Speculation is quickly erupting that top literary prizes were handed to stories containing AI-written text.
The prestigious Commonwealth Short Story Prize, which comes with a winning $5000 (£3700) and publication in respected UK literary magazine Granta, is under fire for handing prizes to what many suspect is either fully or partly written by a large language model such as ChatGPT.
The regional winner for the Caribbean region, Jamir Nazir’s The Serpent in the Grove, was the main story to come under suspicion. Speculation began when a user on X, a researcher and visiting scholar of AI at George Mason University, no less, raised the alarm after winners were announced. The author of the story is reportedly a 61-year-old from Trinidad and Tobago with few publications to his name. A book of love poems from 2018 comes from the author, as does a book of motivational quotes from 2019, but there is little to suggest that the author is anywhere near the level of winning incredibly competitive and prestigious literary competitions.
There were stylistic tells that, taken together, indicated the use of a large language model. The now infamous ‘it’s not X, it’s Y’ sentence structure that AI has so gleefully adopted and can’t let go of saw use throughout the story, along with a myriad of other tells.
Some concluded, even, that the story was not written by a human because it lacked the soul of a human-written piece, AI being so unable to understand the rich meaning of its words that it often produces flowery prose without purpose, in turn meaning little. In the very best scenario for the parties involved, the Commonwealth Short Story Prize is rewarding bad writing, and Granta magazine is platforming it.
Read more:
OpenAI’s Sora was going to destroy Hollywood... until it came abruptly crashing down
Even the first sentence of the story contained a tell: “They say the grove still hums at noon”. With AI fiction writing, particularly when it looks to invoke a sense of place or atmosphere, a trope of things that ‘hum’ frequently occurs. It’s just one of the strange and quirky giveaways, but one that has started to catch out big publishers that have been accused of publishing works written by AI.
Individually, these things would not be a cause for concern: AI is generating its writing from millions and millions of words humans have produced, averaging out all the means of our fascinating stylistic tics. If humans tend to write lists of things in threes, then AI will tell itself to be over-reliant on that format in its quest to accurately replicate how a human writes.
In response, the Commonwealth Foundation seemed embarrassed and woefully underprepared for the onslaught of AI such a competition would attract. Granta, to their credit, left the story up for others to judge. It was not up to them what stories were the winners, only to publish them in the aftermath.
Read more:
Can we pass a law banning the sale of Highland cow AI art?
The story on their website now includes a disclaimer, which ends with: “The suggestion that writers have submitted material not authentically their own is a charge we take seriously, but until definite evidence comes to light we will keep these stories on our website.” Stories, because the winning entry from Canada and Europe, The Bastion’s Shadow by John Edward DeMicoli, is also being accused of being partially written by AI.
Publisher of Granta magazine, Sigrid Rausing, had a peculiar response to the fiasco: “We showed Claude.ai the story and asked whether it was AI-generated … the response was long, concluding that it was ‘almost certainly not produced unaided by a human’”. She said that Claude.ai ended on a ‘cautious note’, citing passages that could possibly be real.
AI detectors are notoriously unreliable, even returning that parts of the Declaration of Independence were written by AI. The extent to which the foundation assessed authenticity seems non-existent, despite their protests to the contrary. A more robust and considered submission process, particularly for a prestigious literary award with a cash prize, that doesn’t rely on flaky verification methods, would be a huge start.
Read more:
On reading Yukio Mishima, the writer who disembowelled himself
In 2026, some form of authenticity checking is essential. Literature competitions have struggled in the past few years, becoming inundated with bogus AI writing submissions and seeing the regular pool of genuine voices diluted. Many have stopped completely because the task of separating reality from the fiction is just too much, too impossible with the shoestrings most literary organisations operate under.
And if the Commonwealth Foundation wishes to trust their judges, as they have said, then they certainly need people on the decision panel who understand how large language models work and how they are used in producing texts of fiction. It’s a new challenge that has suddenly arisen, and one that can’t be ignored only to get worse and cause potential ruin. It is all great for a judging panel to have years of expertise and experience between them, but the influx of AI into the literary space calls for something different to counteract it.
Maybe this all could be a witch hunt, overly critical folk seeing things where there are none. Use coherent sentences and some em dashes online and it does not take long to be accused of using ChatGPT, after all. But if there is something to this accusation, and it has passed through the upper echelons of the prestigious literary world, it is a dire sign of things to come.
Derek McArthur is an artist and arts writer specialising in cinema and culture. He writes a weekly arts column for The Herald.
