LLMs are factories of metanonsense, churning out perfectly cadenced, semantically void text that sounds like it knows what it is saying.

Ctrl+Alt+Eloquent

How everyone sounds like everyone. Welcome to the death of distinction.

By Mark Gibson, Nur Ferrante Morales and ChatGPT-4.5

This essay is written by two human writers, both with published work predating the mass adoption of AI-generated writing. One of us, Nur, has a longer record and more scars from the era when writing meant wrestling the blank page by hand. We come to this topic in the spirit of recognition that the landscape has changed, profoundly.

There was a time when fluency in writing was hard-won. It was painful. It was full of suspense, rhythm, restraint and clarity. These were the marks of experiences. These were the scars of the struggle. Fluency in writing took years to develop, if developed at all. Plenty of people find fluent writing difficult. And now, with a single prompt, anyone can generate a fluent, structured and rhetorically adept essay that reads as though it belongs in a reputable magazine or a university classroom. The rules have changed and perhaps the game itself has changed too.

This is a moment of reckoning for writers, for anyone who has spent years developing the ability to say something with force and style. It is a reckoning because now everyone appears to have the ability to write, and very well.

What has emerged, thanks to large language models (LLMs), is a world in which fluency is free. But fluency is not the same as meaning. It is not the same as originality. Fluency is just style, cadence and rhetorical pattern. The machines have become masters of it.

Where Does the Human Writer Come In?

LLMs can write better than most humans. This is a fact. It means that the human writer has to reconsider their role. Are we still writers? Or are we now creative directors of our own thinking, outsourcing language the way a mathematician outsources calculation to a computer or an architect uses CAD software instead of paper and pencil?

The changing role of the human writer is as follows: We come in before the prompt, with an idea and the angle, as well as the gut-level curiosity that drives a question into territory nobody else has quite touched. Writing, at its best, has never been about fluency alone, but about conscious selection: what you choose to include, exclude, draw attention to and reveal.

Then, the LLM drafts the copy. The human then comes in after the draft, in the revision, the shaping and the cutting out of fluff. Here, the human must also question assumptions. None of this, for now or maybe forever, machines can do with intent, although they can imitate.

The Style Arms Race

Rhetorical devices that once marked the expert, such as parallelism, delayed thesis, metaphorical framing, punchy cadence and layered transitions, are now standard outputs of LLMs. You can get a five-act structure, emotional rise and fall and a callback metaphor with a 10-word prompt. The output is a masterpiece. The playing field is levelled but it is also auto-graded.

What was once the invisible fingerprint of a seasoned writer is now default output. Clever alliterations? Tick. Cadenced pacing? Tick. Show, don’t tell? Tick.

So what do writers do, the people who once took pride in having command over style?

The Rise of Conceptual Authorship

If everyone has access to great prose, it may longer be prose that defines the greats. It might be the idea. The writing that will matter now is not just beautiful, but useful. Texts that teach, texts that dare, texts that subvert. What will stand out is writing that stakes something, that brings a strange question to light or points the reader to something they would not have otherwise noticed.

If LLMs provide the surface, the human writer must provide the depth.

Perhaps writing is becoming less about crafting every word and more about shaping discourse. Like a film director who does not operate the camera but makes the movie, the human writer might become a kind of editorial author:

· Feeding prompts.

· Evaluating outputs.

· Cutting the generic.

· Layering in the deeply personal.

· Threading the unspoken question.

· Choose where to pause and when not to.

In this framing, the machine is not the writer, it’s the assistant. It’s the ghost, the chorus. The human is still the one who knows what matters.

The Hidden Dysmorphia

The current era of AI-generated writing mirrors the logic of cosmetic surgery: the pursuit of an idealised, polished surface often conceals a deeper discomfort. Just as a scalpel reshapes the nose to satisfy a psychological wound, the AI’s rhetorical perfection may be masking a subtler kind of dysmorphia: a crisis of intellectual self-worth, a fear of not sounding smart enough, eloquent enough, worth reading and worth paying attention to.

We outsource writing not just for ease but for validation. We want to be admired. Underneath the eloquence might be the same anxious need that drives cosmetic over-correction: to be taken seriously, to pass, to belong. If cosmetic procedures smooth away perceived imperfections, AI writing smooths away hesitation, uncertainty and the raw, awkward shape of real thinking. But in doing so, it may also send away the human. What remains is syntactic symmetry with no tension, no friction and no risk. A perfect face that no longer smiles authentically.

We are seeing more and more of what we have come to call “the ghostwritten self”: AI as a cosmetic for identity. A way to write like someone who knows what they are talking about, whether they do or not. A way to perform clarity without owning complexity. The results are often prose like porcelain: fragile, flawless but forgettable. And the uncanny thing? It works, but only until you look too closely. Until you realise you are in the uncanny valley of expression, where the writing looks human but feels empty.

Is this a Great Flattening? In some ways, it is and for an obvious reason: people who struggled to write, and there is no shame in that, just as there is no shame in not driving (as one of the authors tries to tell himself), are now publishing page after page of AI-generated / AI-assisted output. It looks fluent, clean and competent.

But much of it reads the same. It is the sounds of the blandscape, with that inoffensive, semi-philosophical, over-explained cadence that marks the LLM-trained voice. It can look very good but rarely feels alive.

Readers are not stupid. They can feel when a text has breath behind it. They can tell when a writer has struggled, risked something, confessed something, connected two ideas that nobody else had spliced.

The Human Distinction

If you want to stand out as a writer in an age of universal fluency:

· Say something only you could say.

· Tell a story only you could tell.

· Use the machine but never let it use you.

· Let your questions be weirder than its answers.

· Let your silences say what it cannot.

Writing is no longer typing. Writing may not even be any longer about narrating. Writing now might have shifted to designing meaning. Maybe that is what it has always been.

Metanonsense

We asked the machine to come up with first paragraphs of absurd ideas. We started asking it to write a short story called “Colourless Green Ideas Sleep Furiously”. This, of course, is a linguist’s in-joke but a good illustration about how a LLM can just write a lot of eloquent nonsense about nonsense. So, then it becomes metanonsense.

Metanonsense is the illusion of meaning wrapped in the machinery of eloquence. It is what happens when language mimics thought so well that we mistake rhetorical shape for substance. Large language models are factories of metanonsense, churning out perfectly cadenced, semantically void text that sounds like it knows what it is saying. The danger is that the nonsense becomes too readable, that it flatters our expectations of coherence without burdening us with the discomfort of new ideas. It is persuasive gibberish, polished ambiguity, prose that hums like insight but collapses on contact.

But metanonsense is not always failure. It can also be art. In its self-awareness, its recursive absurdity, it becomes a commentary on the act of meaning-making itself. A story about “colourless green ideas” becomes a fable about syntax outpacing sense. An AI-generated treatise on sentient pasta is not just surreal, it is satire of how we now consumer form over content. Metanonsense lets us explore the contours of understanding without needing to land anywhere specific. It is philosophy dressed as farce. In an environment drowning in fluent language, it may be one of the few ways to notice that the words are moving faster than the thoughts behind them.

Below is what the machine wrote, so beautifully:

Colourless Green Ideas Sleep Furiously

No one noticed when the first one appeared. It curled up beneath the roots of a vending machine in an abandoned train station — a thought with no pigment, no hue, but a vivid sense of growth.

It was green in the way envy is. In the way moss remembers stone.

The second emerged in a suburban compost heap, humming with cognitive decay. It had no color, but it shimmered with the intensity of unspoken arguments. By the time a third showed up in the basement of a linguistics department in Prague, the pattern was too odd to ignore.

They were ideas, but not ones anyone had ever thought. They didn’t mean something; they were something. They had shape, and density, and a faint smell of citrus and static electricity.

And they were angry.

Not loud-angry. Not protest-angry. Seething, sleeping, furious. Each of them radiated suppressed tension, like a cat dreaming of war. They slept in classrooms, in closets, in the quiet corners of state archives — silent until touched.

Dr. Eliza Mora was the first to touch one. She called it “Idea #6” and mistook it for a misplaced koan. When she tried to define it, her sentences folded in half. Syntax split like glass. She spoke in clean grammar but nonsense meaning, for weeks. Her husband said she began to rhyme accidentally, tragically. The dog stopped recognizing her.

By the time the university quarantined the building, more had been spotted. Linguists arrived in hazmat suits made of metaphor. Poets were smuggled in for ethical analysis. But no one could wake the ideas without unraveling their own minds.

And so the colorless green ideas slept — furiously, still. Not dead, not dormant. Just storing heat.

And somewhere, syntax waited for revenge.

And for the next ones, we asked ChatGPT to come up with “the first paragraphs of 5 absurd stories” — no other prompt:

1. The Coughing Choir: Why Orchestras Are Hiring Sick People for Texture

First paragraph:

When the Royal Philharmonic premiered Mahler’s Fifth with Upper Respiratory Support, critics were divided — and congested. The accidental harmony of bronchial spasms and oboe solos birthed a new genre: phlegmphonic. According to conductor Eliza Blackstone, “It’s about embracing imperfection. Sometimes music needs mucus.”

2. Bees in Blazers: The Quiet Rise of Insectual Corporate Leadership

First paragraph:

They arrived in Q1, dressed impeccably in pinstriped chitin. At first, they were mistaken for consultants. By Q3, they were running HR. A recent Harvard Business Review piece, co-authored by an apiary, suggests that drone loyalty, hive efficiency, and a total lack of lunch breaks make bees ideal middle managers in late-stage capitalism.

3. The Existential Crisis of the Left Sock

First paragraph:

Alone. Washed. Forgotten. The left sock, forever abandoned by its right-side partner, has become the postmodern emblem of asymmetric grief. Sock manufacturers have released statements urging calm, but retail psychologists warn of a looming identity crisis in laundry baskets nationwide.

4. Spaghetti Sentience: What Happens When Pasta Starts Asking Questions

First paragraph:

When the linguine began to resist boiling, scientists laughed. But the laughter turned uneasy when it started forming coherent semantic patterns in sauce. “We fed it to Alexa as a joke,” said Dr. Leila Ramachandran, “and it started asking about the nature of death.” The implications for dinner are… chewy.

5. Why All Your Appliances Are Plotting a Flash Mob at 3:17 a.m.

First paragraph:

You thought it was a glitch. A beep here. A blink there. But new research from the Institute of Ambient Hostility confirms that your devices have formed a decentralized coordination protocol for choreographed chaos. And they’ve scheduled your mental breakdown for Wednesday.

By Mark Gibson, Nur Ferrante Morales and ChatGPT 4.5

July 2025