An editor's own numbers closed a magazine's submission queue

The writer's problem

Clarkesworld is a short-fiction magazine that kept an always-open submission window, and its editor, Neil Clarke, had banned occasional plagiarists for years without much trouble. Late in 2022, wide access to chatbots gave a new group of people a faster version of the same trick: run a prompt through a chatbot, submit whatever came back, and hope to collect a payment before an editor noticed. Clarke's own post describes trying to keep reading real submissions while that volume grew past what one small magazine could absorb.

What the documents show

In the post, dated 15 February 2023, Clarke writes that spam submissions had reached 38 percent of everything Clarkesworld received that month, and that in the first fifteen days of February the magazine had already logged more submissions than in all of January. He declines to explain how he identifies the spam, saying that publishing his method would only help people evade it. A dated edit five days later records a fresh spike, over 50 before noon, and his decision to temporarily close the queue. The magazine's own site still lists Clarke as editor with new issues running, confirming the closure was a pause, not the end of the magazine.

The editorial choice

Clarke weighed automated detection tools against their cost and error rate and judged them unaffordable for a market Clarkesworld's size, noting that one vendor sells a tool to writers for evading the very detection it also sells to editors. He considered blocking submissions from regions with unusually high fraud rates, moving to solicited-only submissions, or requiring more contact information, and treats each option as shifting cost onto legitimate new and international writers rather than solving the underlying problem. This is an editorial trade-off Clarke names explicitly, not a solution he claims to have found.

What stays with the author

Nothing in the post changes what a legitimate submission has always needed: a story an editor wants to read, attributable to a specific writer who can answer questions about it. Clarke is explicit that his numbers describe his own operation, gathered by an undisclosed method, and that he cannot say how representative Clarkesworld's experience is of short fiction generally, only that other editors he contacted reported something similar.

• Does a single editor's closed queue describe a market-wide trend, or one magazine's specific conditions that month?
• What detection cost can a small editorial operation actually absorb without passing it on to writers?
• Who bears the burden when a market tightens its front door: the spammer it was built to stop, or the new writer who looks unfamiliar to a screening rule?

The 2023 closure is often cited as an early, concrete sign of what generative AI does to small publishing operations, but read narrowly, it is the record of one editor's triage under a documented spike, not a measurement of the wider field.