Independent newsroom The Wyre News Network OpEd desk

Perspective 7 min read

The Punctuation That Snitched

The em-dash spent five hundred years doing elegant work in English prose. Then the machines fell in love with it, the internet renamed it the ChatGPT hyphen, and a piece of typography became an accusation. The real history of the dash, the tells that actually give machine writing away / the triads, the hedges, the vocabulary with no fingerprints / and why the giveaway was never the punctuation. Plus a confession: we banned it first.

By Abby Moss for the Wyre OpEd desk

Somewhere in the last two years, a piece of punctuation that predates the printing press's four hundredth birthday became an accusation. Post a paragraph with an em-dash in it / that long horizontal stroke, the one wider than a hyphen / and someone in the replies will announce that a robot wrote it. The dash spent five centuries doing quiet, elegant work in English prose. Then the machines fell in love with it, and it started testifying against everyone.

We build AI systems for a living and we run a site with a standing ban on the em-dash, so consider this both an expert opinion and a confession. Here is how the dash got here, why the models lean on it so hard, what actually gives machine writing away, and why the tell was never really the punctuation.

Five hundred years of respectable service

The em-dash is named for its width: in traditional typesetting it is as wide as a capital M, the way its narrower sibling the en-dash matches the N. It grew up alongside movable type, and by the eighteenth century writers had discovered what it was for / the pause that is not quite a period, the swerve mid-thought, the aside that a comma cannot hold. Laurence Sterne scattered dashes across Tristram Shandy like punctuation confetti. Emily Dickinson built half her music out of them. The dash became the mark of a writer thinking in real time on the page, or at least performing it.

Then the typewriter nearly killed it. No em-dash key, so writers typed two hyphens and moved on, and for most of a century the true em-dash retreated into professionally typeset books and magazines. Word processors brought it back through the side door: type two hyphens and autocorrect quietly upgrades them to the real glyph. Which means that for decades, the em-dash lived overwhelmingly in one kind of text / edited, published, professional prose. Remember that. It is the whole story.

Why the machines love it

Large language models learned to write from the text humans made available, and the best of that text / the books, the magazines, the quality journalism / is exactly the territory where the em-dash flourishes. Casual human writing on the internet barely uses the true dash, because no keyboard offers it. Edited prose uses it constantly, because editors and typesetters put it there. Train a model to sound polished and it absorbs the punctuation of polish along with everything else.

The dash is also, mechanically, a gift to a machine assembling sentences one token at a time. It is the universal connector: it can pivot, qualify, append, contrast, or dramatize, and it is never grammatically wrong. A model that wants to extend a sentence with one more nuance / and the models always want one more nuance / finds the dash sitting there like an open door. Nobody outside the AI labs can fully prove the mechanism, and we will not pretend to. But the observable output is not in dispute: machine prose reaches for the em-dash at a rate no human population ever did, and the internet noticed.

So the noticing became a heuristic, the heuristic became folk wisdom, and by 2025 the em-dash had a new nickname: the ChatGPT hyphen. Five hundred years of service, undone by association.

The rest of the lineup

The dash is the famous suspect, but it never works alone. The tells that actually add up, in rough order of reliability:

The "not just X / it's Y" reflex. Machine prose is addicted to the contrast scaffold: it's not just a tool, it's a transformation. It's not about the product, it's about the journey. One of these is emphasis. Five of them in a page is a template showing through the paint.

The rule of three, everywhere, always. Clear, concise, and compelling. Faster, smarter, and more secure. The triad is a fine rhetorical device that machine writing deploys with the regularity of a metronome, because it is the statistically safest way to end a sentence with a flourish.

The vocabulary with no fingerprints. Delve. Tapestry. Landscape. Robust. Seamless. Elevate. Unlock. Navigate the complexities. Foster innovation. In today's fast-paced digital world. None of these words is wrong; all of them are what a sentence sounds like when it was optimized to offend no one. Humans have verbal tics too / the difference is that a human's tics are theirs, and these belong to everyone.

Hedged symmetry. While X has its advantages, it's important to note the drawbacks. Both sides are presented, weighted equally, and abandoned. Real writers have positions. Machine defaults have balance, because balance is the cheapest way to avoid being wrong.

Structural tells. Every paragraph the same length. Every section the same shape. A bolded phrase at the top of every bullet. An "In conclusion" that summarizes what you read ninety seconds ago. Perfect grammar with no texture / no idiom, no aside, no sentence that exists purely because the writer enjoyed it.

The missing receipt. The most reliable tell of all: prose that is specific about nothing. No number that could be checked, no example with a name attached, no detail the writer could only know by having been there. Machine text describes categories; human text describes Tuesdays.

In defense of the dash

Here is the uncomfortable part for the punctuation police: every one of those tells, including the dash itself, is something good human writers do sometimes. The em-dash discourse has produced a real casualty list / professional writers accused of outsourcing their own voice because they punctuate the way edited English has punctuated for two centuries. A tell that fires on Emily Dickinson is not a detector. It is a mood.

The signal was never any single mark on the page. It is the accumulation: the dash and the triad and the hedge and the fingerprint-free vocabulary, stacked paragraph after paragraph, with no receipt in sight. One dash is typography. Twelve dashes wrapped around zero specifics is a confession.

Why a marketing consultancy cares

Because your customers now read with the detector switched on, and so do the machines. We covered Google's AI-slop penalty in our piece on what AI says about your business: low-value generated content has moved from "ignored" to "actively costs you." The human version of that penalty arrived first. The moment a reader's tell-detector fires, they stop evaluating your argument and start evaluating your authorship, and you have lost them for a reason that has nothing to do with your product.

And the irony is bitter: the businesses publishing the most machine-flavored prose are usually the ones using AI to save money on the exact asset / a credible voice / that AI just made scarce and valuable. Everyone can generate competent paragraphs now. Competent is the new invisible.

Full disclosure: we banned it

Now the confession. This website has a standing rule in its design system, sitting right next to the color tokens: no em-dashes, anywhere, ever. Our house style swaps in a spaced slash instead / like that / and our automated news wire scrubs the dash from every headline that passes through it. Machine-enforced, no exceptions, canon law.

Did we ban a piece of punctuation because the discourse turned it radioactive? Partly, yes. We will admit the dash did nothing wrong. But we publish in a world where readers spend attention deciding who wrote you, and every glyph that raises the question is a glyph working against the copy. The slash reads clean, nobody associates it with a chatbot, and five years from now, when the discourse has moved on to some new suspect, we may quietly pardon the dash. It has been rehabilitated before. It survived the typewriter; it will survive this.

Our position

We use AI in production every day, and here is the honest version: the tell was never the em-dash. The tell is the absence of a human decision. Prose gives itself away when nothing in it required anyone to choose / no stance that risks being wrong, no detail that had to be witnessed, no sentence bent to a particular reader. The dash just happened to be standing nearby when readers went looking for something to blame.

So the fix is not a blocklist of suspicious punctuation, and it is not swearing off the tools. It is editing with a spine: cut the triads down to the one that earns it, replace the balanced hedge with your actual position, trade the fingerprint-free vocabulary for a number with a name on it, and read the draft once asking a single question / did a person decide anything here? If the answer is yes, nobody will care how wide your dashes are.

Write like someone was there. That is the one tell no machine has learned to fake.

Related Reading

More Opinion

From the same desk

Perspective

A Rounding Error

Over twenty-four hours the single largest AI crawler on one consultant's website arrived roughly 1,500 times under a nonprofit research archive's name, sent back nothing, and asked for his SSH keys, his AWS config and his Firebase service account key. Across a hundred such paths: 1,028 requests, 6.7 megabytes, zero referrals, and requests for anything he had actually written rounding to nothing. It appeared in no security log, because security logs record rule trips and he was not blocking it. It appeared in exactly one place: the AI crawler dashboard, counted as an audience. Cloudflare's CFO says humans will be a rounding error on the internet within five years and admits he has been wrong every time, always low. Machine traffic passed human traffic in May 2026, two years earlier than the company forecast. Why robots.txt is a request that OpenAI's own documentation says may not apply, why blocking the obvious way costs you the visibility and keeps the fetches, what it means that agent config files are now on the standard secret-scanning wordlist, and the afternoon of work almost no company has done.

13 min

Perspective

Cut The Price Last

When demand softens, the first lever everybody reaches for is price. It is also the most expensive lever on the board, the hardest to reverse, and almost never the one that was actually broken. Redfin reports buyers have negotiating power across most of the country, the median housing payment fell to $2,575 because sellers cut asking prices to a one-year low, and pending sales hit their lowest since early April. Underneath it, July payrolls came in at minus 23,000 against an expected plus 83,000 with another 103,000 revised away. So the softening is real, and the question is which lever you pull and in what order. Why cutting price assumes a diagnosis you have usually not earned, the cheaper levers ranked, why the gap between the decision and the execution is where the margin leaks, and the measurement problem underneath all of it: cost per lead holds beautifully steady while cost per closed sale doubles.

6 min

Perspective

The Layer Nobody Looks At

A man in Connecticut typed instructions into his own court filing in three-point white text on a white background, where no judge would see them and any software would. They told the system to agree with him and to disregard the court's earlier denials. He was caught, warned, did it again, and lost his e-filing privileges. In the same seven days Google made the visible watermark on its AI images optional while keeping the invisible one, Anthropic shipped an API letting third parties detect text its models probably wrote, a new benchmark found that no frontier model clears sixty percent on basic visual perception with counting, localization and OCR among the weakest skills, and OpenAI started turning Mac users' clicks and keystrokes into a searchable memory their assistant can read, while warning in its own documentation about prompt injection. Five stories, one story. Every document, image and screen you produce now has two readers and they do not see the same thing. What the judge worked out on his own about where these attacks get caught, why provenance just moved into the layer humans cannot check, what a watermark hit actually proves and why a clean result proves nothing at all, and the ten-line diff almost nobody runs.

18 min