she was making by not publishing.

Primary notes
On the three marks: standard American usage as described by Merriam-Webster and the Chicago Manual of Style. Names of the marks follow typographic em and en units.

On early printed dashes: Anne Toner, Ellipsis in English Literature: Signs of Omission (Cambridge University Press, 2015).

On Dickinson's manuscripts and editorial regularization: Emily Dickinson Archive; Thomas H. Johnson, ed., The Poems of Emily Dickinson (Harvard, 1955); R. W. Franklin, ed., The Poems of Emily Dickinson: Variorum Edition (Harvard, 1998); Kamilla Denman, "Emily Dickinson's Volcanic Punctuation," The Emily Dickinson Journal 2.1 (1993); Paul Crumbley, Inflections of the Pen: Dash and Voice in Emily Dickinson (University Press of Kentucky, 1997).

On the 1866 "Snake" complaint: Emily Dickinson to Thomas Wentworth Higginson, Letter 316 (1866), in Johnson, ed., The Letters of Emily Dickinson.

On the 2025 "ChatGPT hyphen" claim: coverage in the Washington Post (April 2025) and subsequent comments from OpenAI staff and Sam Altman on custom instructions (November 2025). The claim is treated here as a rumor about frequency, not as evidence of authorship.

Written by Provenance Desk