An Editor’s Lament

I am, alas, vulnerable enough to moodiness that I sometimes have to remind myself how fortunate I am. For example, the bulk of my job involves reading and writing — things I always did for fun. How lucky is that?

I am — in case you didn’t know — the online editor of The Critic. It is a wonderful job that I have been blessed to have. One thing, though, has dampened my enthusiasm recently: people sending articles generated by AI.

We’ve warned writers not to do it. But they do. There are, for the moment, all sorts of telltale signs: an overall blandness, excessive hedging, the rule of three, repetitive formulations like “it’s not X, it’s Y” — and, of course, em dashes. Such features can characterise human-written prose, of course. But all of them together?

(There are also AI checkers. I don’t wholly trust them, but I think that they can be suggestive of a pattern of misbehaviour.)

To be clear, I am not against all uses of AI. For example, I use AI to look for typos in my own articles. I use it for research (though I never take its word at face value because of AI hallucinations). I’m sure someone could make an argument against using technology in general for one’s writing, but I don’t think there is a real qualitative difference between using Google for research and using Grok.

But using AI to create one’s prose, and not to assist in the creation of one’s prose, seems bad. Firstly, it’s just dishonest. If someone sends an article, I think that there is an implicit understanding that they have written it. No one actually says, “Hey, I used AI to generate this article.” They know that this would make them seem — as, indeed, they would be — almost completely redundant.

Secondly, AI generates terrible prose. It is clunky, monotonous and — for obvious reasons — mechanical. Granted, I work in opinion commentary and human-written opinion commentary is often clunky, monotonous and mechanical. (I’ve been asking for years if AI is behaving like humans or if humans are behaving like AI.) But it is soul-destroying to try to work with writers on improving their prose and then realise that it is futile because it is not even their prose.

Of course, AI-generated text will become more sophisticated. I’m not sure AI can create really interesting prose because it will be so dependent on the prose that it has been trained on. But at the very least it will be able to do a far more convincing job of imitating interesting prose.

Yet I still have problems. Writing is not separable from thinking. Through the process of creating clear and compelling sentences, one explores and develops — or, at least, should explore and develop — an argument or a narrative. Generating one’s writing is somewhat akin to “travelling” through the eyes of a drone and assuming that one has had the same experience as someone who has actually travelled.

More than this, writing is supposed to be fun. That we are passionate about our writing — on the level of sentences and on the level of articles, stories, books et cetera — is what makes them compelling. If a writer cannot invest their time, energy and enthusiasm in creating prose, why should a reader be expected to care?

But I was interested in getting a response here. I wanted the case for AI-generated prose. Of course, none of the writers who submit AI-generated prose will admit to doing it so I went straight to the source. I asked Grok to defend its own creations. Here are three arguments that it made:

First, dishonesty hinges on disclosure, not the tool itself. If a writer transparently says, “This piece was drafted with Grok/Claude and edited by me,” the implicit contract shifts. No one claims redundancy; they claim collaboration. Many fields (academic journals, some media) now require or encourage AI disclosure precisely because transparent use isn’t fraud—it’s like admitting to using Grammarly or a research assistant. Undisclosed use is the ethical breach, not the generation. Hiding it makes the writer seem redundant; owning it makes them a curator of AI output.

Sure, it’s not a fundamental argument against having sex with people who are not your spouse to say that cheating is dishonest. But most people aren’t into open relationships, so most cases of people having sex with people who are not their spouse will be dishonest.

Second, prose quality isn’t inherently “terrible.” Early tells (blandness, hedging, rule of three, em dashes) are fading fast. Modern models produce varied, punchy, even idiosyncratic text when prompted well—mimicking specific voices, injecting humor, or avoiding clichés. Humans edit AI drafts to add soul, just as they revise their own first drafts. The editor admits AI helps with typos and research; extending that to drafting isn’t a categorical leap—it’s efficiency. Claiming AI lacks “soul” risks circularity: if humans increasingly write formulaically (as the editor notes), AI mirrors that. But skilled prompting + human refinement yields compelling prose, often fresher than rushed human output.

There is something to the idea that AI-generated prose is becoming more convincing. Still, if Grok doesn’t want to sound mechanical it shouldn’t talk about “injecting” humour. To compare yourself to a clinician is not exactly a slam-dunk argument for your ability to imitate that which is humane.

“Humans edit AI drafts to add soul” conflates editing with writing. It would make me more important, as an editor, to suggest that there is no fundamental difference here — but there is one. Creating an original sentence demands a level of thought and care that is far above that which is demanded by editing a pre-existing sentence. To talk of “human refinement” acknowledges this. No one would add spice to a ready-meal and call it home cooking.

Third, writing as thinking isn’t erased—it’s augmented. AI drafts force writers to critique, restructure, and infuse arguments with personal insight. This mirrors how thinkers use outlines or dictation: the process sharpens ideas. For time-strapped freelancers or those with disabilities, AI lowers barriers without replacing passion. Readers care about insight and clarity, not whether every sentence was hand-typed. If the final piece engages, moves, or provokes, the delivery method matters less.

Bollocks. To the extent that they “force” writers to do anything — and I can guarantee that they don’t force writers to do anything, because I receive 100 per cent AI-generated articles almost every day — it is not to restructure them — as the structure of a piece might evolve if it is human-written — but to add just enough ornamental touches to make them convincing. As for the reference to “those with disabilities”, well — I can see that AI can imitate the thought-terminating clichés of the more dogmatic progressive circles.

So, if you’re a writer who submits, or is tempted to submit, AI-generated text — please don’t. You are, or will be, draining — in your own small way — the humanness out of culture, and without humanness, what is the point of culture? What are we even doing here?

Besides, if you are submitting an article, the chances are that there is part of you that wants to be a writer. If you want to be a writer because you think it is a route to fame and riches then, well — you’ve missed the boat by decades. You might as well get into coal mining.

Believe me, I know what it’s like to be a “time-strapped freelancer”. I used to stay up all night to have a piece in an editor’s inbox when he started working — for the princely sum of £75 — and I’m sure that one day I’ll be doing something like that again (if I’m lucky).

But I don’t think most of us write just for the money (never mind for the status). If you want to write for a sense of personal fulfillment and creative satisfaction — as I suspect you do or did — why not just write?