Spare a thought for upcoming authors in race to curb AI slop

Anne Carson has been declared winner of the 2026 Nobel Prize in Literature. The announcement found the Canadian poet in Iceland, “listening to the wind”, according to reports. Her award citation praised an oeuvre that, through playful engagement with classical traditions, has created new forms for contemporary literature.

This is great! We may appreciate Edgar Allan Poe’s argument that a short story should be shaped towards a single effect, with every element contributing to it. We may also acknowledge the Russian Formalists, including Viktor Shklovsky, whose 1917 essay “Art as Technique” argued that art makes us see familiar things afresh, and Boris Eichenbaum, whose 1927 essay defended formal literary analysis. But new art forms must be embraced, not dismissed because they defy inherited definitions.

Carson’s books mingle poetry, prose, translation, criticism and drama. Autobiography of Red (1998), for instance, has been hailed for reimagining Greek myth in a novel-in-verse, while Nox (2010) combines elegy, translation and family history in an accordion-fold book. I confess that I have not read much of Carson’s work, but I am keen to do so now. Her Nobel is a timely reminder that literary innovation is not a threat to literature. It is part of its life.

My focus today, however, is how literary prizes have been affected by new technology, particularly generative artificial intelligence. Several recent controversies involving major literary prizes show why editors and organisers are grappling with the issue.

In Japan, Rie Kudan won the prestigious Akutagawa Prize in January 2024 for Tokyo-to Dojo-to, translated as Sympathy Tower Tokyo. She disclosed that around five per cent of the novel consisted of text generated by ChatGPT and used directly. The novel explores AI and the prize committee reportedly did not challenge the award after her disclosure. The case raised a question that remains unresolved. Just how much AI assistance is compatible with a work being judged as the author’s own?

In France, Thélyson Orélien’s novel C’était ça ou mourir (It Was That or Die) was removed from the Prix Goncourt selection in September following allegations of extensive AI use and plagiarism. The AI claim originated with an anonymous online account, while other detection tools reportedly assessed the novel as likely human-written. Orélien denied using AI. The case illustrates the danger of allowing disputed technical assessments to determine literary reputation.

The Commonwealth Short Story Prize also faced allegations against its 2026 regional-winning stories. After examining drafts and other evidence and speaking to the writers, the Commonwealth Foundation said in June that it was satisfied AI had not been used to write the winning stories. It acknowledged the limitations of detection tools and said it would strengthen its processes.

Now, Zimbabwean writer Simbarashe Kundizeza’s Caine Prize-winning story, The Blue Fighter, has become the subject of similar allegations. He has denied using AI and the prize committee is reviewing the claims. The case reinforces the need to distinguish suspicion from proof, particularly when a writer’s reputation is at stake.

We need clear policy guidelines for literary journals and awards. How far may writers go in using tools such as ChatGPT, Claude, Gemini and Microsoft Copilot, or writing assistants such as Grammarly? The last time I received the good news that two poems had been accepted by a leading literary journal in the region, whose policy excludes AI-generated content, the editors encouraged me to run them through Grammarly, a digital writing assistant that checks spelling and grammar and offers suggestions on clarity and style, before resubmission.

A spelling check is not the same as asking a machine to produce a poem. Yet some writing tools go beyond catching typos and recommend changes to wording, tone and structure. Such interventions can alter a writer’s expression. Journals need to state plainly what is allowed, whether proofreading tools must be disclosed and where assistance becomes authorship.

Technology itself is not the enemy. The concern is literary content generated by large language models, not every digital tool. AI detectors remain fallible and concerns have been raised about their treatment of writers whose language varieties are poorly represented in their training data. A detector’s verdict alone cannot establish who wrote a suspicious submission.

Meanwhile, the volume of submissions is putting pressure on journals and prize organisers. Some have restricted open submissions or introduced fees, while editors increasingly require assurances of originality and may ask shortlisted writers to explain their creative process or provide drafts. Assessing thousands of entries takes time and money. It is therefore understandable that introduction of submission fees may discourage mass submissions generated at little cost. But deterrence can also become a pretext for exclusion.

Spare a thought for the talented first-time writer who works on an empty belly and cannot afford an AI subscription, let alone a laptop or desktop computer. Some still write by hand and take their manuscripts to a cybercafé or typing bureau, pay to have them keyed in, then submit them to a journal or competition. These writers may be the least likely to use AI, yet they can be hit hardest by entry fees and extra verification requirements.

 They face the double jeopardy of having not made a name for themselves and must negotiate hurdles that were absent when many established writers began. They may also write in language varieties that AI detectors handle poorly, exposing them to unfair suspicion. A fee that appears modest to an established professional may be prohibitive to a newcomer.

As we devise measures to keep AI-generated slop out of literary competitions, we must not shut the door on the voices prizes are meant to discover. Keep entry routes free or affordable where possible, offer fee waivers, publish precise rules on permissible AI assistance and use human review and evidence of process before making accusations. And, by Jove, do not treat a detector’s score as a verdict.

Literature needs to defend human imagination, labour and experience. It must also protect the chance for an unknown writer to be read on the strength of the work. In the race to curb AI slop, the people we should not leave behind are the upcoming authors whose voices the literary world has yet to hear.