French literature is experiencing one of those controversies that leaves a bad taste in your mouth and, at the same time, forces you to confront an uncomfortable issue head-on. The novel C'était ça ou mourir , by the Canadian-Haitian writer Thélyson Orélien, has gone from being one of the phenomena of the literary revival to becoming the focus of an informal investigation into whether its text was written with the help of artificial intelligence. The accusation doesn't come from a court or a publisher, but from an anonymous account on X that submitted fragments of the work to a detection tool.
The book, also translated into Spanish as Era eso, morir (That's It, To Die ), had sold around 35.000 copies in five weeks and was considered a contender for major awards. Now, the question is no longer just who wrote it, but what exactly authorship means in an industry that is beginning to grapple with language models capable of mimicking human styles. Publishers, agents, and literary juries are watching the case closely because it foreshadows similar problems in other European markets.
A case opened in France: from anonymous complaints to literary prizes
The controversy began when the Balance ton Claude account published several analyses on X using Pangram, a program specializing in identifying patterns in AI-generated texts. According to these results, some passages of the novel had been produced by a machine, and the whole was "almost entirely" written in this way. The post garnered millions of views and brought attention to a book that, until then, had been receiving critical acclaim and strong sales.
French media outlets took up the challenge. Le Monde analyzed more than half of the novel using Pangram and concluded that over 95% of the text could have been generated by AI, with a high level of confidence. This finding was striking because Pangram is considered one of the most accurate AI detectors on the market. Even so, the very nature of these tools prevents us from speaking of conclusive proof: they don't certify authorship, they only indicate probabilities.
The book's publishing trajectory also explains the magnitude of the scandal. It sold some 35.000 copies in five weeks and was shortlisted for prestigious awards such as the Goncourt Prize, in addition to receiving other accolades in France. This commercial and critical success made suspicion about AI a much more uncomfortable issue for the industry.
In response to the accusations, Thélyson Orélien has denied using artificial intelligence to write the novel. He explained that the first draft was finished before ChatGPT's public outcry and that the repetitions, imagery, and rhythms in the text reflect his own way of expressing himself, rooted in his Haitian and Caribbean heritage. Both he and his publishers are preparing documentation on the creative process to try to prove that the work is his own.
Detection tools: clues, not definitive proof
The case has put Pangram and AI detectors in general under scrutiny . The tool boasts a very low false positive rate and has been cited in academic studies, but experts point out that no system can definitively determine who wrote a text. Various media outlets and users have tested the same text fragments with other detectors and obtained mixed results: some classified them as likely human.
Forensic detection specialists have warned that certain writing styles can lead to errors and that accuracy decreases when analyzing languages ​​other than English. Adding to this is a crucial detail: the detector itself is an AI, so its verdict relies on statistical patterns rather than documentary evidence. In a case like Orélien's, the difference between suspicion and certainty is enormous.
In fact, evidence has already been circulated using human texts that tools like Pangram flagged as AI-generated. This doesn't mean the accusation against Orélien is false, but it does necessitate treating the results with caution. There is no consensus among the AI ​​detectors , and for now, the suspicions are based on technical indications that neither confirm nor deny that the author used a language model.
The underlying debate: authorship, awards, and industry rules
Beyond this specific case, the controversy has raised a question the industry has long avoided: where is the line between using AI as a tool and letting it write for you? It's not the same to use it for research, editing, or overcoming writer's block as it is to delegate structure, style, or dialogue to it. French literary circles are debating whether both practices deserve the same criticism and whether readers should know what goes into each book.
Prizes are also under scrutiny. In the debate surrounding literary competitions, it has been argued that the rules must be much more explicit to prevent fraud and that a well-founded suspicion should be enough to disqualify a work . At the same time, some publishers and agents are urging against turning the detection process into a witch hunt, because false positives could harm legitimate authors. The lack of common regulations exacerbates this sense of uncertainty.
The Orélien case has also served to show that AI is no longer a futuristic threat, but a present-day problem in the book industry . From manuscript submissions to promotion, including contracts and copyright, each link in the chain needs clear guidelines. While some authors maintain that writing is a personal and unique experience, others believe that technology can be a useful tool if its use is transparent.
Thélyson Orélien's novel will continue to generate discussion, but its true impact goes far beyond an award or a sales surge. The episode has exposed the fragility of authorship criteria and the difficulty of proving who wrote what when AI is involved. Until shared rules and more robust verification methods are in place, each new case will be resolved amid suspicion, press releases, and an uneasy feeling that the industry is lagging behind technology.