When Philippe Jaenada revealed that he had used ChatGPT while writing his prize-winning novel, it reignited an uncomfortable debate about artificial intelligence and literature. But where exactly should we draw the line?
A French novelist has just won one of the most prestigious literary prizes in the world. Nothing particularly unusual about that. Except that, the following day, he appeared on French radio and explained that he had used ChatGPT while writing the winning novel.
Philippe Jaenada, 62, was awarded the 2026 Prix Goncourt on 3 November for L'Inconnue du quai de Javel, a novel investigating the unsolved murder of a young woman in Paris in 1949. It is his thirteenth novel, a substantial work of more than 500 pages, and the latest in a series of books in which he has explored forgotten crimes and the lives of those caught up in them.
A day after receiving the award, Jaenada told France Inter that he had used artificial intelligence on three occasions while writing the book. On one occasion, he had deliberately inserted a sentence generated by ChatGPT as a game for readers. On another, he had included material to poke fun at the chatbot's tendency to produce absurd answers. He had also used AI to help with technical research.
Predictably, the reaction was swift. Some readers expressed disappointment that a prize-winning novelist had turned to artificial intelligence at all. Others wondered what the fuss was about. If an author can use Google, a spellchecker or an editor, why should ChatGPT be treated differently?
But another, more uncomfortable dimension to the story emerged. Just weeks earlier, another writer had been removed from the Goncourt shortlist amid allegations that his novel had been substantially written using AI.
Suddenly, the question was no longer simply whether artificial intelligence has a place in literature. It was about who gets to use it, how much they can use, and who decides when a writer has crossed the line.
A literary prize with a problem
The Prix Goncourt occupies an extraordinary position in French literary culture. First awarded in 1903, it is arguably France's most prestigious literary honour, capable of transforming an author's career and turning a novel into a bestseller.
Previous winners include Marcel Proust, Simone de Beauvoir and Marguerite Duras. The prize carries enormous cultural authority, even if its nominal financial reward is just €10.
Jaenada's winning novel reconstructs the life and death of Louise Cansot, a young model murdered in Paris in 1949. Drawing on archives, court documents and his own investigations, he attempts to understand not simply who killed her, but who she was.
It is exactly the sort of ambitious, meticulously researched literary work that a prize like the Goncourt exists to celebrate, which is what makes the AI revelation interesting.
Except that revelation is not quite the right word. As the French newspaper 20 Minutes pointed out, Jaenada had already described his use of artificial intelligence within the novel itself. Pages 237 to 240 discuss his exchanges with ChatGPT and Claude, including their attempts to answer a technical question about whether paving stones or hair dries more quickly. He even tells readers that he has left an AI-generated sentence somewhere in the manuscript and challenges them to find it.
In other words, this wasn't a novelist secretly outsourcing his prose and getting caught. It was a writer openly experimenting with a technology, making that experiment part of the book.
That distinction matters. Jaenada's position is essentially pragmatic. AI exists, it isn't going away, and writers should learn to use it without allowing it to replace them. He sees it as something to understand and control, rather than fear or ignore.
It is a reasonable position. But the controversy surrounding the Goncourt suggests that the literary establishment has yet to work out what that means in practice.
When does using AI become cheating?
Earlier in the autumn, Canadian-Haitian author Thélyson Orélien found himself at the centre of a very different AI controversy.
His novel C'était ça ou mourir had attracted considerable attention and was included in the Goncourt's initial selection. But allegations emerged that substantial portions of the book had been generated using artificial intelligence.
Orélien denied using AI to write the novel, saying he had used it for research. His publisher, Éditions du Boréal, later said its own investigation uncovered evidence of AI use and ended its relationship with him. The novel was removed from the Goncourt selection.
The cases are not necessarily equivalent. There is an obvious difference between deliberately including one AI-generated sentence in a novel and having significant passages of prose produced by a chatbot.
But the contrast raises an important question. Where is the boundary? If one sentence is acceptable, what about a paragraph? What about a page? Could a novelist use AI to rewrite an awkward passage, suggest dialogue or restructure a chapter? At what point does the technology stop being an assistant and start becoming a co-author?
These are questions the publishing industry is going to have to answer, and rather sooner than it might like. AI's growing dominance across so many walks of life makes this a pressing issue.
There is also the question of consistency. The contrasting treatment of Jaenada and Orélien has prompted accusations of unequal standards, including concerns about racial bias. Those concerns deserve examination, although the reported differences in the extent of AI use are significant and shouldn't be ignored.
The answer cannot simply be that established writers are allowed to experiment while emerging writers are expected to prove their purity. That makes no sense, and is far from fair.
If literary prizes are going to establish rules about AI, those rules need to be clear, proportionate and applied equally.
Writers have always used tools
Part of the difficulty is that writing has never been an entirely solitary activity. We like to imagine the novelist sitting alone at a desk, wrestling with language, producing sentences through some mysterious combination of inspiration, experience and hard work. It is a romantic image, and one that writers themselves have done plenty to encourage.
But novels are rarely created in complete isolation. Writers use dictionaries, thesauruses, encyclopaedias, research assistants and editors. They discuss plot problems with friends, attend writing workshops and seek feedback from other authors. A good editor might suggest removing an entire subplot, changing a character's motivation or rewriting an ending.
Nobody seriously argues that a novelist ceases to be the author of a book because an editor has helped improve it. The obvious counterargument is that ChatGPT is fundamentally different. A dictionary supplies definitions. An editor offers advice. Generative AI can produce finished prose.
That difference is important. As a novelist, I think there is something fundamentally different about asking a machine to help you research a subject and asking it to write your sentences for you.
The first extends the research process. The second risks handing over the very thing that makes writing an act of individual expression.
But even here, the distinction is not always straightforward. Suppose you're writing a historical novel and need to know how a particular piece of machinery worked in 1915. You could spend an afternoon searching through technical manuals, ask an expert or consult an AI chatbot. As long as you verify the information, does the method really matter?
Now suppose you've written a paragraph describing that machinery and ask ChatGPT to make it clearer. Is that editing, or is it writing?
What if you ask it to produce five alternatives and choose the one you like best?
The technology makes it remarkably easy to move from assistance to substitution without necessarily noticing when you've crossed the boundary. That, I suspect, is where the real danger lies.
The problem isn't AI. It's losing the writer.
I have written before about the strange conventions of contemporary literary fiction, particularly the tendency of some writers to abandon traditional grammar or punctuation in ways that can feel more performative than purposeful.
But whatever we think of those stylistic choices, they are choices. They reflect an author's understanding of language, their aesthetic preferences and the effect they want to create.
The same is true of awkward sentences, peculiar rhythms, and the occasional imperfections that make a writer's voice recognisable. Think about the long, intricate sentences of Henry James, the spare precision of Ernest Hemingway or the extraordinary verbal energy of Martin Amis. Their prose is inseparable from the way they see the world.
That is what readers respond to. Not simply the information being communicated, but the particular consciousness communicating it.
Generative AI can imitate many of these characteristics. Ask it to produce a paragraph in the style of Hemingway, and it will probably give you something short, declarative and stripped of ornament. Ask for something resembling James, and it may produce a sentence that requires a map and a packed lunch to navigate.
However, imitation is not the same as artistic intention. Hemingway's restraint wasn't simply a collection of stylistic rules. It emerged from his understanding of what should remain unsaid. James's elaborate syntax reflected his fascination with perception, psychology and the complicated ways people misunderstand one another.
A chatbot can reproduce some of the visible features of those styles. It cannot draw on the lived experience that produced them. There is also a practical problem. AI-generated prose often gravitates towards the plausible and familiar. It is very good at producing sentences that sound like writing. It is less reliable at producing sentences that sound unmistakably like one particular writer.
For commercial copy, that may be perfectly adequate. For a novel, it seems a rather depressing ambition. Literature should do more than sound competent.
Does the reader have a right to know?
One of the more interesting aspects of Jaenada's case is that he appears to have treated AI as part of his novel's subject rather than simply a convenient shortcut.
He incorporates his exchanges with chatbots into the narrative. He comments on their shortcomings and makes a game of the AI-generated sentence he has hidden in the text.
That feels rather different from an author secretly generating chapters and presenting them as entirely their own work.
It also suggests a possible way forward: transparency. Readers do not necessarily need a detailed account of every tool an author used during the writing process. I don't particularly care which word processor a novelist prefers, whether they use Scrivener or Word, or how many times they consult a thesaurus.
But if AI has generated substantial portions of a novel, I think readers have a legitimate interest in knowing. Not because the book is automatically worthless, but because authorship is part of what we believe we are buying.
When I pick up a novel by Donna Tartt, Ian McEwan or Jonathan Franzen, I am not simply purchasing a sequence of words arranged into a story. I am interested in their imagination, their observations and the particular way they transform experience into fiction.
If significant parts of that experience have been delegated to a machine, it changes the relationship between writer and reader. It doesn't necessarily mean the resulting book cannot be entertaining, moving or even artistically interesting. But it does mean that the nature of the work has changed, and readers should be allowed to make their own judgements about that.
The next literary scandal is already being written
What makes the Goncourt controversy particularly revealing is that it has exposed the absence of any settled understanding about AI and literary authorship.
Jaenada's limited and openly acknowledged use of AI seems to me entirely defensible. The fact that he incorporated the technology into his novel as an object of curiosity and humour makes the experiment more interesting, not less.
The allegations surrounding Orélien's novel raise a different set of questions about the extent of machine-generated writing and the responsibilities of authors and publishers. Those questions deserve proper scrutiny rather than a rush to judgement based on suspicion alone.
But neither case offers a simple rule to settle the argument, and the problem will only become more complicated.
As AI tools become integrated into word processors, editing software and research platforms, the distinction between writing with AI and writing without it will become increasingly difficult to maintain.
We may eventually reach a point where asking whether a novelist has used artificial intelligence is about as meaningful as asking whether they have used the internet.
The more useful question will be what they used it for. Did it help them discover something? Did it challenge their assumptions? Did it make their research more efficient? Or did it do the imaginative work that we expect a novelist to undertake?
For me, that is the distinction worth defending. I don't particularly care whether a writer uses AI to research how a nineteenth-century railway engine works. I care very much if they use it to manufacture their characters' thoughts, emotions, and language.
Because the value of literature lies not simply in the finished sentences, but in the human imagination behind them, and if we lose that, we may find ourselves surrounded by more novels than ever before, with increasingly little worth reading.
Perhaps the most revealing thing about Philippe Jaenada's experiment is that nobody yet knows which sentence in his prize-winning novel ChatGPT wrote.
The temptation, of course, is to find it, but perhaps the more interesting question is whether we would recognise it if we did.