When writing needs an alibi

Literature has survived censorship, plagiarism, ghostwriters, bad editors, worse reviewers, and the occasional panic that a book might be dangerous enough to ban. Its latest problem is apparently the suspiciously well-behaved sentence. Too smooth, too balanced, one metaphor behaving a little too well, and somewhere a tab opens. AI detector. Paste. Scan. Apparently “this sounds like AI” now sits somewhere between literary criticism and forensic evidence. Somewhere between the blinking cursor and the published page, authorship has acquired a new anxiety, not simply whether writers are using machines, but whether anyone can prove when they are.

Tap here to add The Daily Star as a trusted source

I do not think the answer is to ban every machine from the writer’s desk, nor do I think calling AI “just another tool” settles much. A dictionary does not offer to invent my childhood. A search engine does not usually ask whether I would like the ending made more devastating. Generative AI can move from servant to collaborator without much ceremony.

The 2026 Commonwealth Short Story Prize turned that anxiety into a public spectacle. Suspicion first gathered around Caribbean regional winner Jamir Nazir’s “The Serpent in the Grove” (Granta, 2026), where readers pointed to repeated “not X, but Y” constructions, clusters of three, heavy metaphors, and other features they considered familiar from generative AI. Detection software was then brought into the discussion, and allegations spread to two other regional winners: John Edward DeMicoli’s The Bastion’s Shadow (Granta, 2026) and Sharon Aruparayil’s Mehendi Nights (Granta, 2026). Suddenly literary criticism had developed the atmosphere of a crime laboratory. Perhaps future writers should leave one terrible metaphor in every story as proof of life?

The difficulty is that suspicion, detection and proof are not the same thing. Nazir denied using AI and explained that chronic health conditions had led him to compose through speech-to-text on his phone, followed by editing. The Commonwealth Foundation did not let an AI-detector percentage settle the matter. Instead, it looked at drafts, time-stamped documents, notes, and how the writers said the stories had actually been written. On June 22, the Foundation announced that it was satisfied AI had not been used to write the winning stories; Nazir later went on to win the overall prize. That conclusion does not magically settle every philosophical question raised by the controversy, but it does establish one factual boundary: accusation was not the institutional finding.

Granta reportedly went one step further and asked Claude, an AI chatbot developed by the company Anthropic, to assess Nazir’s story, producing an inconclusive answer. We are therefore approaching the magnificent technological achievement of asking one machine whether another machine wrote a paragraph while the human author waits outside the courtroom. A detector can become a dictator remarkably quickly when a percentage starts sounding like a verdict, can’t it?

Olga Tokarczuk, the 2018 Nobel Prize in Literature winner, spoke about AI at Impact’26, a large international business, technology and ideas conference held in Poznań, Poland, on May 13 and 14, 2026, where speakers from culture, politics, science, and business gathered to discuss subjects including AI and digital change. During her talk, she described experimenting with an advanced language model and said she sometimes gives it an idea and asks, “Darling, how could we develop this beautifully?” She also gave a much simpler example: asking the model what songs her characters might have danced to decades earlier. The answers were useful, though not always correct, and Tokarczuk stressed that anything supplied by AI still had to be checked. She also spoke about writers possibly developing a closer working relationship with AI in the future, while saying that present-day chatbots still cannot match the kind of literary language she values.

After those remarks began circulating, Tokarczuk clarified what she meant by using AI. She said her forthcoming novel had not been written with AI, and described the technology’s role as preliminary research, documentation and fact-checking, with the information independently verified. She also insisted that none of her literary texts had been written by AI. Fair enough. Still, her earlier question, “How could we develop this beautifully?”, shows why the phrase using AI can become confusing very quickly. Looking up a period-appropriate song is one thing; asking a machine to help develop an idea feels a little closer to the writing desk. Tokarczuk even added that although dreams sometimes enter her work, those dreams are still her own. Apparently, in 2026, even the subconscious may need an authorship declaration.

The Commonwealth controversy therefore looks different to me after Tokarczuk. “Did they use AI?” is too small a question. Did they research with it? Brainstorm with it? Revise through it? Generate language with it? What did the competition rules actually prohibit? And what evidence should be enough to publicly question a writer’s authorship? The Foundation cleared the regional winners after examining process evidence rather than relying on detector scores. Granta nevertheless ended its publishing partnership with the prize, saying it would no longer participate in external publishing arrangements where it lacked editorial control. The Foundation has since postponed the opening of the 2027 prize until January 2027 while it reviews rules, submissions and procedures. Apparently, the easiest way to modernise literary administration was to make everybody suddenly wish manuscripts still arrived in envelopes?

I do not think the answer is to ban every machine from the writer’s desk, nor do I think calling AI “just another tool” settles much. A dictionary does not offer to invent my childhood. A search engine does not usually ask whether I would like the ending made more devastating. Generative AI can move from servant to collaborator without much ceremony, and that quiet movement is what interests me. Perhaps authorship will eventually depend less on whether AI entered the room than on what the writer refused to hand over once it arrived.

The Commonwealth incident shows how dangerous it is to mistake stylistic suspicion for proof. Tokarczuk shows how inadequate the simple confession “I used AI” has become. Somewhere between research and surrender is a boundary we have not named properly yet.

The problem is that writing has never been the completely private activity we sometimes pretend it is. Writers borrow; not sentences ideally, but rhythms, structures, memories, overheard remarks, myths, newspaper stories and things friends had said over dinner. They use editors, dictionaries, workshops, translators, librarians and search engines. So “a real writer never receives assistance” is not a serious standard. If asking a librarian for a historical newspaper is acceptable and asking an AI for the same fact is suspicious, perhaps the machine itself is not the entire issue.

This is where my own comfort with AI becomes inconvenient. I can defend using it to locate information, organise research, test an argument, or point out something I have missed. I become less comfortable when it starts providing the thing I was supposed to discover by writing. Those annoying hours matter. Staring at the page, deleting a sentence, bringing it back, getting something wrong, following that mistake somewhere unexpected, and finally finding an image that was not there before. AI promises to make some of that easier.

Writers have complained about writer’s block for centuries, and now that there is a tool willing to help. But for me, apparently getting stuck has developed sentimental value. Inefficiency is not always a defect. Sometimes the wasted hour is where the sentence comes from.

Nayer Sultana connects bridges between poetry and prose. When she’s not being her lecturer-self at North South University, she’s stitching together a non-fiction memoir that refuses to stay in one genre. Her work wanders through belonging, becoming, and the shifting idea of home.