With this fine essay by a Nigerian writer we kick off PREE VIEWS 5, a forum on the contentious subject of AI and its intersection with writing. Over the next few weeks we will add other contributions. We welcome well-articulated responses to this essay or on the general subject under discussion. Elnathan John’s essay was first published on his substack which he has since discontinued. PREE is proud to republish it here with his permission. We believe it adds intellectual heft and complexity to a discourse we cannot afford to ignore or reduce to simplistic black and white binaries.
I am seeing a particular kind of confidence everywhere now, and I recognize its scent because history has worn this perfume before. It arrives in the language of certainty. It speaks in final nouns. It says: AI can never write real poetry. AI can never produce literature with human depth. AI can never do this, can never do that, can never cross this line I have drawn today with a marker I will pretend is permanent.
Never is a seductive word. Never gives us posture. Never gives us immediate relief from ambiguity. Never lets us stand in the storm and call ourselves weatherproof. Never lets us feel morally clean while the world becomes more complicated around us.
What never almost never gives us is accuracy.
I want to be precise from the beginning, because this argument is easy to misread by people who enjoy misreading. This is not a defense of AI companies. This is not an advertisement for AI use. This is not a cheerful hymn to technological inevitability. It is not a claim that “progress” is good because it is progress. It is not a claim that writers should surrender to tools, nor a claim that we should stop caring about craft, or labor, or theft, or meaning, or energy, or power.
Human Exceptionalism
This is a narrower and harder claim: if we are speaking specifically about literature and poetry, the confidence with which many people assert permanent limits for AI is intellectually weak. It is emotionally understandable, yes. It is rhetorically satisfying, yes. It is historically and philosophically weak.
And if we are serious people living in a serious time, weak arguments are luxuries we can no longer afford.
I am not an AI skeptic in the specific literary sense in which many writers are currently performing skepticism. I do not believe AI will remain permanently incapable of producing writing that moves human beings. I do not believe there is a metaphysical wall around poetry that machine-assisted language can never approach. I think AI will get better and worse at the same time. Better at producing outputs we once assumed required exclusively human touch. Worse in the sense that it can become an extraordinary instrument in the hands of corporations and states that already possess too much power in a world where freedoms are thinning and public institutions are outmatched by private infrastructure.
Both things can be true together. In fact, both things are already true together.
The difficulty is that contemporary discourse prefers one-note performances. You are either a naive worshipper of technology or a pure defender of humanity. You are either for the machine or for the soul. You are either a futurist with a demo reel or a guardian of civilization with a Substack.
I decline this casting call.
I am more interested in what survives time and in claims that still stand when the software improves, when the market consolidates, when the novelty evaporates, when the children who grow up with these tools become adults and stop asking permission.
A large part of what I am resisting here is the old doctrine of human supremacy hiding in artistic language. We smuggle it in gently. We call it craft, authenticity, soul, human depth. Often those words are sincere and useful. Sometimes they are camouflage for a deeper claim: there are realms of value that belong to us alone, by right of species, and poetry is one of the final castles.
I do not share this confidence in human exceptionalism.
Humans might think of themselves as the apex species of the universe. But I do not assume that our present cognitive architecture is the final and best expression of intelligence. I do not assume that the qualities we prize in ourselves are beyond simulation, approximation, recombination, or surpassing. I certainly do not believe or assume that because we feel profound, we are therefore ontologically protected from competition.
We have declared final boundaries before. We declared boundaries around flight, and then we built wings of metal, declared boundaries around distance, and then we spoke across oceans as though across a table. We declared boundaries around sight, and then we built instruments to see cells, galaxies, radiation, wavelengths our naked eyes cannot even name. We declared boundaries around the body, and then we transplanted organs, replaced joints, edited genes, learned to intervene in life in ways that would have sounded like blasphemy or sorcery three centuries ago and now sound like medicine.
The pattern is old. Too often, we confuse present incapacity with permanent law and convert technical lag into metaphysics. Then history arrives with bad manners.
Guild anxiety?
So when someone says AI can never write a poem that moves people, I ask what evidence they think they have. Not vibes. Not wishful ethics. Not guild anxiety. Evidence.
Because if the evidence is human emotional reality, the argument runs in the opposite direction.
Humans are extraordinarily moveable creatures. We are moved by truth, yes, but not only by truth. We are moved by fabrication, staging, exaggeration, distortion, half-truth, sentimental editing, myth, prophecy, ideology, ritual, rumor, slogan, anthem, confession, and strategic performance. We are moved both by high art and we are moved by cheap manipulations. We are moved by beautiful lies and ugly lies. We are moved by stories we know are not empirically verifiable and by stories we do not care to verify because they serve a psychological or social function we are unwilling to surrender.
Politics is full of this. Intimate life, religious life, national life, all full of this.
I am not being flippant when I say that religion alone should force us to treat literary impossibility claims with caution. Billions of people orient their lives around supernatural narratives and miracle claims with total sincerity and existential commitment. And among those very many would rather die than abandon beliefs modern empiricism would classify as unproven or impossible. Whether one sees those beliefs as sacred truth, symbolic truth, community truth, private truth, or empirical fiction, the human fact is unchanged: narrative can move people beyond argument. Narrative can reorder behavior, loyalty, sacrifice, fear, tenderness, and courage.
If humans are already this narratively porous, this symbolically available, this emotionally responsive to structured language and collective meaning, on what basis do we insist AI-assisted language can never produce poetry that moves millions if not billions? Why is that impossible? What cognitive law forbids it? What part of the nervous system has sworn an oath to authenticate biological origin before feeling grief?
At least some part of the answer lies in status.
I think the “AI can never write real poetry” statement often carries a second, unspoken follow up: and therefore my role as a poet or writer remains protected. My gate remains relevant. My credentials remain scarce. My authority to define the real remains intact.
I understand this anxiety. I am not mocking it from outside. Writers are not fools to worry. Any system that can generate language at scale can flood attention markets, depress compensation, blur credit, and intensify already brutal precarity. Writers are right to worry about theft, style extraction, devaluation, and the replacement of living labor with synthetic output optimized for volume over meaning.
But there is a difference between warranted fear and bad philosophy.
I do not think good writers should capitulate to AI. However, while “If you use AI, you are not a writer” sounds righteous, it collapses distinctions we urgently need. It cannot distinguish between a student outsourcing thought and a novelist using a model as a pressure-testing interlocutor. It cannot distinguish between theft and transformation, between plagiarism and influence, between noise generation and careful revision, between cynical deployment and ethically disclosed collaboration. It cannot distinguish between people using tools to avoid thinking and people using tools to think harder.
Demand responsible use, not purity
Purity language has this problem. It feels clean, so it ignores complexity and it gives us a moral performance while practical reality keeps moving.
I have watched this unfold in other industries. A new tool appears. The guild panics. The gatekeepers announce a death. The market mutates. The craft survives but changes location. The definition of skill shifts. New mediocrities appear. New excellences appear. Then, twenty years later, people speak as if the old panic was unreasonable and inevitable and somehow unrelated to them.
Examples abound. Food processing machines in the kitchen? Not a real cook. Digital editing of photos? Not a real photographer. Oh and when the word processor arrived and you moved from the typewriter to something that could spellcheck? Not a real writer. Because, remember? A real writer was to write by hand and only then allow the typewriter copy exactly what the sacred thoughts of the writer was.
The pattern is not identical each time, but the rhythm is familiar enough that we should be embarrassed by our refusal to learn from it.
None of this means we shrug and surrender. This is where I want to be very clear: realism is not capitulation. Saying “we cannot wish AI away” is not surrender to corporate rule. It is the first sentence of strategy.
If a technology is already deployed, already capitalized, already integrated into institutional systems, already crossing borders faster than regulation, then serious thinkers have two choices. They can perform moral disgust from a shrinking island, or they can engage reality with enough rigor to shape norms, laws, and practices while there is still time.
I choose the second.
And this is exactly why this essay cannot remain only a literary argument. The literary argument sits inside political economy whether we like it or not.
The environmental cost is real. Large-scale computational systems draw enormous energy and water resources, and those costs are not distributed evenly. The communities paying those environmental costs are often not the communities owning the upside. The ethics of extraction are familiar because the structure is familiar: invisible infrastructures, opaque accounting, concentrated ownership, public externalities.
The corporate ethics are not peripheral. They are central to this argument. We are dealing with entities that routinely ask for trust while resisting transparency, ask for public adoption while minimizing disclosure, ask for regulatory flexibility while maximizing market capture. They speak in the language of empowerment while building dependencies hard to reverse. They frame themselves as inevitable, then ask to be treated as benevolent.
This is not a new behavior in capitalism. It is a new scale of that behavior.
In a world with shrinking freedoms, brittle democracies, weakened labor protections, and hyper-concentrated media ecosystems, AI is not just a writing tool. It is an infrastructure of persuasion, surveillance, ranking, and productivity enforcement. It can be used to open access and challenge gatekeeping. It can also be used to tighten control over workers, publics, and truth itself. It can democratize drafting and centralize authority at the same time.
That is the paradox. Ignore either side and you become propaganda for the other side.
I keep returning to one practical question: if we cannot presently destroy every corporation building these systems, and if we are not on the verge of abolishing the economic order that rewards this concentration, what does responsibility look like now?
It must look like demands, not vibes. I can think of a few.
Demand disclosure when AI is used in published work, especially journalism, education, and public communication where epistemic trust matters. Demand provenance standards so people can trace what sources shaped outputs and under what permissions. Demand compensation frameworks where creative labor was used in training or imitation. Demand stronger labor protections for workers pressured into invisible model maintenance, moderation, and annotation. Demand environmental accounting that is auditable, comparable, and tied to enforceable thresholds. Demand interoperability and anti-monopoly measures so infrastructure ownership does not become a permanent private constitution. Demand reader education so publics can interpret synthetic persuasion without collapsing into nihilism.
I know I just used a sequence there. (Does it feel like AI? I apologise.) The point stands: responsibility is a design problem and a power problem, not a purity slogan.
Moral hierarchies
There is another part of this conversation that I care about very personally, because I am a writer and I have spent my life among writers and readers and institutions that love to convert cultural habits into moral hierarchy.
I do not think literature makes me a better class of human. I do not think the ability to read complex prose is proof of ethical maturity. I do not think people who do not read books are therefore less adjusted or less capable of love, conscience, or communal responsibility.
Everything I know about humans tells me otherwise. I have spent time in universities among intellectuals and academics and I have spent time interviewing and talking to people who cannot read or write. I grew up around a lot of people in the latter category. Nothing in my experience tells me that books and university degrees make kinder or more moral humans.
Many of the most devastating atrocities in recent human history were planned by literate people, educated people, decorated people, people trained in law and philosophy and science, people who could quote and write poetry while authorizing cruelty. Race science was not invented by the illiterate. Industrial-scale environmental destruction was not designed by people outside institutions. Weapons systems are not built by villagers who cannot parse a sonnet. Bureaucracies of death are staffed by professionals who can write impeccably.
Meanwhile, forms of care that keep human life possible in vast parts of the world are carried by people with little or no formal literacy. People who cannot read can still enact ethical intelligence daily: restraint, generosity, hospitality, honesty, dignity, repair. They keep families and neighborhoods from collapse without ever publishing an essay about moral psychology.
So when literary culture speaks as if books confer moral aristocracy, I hear insecurity dressed as enlightenment.
This matters for the AI argument because the same insecurity often sits behind poetry exceptionalism. If literature is your temple and your badge, then the possibility that nonhuman systems might produce emotionally effective language feels like desecration. But desecration of what, exactly? Of beauty? Or of status?
Poetry is not diminished because more entities can generate verse-like forms. Poetry is diminished when readers stop caring about precision, when editors stop caring about rigor, when institutions reward volume over depth, when the conditions for patient attention collapse. That collapse can happen with or without AI. In many places, it was already happening.
A writer with no ear remains a writer with no ear even with a model. A writer with no ethics remains a writer with no ethics even with a model. A writer with judgment, formal curiosity, and responsibility can use any tool and still produce work with charge, friction, and consequence.
Tool access does not equal aesthetic mastery. Never has. (Look at me using never after railing against it in the beginning. Silly me. I am after all human.)
So what happens when AI begins producing poems that large numbers of people genuinely love? Some of this is already visible in small ways: synthetic lines that circulate because they are simple, resonant, and timed to social mood. Most of it is shallow now. Some of it will remain shallow. But shallow has always had a market. In fact shallow has always had the biggest market share. There is a reason more people have bought Fifty Shades of Grey than most of the remarkable books of literature published at the same time. Or why more people were moved by Psy’s Gangnam Style than any of the phenomenal classical or jazz music available at the time. The deeper question is whether work of genuine complexity and power can emerge from AI-assisted processes under serious human stewardship.
I think yes. I think it already can in fragments and will increasingly do so in full forms. Not because machines have souls. Because literature is an encounter between language and reader, mediated by context, form, expectation, and need. If a text arrives with cadence, image pressure, emotional intelligence, and conceptual surprise, readers respond. People do not run a blood test on the author before their tears form.
Human writing and literary culture
Will there still be value in unassisted writing, in slow apprenticeship, in tradition, in hand-built sentences forged through solitude? Absolutely. I would and do defend that value fiercely. But defending that value does not require metaphysical denial of everything else. It requires confidence in the thing itself.
If your faith in human writing requires the claim that nothing else can ever approach it, your faith is already fragile.
A more durable stance is this: human writing remains valuable because of responsibility, situatedness, accountability, risk, memory, and relation, not because of a mystical monopoly on pattern generation. Human writers answer for what they write. They stand inside history with names, bodies, obligations, and consequences. Those dimensions matter profoundly. They can coexist with the reality that machine-assisted language may become aesthetically and emotionally formidable.
I am aware of some of the pushback that will come, so let me meet some of it directly.
One pushback will say: you are normalizing mediocrity. I am not. Mediocrity needs no normalization; it was already normalized by platform incentives long before AI. I am arguing for higher standards, not lower ones. Precisely because cheap output is easy, editorial rigor becomes more valuable, not less.
Another pushback: you are dismissing plagiarism and theft. I am not. I am saying authorship ethics must become stricter, disclosure norms clearer, consent regimes enforceable, and compensation frameworks real rather than symbolic.
Another: you are reducing art to effect. I am not. I am refusing to reduce effect to origin. Art includes process, intention, relation, labor, form, risk, and community. Origin matters, but it is not the only dimension that matters.
And then there is this one: this is techno-determinism. I say, no. Techno-determinism says whatever technology enables will happen and should happen. I am saying what happens is political, institutional, and contested. We still choose norms. We still choose laws. We still choose incentives. We still choose what we reward and what we ban.
And finally, my dear fellow writer might say: this argument betrays writers. I will hold your beautiful writer hands and say this: what betrays writers is false certainty. Bad strategy. And pretending the future will obey our nostalgia. Writers deserve arguments that survive contact with reality.
I suspect part of my own posture comes from refusing to romanticize either human nature or literary culture. I write, I teach, I love language, I live by words, and still I do not believe the world owes me a metaphysical exemption. I do not believe my vocation grants me higher moral altitude. I do not believe the sentence is a sacrament that only one species can touch. I do not believe art is worthless either. I believe art is powerful and precarious. I believe power without ethics ruins art. I believe ethics without realism becomes performance.
So where does this leave us? At this risk of sounding like AI, here is another list:
It leaves us with a difficult discipline: Hold the contradiction without collapsing it. Hold that AI can become capable of producing poetry that moves vast audiences. Hold that this does not make AI good. Hold that corporate concentration around AI is dangerous. Hold that wishing AI away is not a plan. Hold that resisting corporate power requires policy, institutions, and public pressure, not literary incantation. Hold that writing is still craft, responsibility, and relation, even when tools change. Hold that readers are not fools, but are also not immune. Hold that humans are astonishing and brutal, creative and credulous, and that our susceptibility to story is precisely why this debate matters.
I return again to that first word: never.
Never is easy when you need comfort. Never is expensive when you need truth.
I am less interested in being comforted than in being correct enough to act. If we are thinkers, if we are writers, if we are people who claim to care about language and public life, then our task is not to guard a shrinking shrine of purity. Our task is to shape conditions under which language remains accountable, human dignity remains non-negotiable, labor remains visible, power remains contestable, and beauty remains possible.
The machine is here. The corporations are here. The harms are here. The possibilities are here. The question is whether we can build a future where neither propaganda nor nostalgia gets to decide what literature becomes.
-
If there are any typos, it is just so you don’t think I wrote this with AI.
