AI isn't destroying writing. It is destroying the monopoly on polished prose.
Imagine reading an essay and enjoying it. The argument is clear, the examples work, and a few lines make you laugh. You reach the end feeling that the writer has explained something you had vaguely sensed but never quite put into words.
Then you discover the essay was AI-assisted.
Suddenly, the prose feels contaminated. The argument seems less impressive. You begin examining the sentences like an official at the literary Olympics. Was this adjective performance-enhanced? Should the metaphor provide a urine sample?
Part of your reaction is perfectly sensible. For centuries, polished prose was reliable evidence of education, effort and thought. You have just learned that this evidence no longer proves what it once did. A signal has broken, and recalibrating is what intelligent readers do.
But notice that nothing in the essay has changed. The facts and the argument remain the same. And the intensity of the response — contamination, fraud, betrayal — suggests that something more than recalibration is at play.
Perhaps we are not only protecting writing. Perhaps we are protecting the status once derived from being able to do it well.
Language was the road and the toll booth
Language has always been the primary currency for spreading ideas. You could understand illness, agriculture, poverty or human nature better than almost anyone, but unless you could explain what you knew, your ideas were unlikely to travel far.
The ability to write well was scarce. It required education, practice, time and a particular sort of mind. It also helped enormously to be born into the right country, class and language. A person with brilliant insights and poor grammar could be dismissed as unintelligent, while a person with ordinary insights and beautiful prose could build a career.
Language was the road along which ideas travelled, but it was also the toll booth. Not everyone could afford to pass through.
Some people found ways around it. The wealthy employed editors, speechwriters and ghostwriters. Politicians delivered stirring speeches they did not write, and celebrities produced autobiographies despite showing no prior evidence of literacy. I spent years working in radio and television, where nothing goes to air without producers shaping it. Nobody ever accused me of fraud because a producer tightened my script.
We mostly tolerated assistance when it was human, expensive, and reserved for the powerful. Mostly. When John F. Kennedy won a Pulitzer Prize for Profiles in Courage — a book largely drafted by his aide Ted Sorensen — the accusations, when they came, were not about receiving help. They were about claiming a prize for craft that was not his. Hold that distinction; we will need it later.
Now that assistance is cheap and available to almost everyone, it has suddenly become a moral crisis.
Beautiful language, ordinary ideas
We have spent centuries confusing verbal sophistication with intellectual depth.
The confusion is forgivable, because the two often travel together. Clear writing often reflects clear thinking, and writing can be how an idea is discovered. I rarely know exactly what I think about a topic until I have tried to write it down and watched half of it collapse.
But polished prose can also mask a lack of substance. An elegant sentence can sound intelligent even when it carries little weight. Confidence can lend the dubious an air of authority. A well-placed metaphor can distract readers from the fact that the argument beneath it is shallow.
Meanwhile, people with genuine expertise go unheard because they cannot present their knowledge in an engaging way. I have spent three decades reading referral letters from senior clinicians whose prose is as impenetrable as their handwriting — colleagues who understand human suffering far better than a columnist capable of producing three metaphors over one chardonnay.
Until now, the columnist had the advantage. AI is beginning to remove it.
The price of competent prose has collapsed
AI has made polished, readable language abundant.
A scientist can turn rough notes into an explanation everyone can follow. A second-language speaker can argue without sounding less intelligent than they are. A clinician can describe a patient without first completing a minor in creative writing.
AI can help people say clearly what they previously could not say clearly enough to be heard.
The printing press reduced the cost of copying language. The computer reduced the cost of editing it. The internet reduced the cost of distributing it. AI reduces the cost of producing competent language in the first place. Whether that constitutes a bigger revolution than the printing press is a question for future historians. It will certainly be a noisier one.
This does not mean everyone will suddenly have good ideas. It means everyone will have access to the language in which good ideas compete. Poor writers with good ideas will gain entry. Good writers with good ideas will have a more powerful tool. Poor writers with poor ideas will produce enormous amounts of highly polished rubbish.
The people in the most trouble are good writers with mediocre ideas.
Their fluency once served as a moat. It gave their ideas an appearance of intelligence and authority that less articulate people could not easily challenge. Now almost anyone can produce 1,500 polished words with a respectable opening, three subheadings and a conclusion suggesting the issue is more complicated than it first appears. (This essay has six subheadings. Twice the risk?)
AI will not abolish writers. It will erode the advantage of merely sounding like one, because polish alone will no longer be enough.
Which predicts, rather neatly, who will be angriest.
The grief behind the outrage
If fluency is a moat, the loudest fury about AI should come from the people whose living depends on it. Look around: the fiercest anti-AI columns are being written by columnists. Anyone in the writing community knows how hot this topic has become. Substack is loaded with debate. We are not quite taxi drivers blockading the streets against Uber, but we are getting close.
As a psychiatrist, I see grief beneath the fury — and I think there are two kinds.
The first is grief for lost meaning. Humans have long regarded creation as our most sacred capacity. Other animals could run faster, smell better and bite harder, but we could write symphonies, prove theorems, paint ceilings and produce novels about middle-aged people having affairs in coastal towns. For those who have devoted their lives to discovery — in writing, mathematics or music — the prospect of a machine doing it better is not merely a career problem. It is a wound to the meaning of the work itself. That grief is sincere and deserves respect rather than mockery.
The second is grief for lost status. Creation was not only what made humans special; it was what made some humans special among humans. Over the past couple of years, we have lost confidence that the best sentence, image or idea in the room must have come from a human mind. For those whose identity rests on being the best writer in the room, that is an injury of a different kind — and it produces less mourning than rage.
The reaction follows a familiar pattern. First comes denial: AI merely copies; it does not understand; it has no soul. Then anger: its work is theft, fraud, “slop”. Then the purity rules arrive: real writers choose every word themselves; real thought must emerge untouched from the sacred human interior.
Some of these criticisms are reasonable. Much AI-generated work is bland and repetitive. But much human writing is bland and repetitive too. We should criticise generic writing for being generic, false claims for being false, and bad arguments for being bad. We do not need to conduct a paternity test on every adjective.
Assistance is not abdication
And before you ask: yes, this essay is AI-assisted. Consider this my doping declaration. Substack has just installed an AI detector, so you no longer need to take my word for it — scan away. This essay will test positive. I have never been prouder to fail a drug test.
For the record, I was writing books and articles long before AI arrived to help, and I have been open about my use of AI — my website includes a full disclosure. I love sharing ideas, but I'm a mediocre writer. My English scores at school were terrible. Both my books were co-written with people more skilled than I am and then extensively edited by the publishers. My columns for The Conversation were edited by professional editors. I have always needed help; it just used to be human.
By the time AI arrived, I was semi-retired and writing almost full-time. It was a godsend: high-level editing, honest feedback, and an easy research tool. I experimented — some pieces drafted by AI, some merely assisted, and some entirely solo. At this point, refusing a tool this powerful would require me to be an idiot, a purist, or an extreme conservative. I have been all three at various points in my life, but I try not to combine them.
The important distinction is not between writing touched by AI and writing produced in a state of technological virginity. It is between assistance and abdication. Using AI to interrogate, organise and express your thinking is assistance. Asking it to manufacture beliefs you do not hold, expertise you do not possess and claims you have not checked is abdication. The question is not whether every sentence emerged unaided from the author's brain. It is whether the author remained intellectually present.
When I write with AI, the arguments are mine. The examples, judgements and tone are mine. AI helps me organise the structure, test the claims and refine the language. I reject suggestions, revise sentences and remove passages that do not belong. Most importantly, I take responsibility for the finished piece.
That is authorship. Not everyone will agree, and that's fine: writers can work however they like, and readers can judge accordingly. (You may also have worked out which quadrant of the earlier matrix I write from. I like to think it's one with the good ideas.)
There will still be places where process matters — exams, journalism, research, prizes. Kennedy's Pulitzer is the template: the sin was not the help but the claimed credit. Literary writing is the genuinely hard case, because the precise arrangement of words may be the artistic achievement itself.
One more objection deserves a straight answer. Some readers value human origin the way people value handmade furniture or live music: provenance is part of the product. Fair enough — I occasionally feel the same. But be clear about what you are paying for. What a reader buys from a writer worth reading is judgement, experience and responsibility for the claims. None of that disappears with assistance. What disappears is the hand-carving of individual commas, and I suspect fewer readers were paying for that than writers like to think.
The new scarce currency
Competent prose is becoming cheap. That does not mean words no longer matter. It means that neatly arranging them will become a poor proxy for intelligence.
Do not expect readers to judge every argument purely on its merits; nobody has the time. The toll booth will not disappear — it will relocate. When anyone can produce competent prose, the scarce resource becomes attention, and the new toll collectors will be trust: named authors, track records, editors, curators, and the friend who says “read this”. The question shifts from “Can this person write?” to “Has this person earned the right to be read?”
That is uncomfortable because we liked the old shortcut. Polished language told us who was educated, authoritative and worth listening to, and it let institutions decide whose ideas counted before engaging with the ideas themselves. That system is weakening. Good.
So let’s worry less about who chose which word and in what order, and more about whether the ideas are true, useful or interesting. The words still matter. But the fingerprints on the words matter far less than we think.
When everyone can write well, we may finally have to ask who has something worth saying.