In debates about artificial intelligence and writing, one question dominates: ‘Was this written with AI?‘ In some settings, that question is important, especially when academic integrity, competition rules, or authorship claims are at stake. But as a general approach to reading and assessing texts, it is surprisingly limited. What ultimately matters is whether a piece of writing is thoughtful, accurate, and worth engaging with, and not whether a language model touched it somewhere in the process. This article focuses foremost on creative writing like literature, fiction, and poetry, where questions of authorship, originality, and artistic voice are central. Writing is a vast field that includes journalism, blogging, technical writing, diplomacy, legal work, and marketing, among many others. In each of these areas, AI is viewed differently, with varying degrees of judgement and different ethical concerns. The opinions and cautions I raise here apply primarily to the context of literature, though some principles may resonate more broadly. The anxiety around AI in writing often reminds me of earlier moments when new AI tools arrived in creative fields. We have been here before: arguments that animation should only be hand-drawn, that digital photography is less ‘authentic’ than film, or that spellcheck and grammar tools somehow weaken language. Tools do change practice, sometimes profoundly. But their presence does not automatically diminish the result. The key questions are how they are used, what human judgement guides them, and whether the final work has substance. Despite this, a kind of informal ‘AI police’ is emerging around writing. Institutions experiment with AI-detection systems; social media users scrutinise sentences for signs of machine-generated style; accusations of ‘this is AI‘ are thrown around with little context. There is a legitimate concern here: much low-value, generic AI-generated text is flooding the internet. But the growing reflex to treat any AI involvement as automatically wrong risks turning a complex challenge into an oversimplified one. It focuses attention on the tool rather than on the quality and integrity of the work.
The recent controversy around the Commonwealth Short Story Prize illustrates why the conversation has become so charged. In 2026, one of the regional winners, Trinidadian writer Jamir Nazir, was celebrated for his story ‘The Serpent in the Grove‘. Within days, critics began alleging that the story bore features typical of AI-generated text and circulated screenshots from Pangram, an AI-detection tool, suggesting that the piece was likely machine-written in its entirety. The publisher, Granta, publicly acknowledged the debate and even asked an AI system whether another AI had written the story, receiving a characteristically inconclusive answer. The Commonwealth Foundation has since been forced to investigate, with the uncomfortable admission that it may never be possible to know for sure.
This case raises real concerns. If a text submitted for a prestigious prize is largely or wholly produced by an AI system, then the issue is not only stylistic. It goes to the heart of what such prizes are meant to recognise: human creativity, craft, and labour. When the human role shrinks to prompting, selecting, and minor editing, we are no longer discussing AI as a tool within a human-led process. We are close to presenting a system’s output as personal achievement. In that narrow context of competitions and other settings where the goal is to evaluate original human work, questions about ‘how much AI?’ are justified and necessary. However, it would be a mistake to generalise from this kind of extreme case to a blanket suspicion of any AI use in writing. A more constructive approach is to distinguish between AI as a primary author and AI as an auxiliary instrument. That distinction becomes much clearer when we look at a second recent controversy, this time involving a writer whose authorship is beyond dispute. In May 2026, Polish novelist Olga Tokarczuk, winner of the 2018 Nobel Prize in Literature, sparked a debate after publicly discussing how she uses AI in her practice. She described purchasing access to an advanced language model and asking it to analyse or expand on her ideas, including details such as what music fictional characters might have danced to in past decades. Tokarczuk spoke of being ‘shocked’ by how broadly the tool could deepen her creative thinking, even joking that she addresses it affectionately: ‘Darling, how could we develop this beautifully?’.
Some readers and commentators interpreted her remarks as an admission that AI had helped write her forthcoming novel. Tokarczuk responded with a detailed statement clarifying her position: she treats AI as a tool for faster preliminary research and inspiration, verifies all the information it provides, and does not use it to write the text of her books. ‘None of my texts, including the novel that will appear in Polish this fall, has been written with the help of artificial intelligence, ‘ she emphasised, drawing a clear line between idea development and authorship. Taken together, the Commonwealth case and Tokarczuk’s comments point to the distinction we urgently need to make. In one pattern, AI appears to supply much of the narrative substance and phrasing, with the human acting mainly as a curator. In the other, a human author brings the concept, perspective, and voice, while using AI as an enhanced notebook, research assistant, or sparring partner. Both involve AI, but they do not involve it in the same way, nor do they raise the same ethical questions. From this perspective, the discussion shifts toward authorship. Who is doing the thinking here? If a person starts with a genuine idea, uses AI to explore angles, gather references, or generate a rough draft, and then substantially rewrites, edits, and owns the final text, there is a strong case for viewing the work as human-authored. The tool accelerates and expands what is possible, but it does not replace human judgement. By contrast, when a model produces a near-finished story or essay from a short prompt and the human contribution is minimal, the claim to authorship becomes much harder to defend, especially in competitive or academic environments. Authorship has never been defined solely by who physically produces the words. Writers draw on conversations, books, editors, archives, research assistants, and countless external influences. What ultimately makes a work their own is not that every sentence emerged in isolation, but that they exercised judgement over the ideas, structure, arguments, and final form. Therefore, the challenge posed by AI becomes one of authority over how direction is established, how meaning is interpreted, and who carries responsibility for the final work. This argument builds on earlier reflections on AI and creativity, where the central issue was not whether tools are used, but whether they displace human intent and control. This is why an overemphasis on AI detection can be counterproductive. If every trace of AI is treated as problematic, writers may feel pressure to conceal how they work rather than to be open about their tools. Similar tensions have accompanied earlier technological shifts, though AI introduces challenges of its own. Detection also risks becoming a substitute for evaluation. Instead of discussing whether a text is insightful, original, or persuasive, attention shifts toward proving how it was produced. In the long run, institutions are likely to benefit more from clear expectations about acceptable use than from an atmosphere of distrust. Instead of asking ‘Was this AI?’ first, readers, editors, and educators should start with questions about the content itself. Is the piece informed, coherent, and honest? Does it show a human perspective, or is it just rephrasing common ideas? Would the argument still make sense if you stripped away the polished phrasing? Only after answering those questions does it make sense to ask how AI was used, whether that use fits the context, and whether it is disclosed where it matters. This is important in spaces like diplomacy, education, and policy, where writing is both a tool and a record of how decisions are reasoned. If AI can help more people articulate complex ideas, explore scenarios, or organise their thoughts, that is an opportunity. But it remains essential that institutions are clear about when human authorship is being evaluated and what forms of machine assistance are compatible with that goal. Transparent norms and shared expectations are more promising than a culture of suspicion.
The controversies around the Commonwealth Short Story Prize and Olga Tokarczuk show two ends of a new spectrum. At one end, we rightly worry about cases where AI may have replaced much human creative effort. On the other hand, we see an established author using AI as a research and brainstorming aid while insisting that the writing itself remains her own. If we treat these situations as identical simply because ‘AI was involved’, we lose sight of the real questions about authorship, honesty, and value. While this discussion focuses on creative writing, the broader challenge of distinguishing human authorship from machine output applies across journalism, technical writing, legal documents, diplomacy, and many other forms of writing. Each field will need its own standards and transparency norms. AI will remain part of the writing process, just as earlier technologies have done. The task ahead is not to exclude it entirely, nor to accept it uncritically, but to refine our understanding of what it means to write and to take responsibility for what we publish, in an era where human and machine capabilities are increasingly intertwined. Author: Slobodan Kovrlija
The commonwealth short story prize scandal
AI as research assistant, not author
Who is doing the thinking?
Changing the questions we ask
Beyond simple categories
