Editors Shouldn't Care Whether Writers Are Using LLMs

"You must promise to forget this fellow — to illiterate him from your memory!” - Mrs. Malaprop in “The Rivals” (1775)

One thing that often happens whenever a new technology emerges that is capable of simplifying a process that used to requires years of skill to learn is that doing things the old-fashioned way retains a certain antiquarian prestige. There are all sorts of reasons for this, not least that those who were trained in the old ways, and whose social status is tied up with have mastered its methods and codes, wish to hold onto that status.

So it is with LLM-generated prose today, which many an old-timey writer and editor is regarding with a mix of disgust, fear, and horror. Are we all about to be rendered redundant by these machines?! I know what to do: let’s shame people into not using them! Many publications today (here, in the middle of 2026) are demanding that writers not use LLMs or at minimum disclose the use of LLMs, which admission is then appended to the article like a scarlet letter. Like it or not, there’s no question that (here today, in the middle of 2026) writing things “artisanally,” e.g. without LLM assistance (though perhaps not literally by hand), continues to carries more prestige than writing done with AI assistance.

But this present social convention is creating an incentive problem. As long as writing without AI carries more prestige than writing with it, but writing with AI is both labor-saving and easy to conceal, writers will have a reason to be less than transparent about their process — even when transparency is technically permitted. No disclosure pledges or honor codes are going to change the underlying logic. As the editor of a magazine myself, where we at present permit the use of LLMs but demand disclosure and then note in the publication what the authors admit about their use of LLMs, I am constantly seeing authors hedging on this front.

In light of this perverse situation, allow me to propose something that I expect may be controversial (here today, in the middle of 2026) but which I am highly confident will seem utterly obvious in a year or two, if not sooner: the solution is to treat AI-assisted writing as so routine and unremarkable that we stop thinking about it — and return, as editors and readers, to doing what we’ve always done: judging writing on its intrinsic merits.

In other words, there’s no need to make up any new rules.

LLM slop-writing is not a distinct problem. The problem is bad writing, period.

Let me put this as bluntly as possible: if the writing is bad, who cares how it was produced? As editors (and readers), we’ve always rejected stale prose, trite observations, weak research, and poor diction regardless (I’m tempted to say “irregardless”) of the process behind them. That standard doesn’t need to change.

The rise of LLMs presents novelties only in that, first, some of the bad writing has a peculiarly specific “smell,” and second, there’s more of it than ever because that kind of bad writing is almost free to produce.

But the failure mode is identical: shitty writing. And the editorial response should be too: reject it.

If the writing is bad — padded with the characteristic not-this-but-that constructions, stuffed with plausible-sounding but thin research, beset by somethings-off-here metaphors, and festooned with M-dashes like this one → — then reject it. For exactly the reasons we’ve always rejected cheesy writing when it came from humans.

As I say, there’s no need to make up any new rules.

So far, I doubt I’m really going to have lost too many readers, even the antiquarian defenders of the old ways. So here’s the controversial part of my argument: it cuts both ways. If the writing is original, precise, delightful, and droll, then who cares how it was produced?

Here’s the point: what matters is quality, not provenance. This has always been true of the writer’s craft. We don’t ask whether a novelist used index cards or Scrivener, whether a journalist dictated notes or typed them, whether a poet drafted by hand or on a phone. What we do is read the work and judge whether it is good.

This approach has a practical virtue: it aligns incentives correctly. Writers will still need a style that isn’t clichéd or generic. If they use AI to generate research, brainstorm examples, produce structural drafts, or accelerate revision, but then bring taste and judgment to bear by introducing a distinctive voice, cutting what’s weak, and sharpening what remains, such that the output is excellent, then great! If on the other hand they attempt to outsource not just the typey-typey labor but actual discernment and style, it’s extremely likely to show. The “LLM-prose tell” is a quality signal, and editors already know how to read quality signals, and what to do when they see low quality.

The writers who will thrive in this environment are the ones who treat AI as what it actually is: a powerful but stylistically generic tool that at least for now requires human judgment and polish to yield something worth reading. Those who simply outsource their prose and hope no one notices will produce work that reads exactly like that. Which we as editors will reject.

For all intensive purposes, it’s a doggity-dog world out there, and there’s no need to make LLMs an escape goat!