When Commonwealth Short Story Prize winners were accused of using AI, someone asked editors what to do about it. One expert said they should learn to spot AI writing themselves. Another said "absolutely not" — and offered nothing in return.
That exchange stuck with me. Because "absolutely not" is a position, not a policy.
Banning AI on paper is easy. Enforcing it is nearly impossible. The author says "this is my work, prove otherwise." Most publications won't pay for detection tools — that costs money. So they may end up awarding AI-written work without ever knowing it. Authors aren't villains; they're just people. If no one's checking, they'll use what helps them.
Some publishers are trying to take this seriously. Microcosm Publishing runs submissions through Pangram. The tool isn't perfect, but it flags suspicious work for human review — which is more honest than writing "AI prohibited" in the guidelines and hoping for the best. Another option is third-party certification from the author. But that shifts the cost onto them. And something like the Authors Guild's honor-system version doesn't actually prove anything.
Honestly, I think most of these anti-AI policies will quietly disappear. Detection tools can't keep up with the models. And where exactly is the line — is 30% AI assistance acceptable? 60%? Who decides?
It may just be simpler to judge the writing. Bad? Reject it. Good? Take it. Stop spending resources playing AI police.
At KF Books we landed somewhere different: authors tick a box in the submission form. No interrogation after that.