A Prize Winner Was Accused of Using AI, So They Asked a Chatbot to Judge

Posted May 26, 2026

Here is a sentence that should stop every writer, editor, and contest judge in their tracks. After a short story won a regional 2026 Commonwealth Short Story Prize and was then accused of being written with artificial intelligence, the publisher's way of investigating the accusation was to ask an AI chatbot whether AI had written it. Read that again. The proposed cure for the AI problem was more AI. The arsonist was handed the fire investigation.

The story is "The Serpent in the Grove" by Jamir Nazir, a writer from Trinidad and Tobago who was named one of five regional winners of the prestigious prize, administered by the London-based Commonwealth Foundation and announced on May 14. One judge praised the language as sublime, precise yet richly evocative, conjuring vivid, lush imagery with remarkable economy. That praise is exactly what lit the fuse. The prose was so polished that readers online began asking whether a human had really written it, and the suspicion spread fast. The overall winner is not due to be announced until June, which means this cloud is now hanging over the entire prize.

The Detail That Belongs On This Site

We do not cover literary prizes here. We cover the ways people are misusing AI and getting burned, and this story qualifies on a technicality that is the whole point. When the controversy went viral, the publisher issued a statement noting that it had asked Claude, an AI chatbot, whether artificial intelligence was used to create the story. That is the part that should set off alarms in every newsroom and every contest committee on earth. A chatbot is not a forensic tool. It cannot reliably tell you whether text was machine-generated, and it will answer your question with the same fluent confidence whether it is right, wrong, or guessing.

This is the exact failure mode we document over and over. A large language model produces output that sounds authoritative regardless of whether it has any basis in fact. Ask it whether a passage is AI-written and it will give you a clean, plausible verdict, complete with reasoning that reads like analysis. But that verdict is not a measurement. It is a prediction of what an answer to your question tends to look like. Treating it as evidence about a real person's authorship is not investigation, it is laundering a guess through a confident interface.

AI Detectors Were Already A Disaster

Even the purpose-built tools do not work. The entire category of AI-writing detectors has a documented history of false positives, flagging human writing as machine-made and clearing machine text as human. They are notoriously biased against non-native English writers, whose more formal or carefully constructed prose trips the same signals these tools associate with generated text. Sit with that in the context of this case. A Caribbean writer's prize-winning English is described by a judge as unusually precise and economical, and that very precision becomes the reason to suspect a machine. If a dedicated detector would be on shaky ground here, a general-purpose chatbot asked an offhand yes-or-no question is not even in the same universe of reliability.

The damage runs in both directions. A false accusation can end a writer's reputation over a story they genuinely wrote, with no recourse and no real evidence, just a vibe and a chatbot's say-so. And a false clearance lets actual AI-generated work collect a prize while everyone points at the machine that supposedly checked it. Either way, the chatbot becomes a way to avoid the hard human work of actually adjudicating the question, and the institution gets to say it did something technical and modern while having proven nothing at all.

The Real Problem Is Procedural, Not Technological

What makes this a genuine mess is that there may be a real question buried under it. Maybe AI was involved in some of these entries. Maybe it was not. The point is that the moment you reach for a chatbot to settle it, you have made the answer unknowable, because you have introduced a tool that produces confident output untethered from truth. The accusation and the supposed verification now carry exactly the same evidentiary weight, which is to say none. We have written about how language models invent facts with total fluency in our coverage of lawyers getting sanctioned for AI-fabricated citations, and the parallel is brutal. The lawyers trusted a chatbot to produce real cases and it produced fiction. A contest trusting a chatbot to produce a real authorship verdict is making the identical bet with someone's career as the stake.

It is also part of a broader panic we have tracked in the creative world, from the literary scandal around a Nobel laureate's AI-touched novel to the steady drip of accusations chronicled in our running file on AI controversies piling up through 2026. The common thread is institutions reacting to AI anxiety by deploying more AI, faster, with less scrutiny, as if the technology that created the uncertainty can also resolve it. It cannot. That is the disaster.

What A Sane Process Would Look Like

If a prize committee genuinely suspects an entry was machine-assisted, the answer is human and procedural, not a chatbot prompt. Ask the writer about their drafting process. Request earlier drafts, notes, and revision history. Look at the rest of their body of work for consistency of voice. Have experienced editors read the piece with the question in mind. None of that is foolproof, but all of it keeps a human being accountable for a human judgment about another human's work. The chatbot shortcut does the opposite. It hands the most consequential decision to the least accountable participant in the room, a system that cannot be cross-examined and does not know what it does not know.

The overall Commonwealth winner will be named in June, and whatever happens, the lasting lesson is not about one short story. It is about a reflex. Faced with the possibility that a machine wrote something, an institution's first move was to ask a machine to rule on it. That reflex is going to spread, into hiring, into grading, into journalism, into every place where someone needs a quick verdict on whether text is real. And every time it does, a confident, unaccountable, frequently wrong system gets to decide a human being's fate while everyone pretends a question was answered. Asking the chatbot is not the safeguard. It is the failure.