My academic speciality was text linguistics—old-school text linguistics. I was fascinated by the idea that a well-formed text could be recognised by its being both coherent and complete. Its parts belonged together; its argument or narrative developed intelligibly; and the reader was not left with the sense that something essential had been omitted.
That definition was useful, but it also left out whole layers of what language is. It said little about style, register, rhythm, resonance or voice. It treated language chiefly as a means of organising and transmitting meaning.
For much expository writing, that may be enough. My own observation is that, when I have coherent ideas, it is sufficient for them to be expressed in plain English. I do not need stylistic frills. The chief task is to ensure that the ideas are connected, that the argument progresses clearly and that the whole says what I intend it to say.
In this respect, I have no reservations about using artificial intelligence.
AI can help remove repetition, disentangle awkward syntax, strengthen transitions and reveal the underlying structure of an argument. It can often express more clearly a distinction that the writer has already perceived but has not yet formulated precisely. The thought remains the writer’s; AI assists in making its structure explicit.
Indeed, this is one of the strongest arguments for its legitimate use. Where language serves thought, AI may properly assist expression.
Plain English is not an impoverished form of language. On the contrary, it is often evidence that the writer knows what he means. Clarity is not the enemy of intelligence. A simple sentence may embody a complex distinction more successfully than a paragraph laden with rhetorical decoration.
But language does not always serve merely as a container for ideas. It can also be an artistic medium.
In literary, poetic or deeply personal writing, the language is part of the meaning. Rhythm, sound, ambiguity, allusion, register and cadence cannot be detached from the experience being expressed. Particular words and phrases carry historical and emotional associations. They resonate. They press buttons. Their effect arises not only from what they denote, but from everything they have meant elsewhere and from the particular consciousness that has chosen them here.
AI can imitate these effects. It can produce lyricism, solemnity, intimacy, indignation, tenderness or grandeur. Sometimes it can do so remarkably well.
Yet imitation is precisely where the difficulty begins.
AI can reproduce the outward signs of emotional or literary authenticity without having undergone the experience from which such language would ordinarily arise. It can simulate grief without grieving, faith without believing, love without loving and memory without remembering. The result may be effective, but its provenance is questionable.
When language itself constitutes the art, extensive AI intervention risks replacing authorship with simulation.
This does not mean that style is merely an optional embellishment added after the thought is complete. Even plain English involves choices of emphasis, cadence and tone. Nor does it mean that a poet or novelist must never use AI. A writer might use it experimentally, as one might use a dictionary, a set of formal constraints or even a collaborator.
The decisive question is not simply whether AI has touched the words. It is where the intellectual and creative responsibility resides.
Did the writer originate the thought, make the distinctions, judge the argument and accept responsibility for what is said? Or has the machine supplied not only the wording but the apparent depth, emotion and resonance?
There is therefore an important distinction between AI as a servant of coherence and AI as a counterfeit source of resonance.
Where language serves thought, AI may help us say more clearly what we already mean.
Where language is itself the art, we should be far more cautious.
