Let us make one crucial point clear straight away: there is currently no evidence to suggest, as an established fact, that Thélyson Orélien wrote his novel using artificial intelligence. The author categorically denies this. He claims to have started the book in autumn 2017 and completed a first draft in 2019 – in other words, before the current wave of generative AI tools became widely known. Grasset has publicly reaffirmed its confidence in him.
However, the suspicions have not entirely come out of the blue. Several passages submitted to Pangram have been flagged as likely to have been generated by AI, and an expert interviewed by *Le Nouvel Obs* considers certain clues to be troubling. But a detector analyses the statistical characteristics of a text: it does not physically reconstruct the process by which it was created. On its own, it cannot show who typed a particular sentence, what versions preceded the final text, or how it was revised.
But let us suppose, purely for the sake of argument, that Orélien did indeed use AI. So what?
This is perhaps where this controversy becomes far more interesting than the petty technological tribunal to which it gives rise. Since when have we judged a work solely on the basis of the tool used in its creation? A writer uses a spell-checker, a dictionary, word-processing software, feedback from their editor, and sometimes from a proofreader, a research assistant or trusted readers. None of these intermediaries alone is sufficient to determine the literary value of the result.
Artificial intelligence obviously raises new questions, particularly those of originality, attribution, copyright and transparency. Having a machine generate an entire novel and then publicly claiming to have written it alone would not be the same as asking an AI to correct, rephrase, research, suggest alternatives or help unblock a passage. Lumping all these practices together under the term ‘written with AI’ has virtually no meaning anymore.
Above all, literature should perhaps return to its fundamental question: what effect does a book have on the reader?
A novel either moves us deeply or leaves us cold. It makes us think or bores us. Its language may or may not possess a distinctive quality. Its characters may haunt us or vanish as soon as the book is closed. Its message may strike us as powerful, banal, unsettling or illuminating. This is what literary criticism should examine first and foremost.
The case of *C’était ça ou mourir* makes the debate all the more fascinating. Even before the controversy erupted, the novel had already won over readers, booksellers and various judging panels. The Fnac Novel Prize is awarded based on votes from hundreds of booksellers and readers, and the book had made it onto the shortlists for several major autumn literary prizes.
These readers therefore felt something when reading the text. A genuine emotional response to a work—even one potentially aided by a machine—would still be a genuine emotional response. The reader’s experience does not become invalid in hindsight simply because software may have been involved at some stage of the production process.
This is, in fact, an old story in the history of art. Photography was long accused of killing painting. Cinema was viewed as a mechanical attraction before being recognised as an art form. Digital technology has provoked the same accusations. Photoshop was said to have destroyed photography, synthesisers were said to have distorted music, and digital effects were said to have replaced cinema. With every technical revolution, the same confusion arises between the tool and the work.
Yet a camera does not make a photographer. A camera does not make a filmmaker. A piano does not make a composer. And ChatGPT or Claude do not automatically make a writer.
You can ask an artificial intelligence to produce a hundred novels: this in no way guarantees that even one will make anyone want to turn the page. The artistic act also lies in choice, intention, editing, omission, the direction given to the material and, ultimately, in the relationship created with the recipient of the work.
This does not mean that all requirements for transparency should be dispensed with. When a prize explicitly recognises an individual’s written work, knowing what role AI actually played may become a legitimate question. Publishers and judging panels will probably need to establish much clearer rules: linguistic assistance, research, partial rewriting, the generation of individual passages, or the mass generation of a manuscript are not equivalent practices.
But let us turn this question into an intellectual debate rather than a witch-hunt.
Otherwise, we risk entering a rather absurd era where every metaphor that is too perfect, every sentence structure that is too regular and every sentence that flows too smoothly will come under suspicion. Will writers soon have to provide their drafts, their word-processing history and the various versions of each chapter to prove that they are indeed human?
The greatest irony would ultimately be if this obsession with detection ended up standardising literature even more than artificial intelligence itself.
Whether or not Thélyson Orélien used AI remains, for now, a matter of disputed fact. But the controversy raises a much broader question, one that now concerns all creative work: should we judge a work on the basis of the tools used to produce it, or on the basis of what it offers?
For us, the answer is quite simple. Let’s look at the book. Let’s read it. Let’s discuss its language, its weaknesses, its flashes of brilliance, what it tells us and what it evokes.
And if one day a writer manages to move us deeply by making clever use of artificial intelligence, the real critical question may not be which key they pressed.
It will be to understand why, despite everything, we were so deeply moved.
