At times like this I find myself praying for the robot revolution to round up these lunatics and put them in the grammar asylum where they can no longer harm themselves and others.
Somewhere on LinkedIn this week, a writer who bills herself as a grammar authority proudly announced she can spot AI-drafted prose the moment it lands in her inbox.
Not by reading it, but by looking at it. The font gives it away, apparently. Aptos, specifically. Pair that with an em dash and the construction “it’s not this, it’s that,” and she claims to know everything she needs to know, without reading a single word (not quite sure how she spotted the “it’s not this, it’s that” without reading a singe word, but you know how these folk love to exaggerate their prowess.
But seriously. What an arrogant f*****g idiot!
The aptos font, beloved of millions, and the em dash, beloved of many more, are now definitive signs that anyone using either used AI.
As Henry II almost said, Who will rid us of these turbulent grammatists?
Her suggestion? Remove the evidence. Check your document for em dashes. Check it for “genuinely.” Change the font. Make sure, in her words, that it no longer sounds like anything other than “you”.
FFS! The goal is not better writing. It’s not getting caught!
At times like this I find myself praying for the robot revolution to round up these lunatics and put them in the grammar asylum where they can no longer harm themselves and others.
I’m not going to name her, because the post itself doesn’t much matter – it will be forgotten by the time you read this, replaced by an all-but identical one under a different byline. What matters is that this nonsense is replaying in a continuous loop across the writing professions, in ELT publishing, in journalism, in academic marking, book publishing and education.
And the reason it matters is that this is not literary criticism. It’s a retreat into what linguists call folk taxonomy, and it’s happening for a reason that has sweet f.a. to do with prose quality.
What folk taxonomy is, and why we need to call it out as such
A folk taxonomy is the classification system ordinary people reach for instead of a technical one – sorting animals into “creepy-crawlies” and “cuddly things” rather than phyla, sorting food into “healthy” and “junk” rather than macronutrients. It’s fast, it’s vivid, and it doesn’t require anyone to actually know anything.
That’s precisely the register the AI-detection cottage industry operates in.
Nobody arguing that an em dash is a tell can point to a corpus study showing em-dash frequency correlates with model output rather than with, say, Bill Bryson, Joan Didion, or every broadsheet leader column of the last hundred years.
Nobody flagging “delve” or “genuinely” as machine vocabulary has checked a frequency table against pre-2022 English. They don’t need to, because folk taxonomy isn’t measuring the text. It’s measuring a feeling – this reads too clean, too competent, too put-together, too whatever (I added a fourth too to pre-empt some idiot shouting “Gang-of-three! Proof of AI!”) to have come from an ordinary, harried human – and then reaching for the nearest available word to dress that feeling up as faux expertise.
And anyone with half a brain cell would think twice before claiming a font was proof of AI. But here we are talking the very outer reaches of the Luddite Fringe, where even half a brain cell is unlikely. Which of course means thinking at all, let alone twice, is not going to happen.
The Aptos font, for those unfamiliar, came along in 2019 (maths memo to the Luddite Fringe – 2019 happened before ChatGPT appeared in November 2022). It was called Bierstadt and Microsoft officially introduced Bierstadt to the public in April 2021 (another maths memo to the Luddite Fringe – 2021 ((all of it, not just April)) happened before Chat GPT appeared in November 2022). It was renamed in July 2023 and presumably the Idiot Squad googled when Aptos appeared, saw 2023, and immediately decided ChatGPT invented it because, well, why not?
In any case – and this is where it gets sooo funny – the standard default font for LinkedIn, whatever we paste in from any source, is Aptos!
Don’t believe me? This line you are reading was typed in Word in Times New Roman and pasted here. Now it’s Aptos.
In summary, then, the tell was never for one second the em dash or the Aptos font, but coherence itself. Anyone remember coherence? Structure, correct punctuation, a paragraph that goes somewhere… In other words, what’s being flagged as artificial is, overwhelmingly, just edited writing.
And the fact that so many professional communicators now find edited writing suspicious tells us far more about the state of the average unedited LinkedIn post than it does about any language model.
Why now, and why this vocabulary?
For as long as there have been gatekeepers of the written word, their authority has rested on a simple claim: that producing clean, structured, persuasive prose takes a rare and teachable skill, one they happen to possess and are willing to certify in others, for a fee, a byline, or a following.
That claim has just taken the single biggest hit in its history. Knocked for six as they say in the cricket world. (Just throwing that in for Carlo Carrenho and Richard Charkin – the former a one-time Brazilian so obsessed with any sports nonsense involving a ball, the latter British so pretends cricket is a sport). A model with no schooling, no red pen and no LinkedIn following can now produce publication-ready prose to order, in seconds, for anyone regardless of what school they went to or what their first language was.
That’s a genuine, structural threat to a certain kind of professional identity – not to writing, but to the gatekeeping of writing – and folk taxonomy is what a threatened gatekeeping class reaches for when the technical ground has been cut away from underneath it.
You can’t credibly say “only I can produce a well-formed sentence” anymore, so the goalposts move to something unfalsifiable: soul, spark, authenticity, the ineffable human touch that definitely isn’t present in this em dash but definitely was present in the same em dash last year.
It’s a defensive pivot, and a totally transparent one once you notice the pattern: first insist that mastery of form is what separates the talented from the unwashed masses who only think they can write, then, the moment a machine masters the form, insist that form was never the point.
But here’s the thing (and yes, AI says that, but yes too, Bill Bryson used it in his books likely before Sam Altman was born, so call off the anti-AI rabid dogs, please): I don’t think this is cynical, mostly. I think it’s frightened.
There’s an uncomfortable realisation sitting (Oops! Another AI frequency word. Down boy!) underneath a lot of these posts, which is that much of what got treated for decades as evidence of “good writing” – structure, transitions, varied sentence length, the rule of three, a point that resolves – was always more mechanical than its practitioners liked to admit. A large language model holding a mirror up to that is not a flattering experience for anyone whose professional worth was built on being one of the few people who could do it.
Same instinct, different target
I’ve made this point before, about a different symptom of the same disease. When commentators reach for “Gen Z” as an explanation for falling literacy – as though an entire generational cohort simply decided, collectively, to stop reading – they’re doing exactly what the grammar police are doing with em dashes: reaching for a visible, labellable scapegoat instead of looking at the actual mechanism.
In that case it’s a curriculum built on debunked synthetic-phonics methodology and the policy failures sitting behind it. In this case it’s a font. Both moves do the same job. They let the person making the accusation avoid the less flattering diagnosis – that the system they were trained in, or the standard they were enforcing, was never as robust as they desperately want to believe.
The part where I disclose an interest
I make my living from writing, as an author, novelist, journalist and one time playwright and TV writer, and I’m also a teacher. I have have been and have done so since before most of today’s AI critics could one-finger-type, and I don’t feel remotely threatened by any of this. The opposite, in fact. Empowered is the word!
I’m a hopeless illustrator, so I use AI for the visual work I could never do myself and could never afford to commission – the same way I use a supermarket instead of growing my own food, or buy clothes instead of weaving them. Nobody today rides a horse and cart to the office out of loyalty to pre-industrial values, and nobody who actually wants their writing read is going to hobble it in case an unqualified font detective with an insecurity complex and lots of LinkedIn followers mistakes competence for machine output.
Every one of the technologies that has historically unsettled a trade eventually created more opportunity than it destroyed, for the people willing to pick it up: the photocopier didn’t end publishing, the spellchecker didn’t end editing, the camera didn’t end illustration.
Car automation put coachmen out of work and created an entire automotive economy. The mobile phone did the same to switchboard operators and created app development, mobile publishing, and what have you. New industries orders of magnitude larger than those they replaced.
Whaling collapsed once kerosene arrived, and nobody today mourns it as a lost craft, however skilled the whalers were. In every case the pattern is identical: the technology creates new terrain, and it’s only the people who insist on defending the old terrain, on principle, past the point it can support them, who fail to make the crossing.
Writing is not exempt from that pattern just because the people it’s currently unsettling used to be the ones handing out the grades.
None of this means anything goes
There are real questions about AI and writing worth taking seriously – disclosure in academic and journalistic contexts, perhaps; the treatment of non-native English writers who get disproportionately flagged by detection tools precisely because their prose is “too clean” by native-speaker folk standards; the difference between using a tool and outsourcing your judgement entirely.
Those are structural, checkable, falsifiable questions, and TNPS has spent plenty of column inches on exactly that ground.
Whether an em dash showed up in your document isn’t one of them. Never was. The people telling folk to change their font aren’t defending literacy. They’re defending a hierarchy that used to run through them, and folk taxonomy – soulless, hallucination, slop, tell – is simply the vocabulary available to anyone who’d rather not face up to the fact that the world has moved on without them.
On the bright side, they’ll have plenty of former whalers, coach & horse drivers, chimney sweeps, and typewriter manufacturers to keep them company.
This post first appeared in the TNPS LinkedIn Analysis Newsletter.
