Publisher dates Orélien's Goncourt manuscript to 2019 – before commercial generative AI became widespread

Publisher dates Orélien's Goncourt manuscript to 2019 – before commercial generative AI became widespread

The Académie Goncourt has removed Thélyson Orélien's novel "C'était ça ou mourir" from its 16-title longlist, four days after an anonymous X account claimed the book was written by artificial intelligence. The author denies it, and the case has ignited a French debate over the reliability of AI detectors – a question that remains unresolved when the Académie announces its shortlist on 6 October.

What has happened

On Friday 25 September 2026, the Académie Goncourt announced that it had removed the Canadian-Haitian writer Thélyson Orélien's debut novel from the prize's longlist. The decision came days after accusations of AI-generated text and plagiarism surfaced online, according to France 24/AFP.

The Académie justified the removal by citing its wish to "preserve the integrity of the Prix Goncourt and of the Académie itself, whose central mission is to promote and celebrate literature written by women and men." It pointed to findings suggesting the book is "in all likelihood, very largely the product of artificial intelligence," and also referenced plagiarism accusations (France 24/AFP).

In a statement to CBC News, the Académie said the decision was based on a review of the work of several "credible" researchers and journalists. At the same time, it said it would not comment further on the case before 6 October, when the shortlist is announced (Reuters, via Global Banking & Finance).

The stakes were high: the book had already earned Orélien, 38, the Prix du Roman Fnac, and it was a finalist for the Renaudot prize, according to BBC News, via Yahoo News. The Goncourt involvement was, in other words, a longlist place – now withdrawn.

How the accusations arose

The accusations first came from an anonymous X account called "Balance ton Claude" – roughly "Turn in your Claude." The account said it had run passages from various parts of the manuscript through the AI detection tool Pangram, which it described as highly accurate: "We tested several passages, from different parts of the manuscript, and got the identical result: 100 percent AI," the account said, according to BBC, via Yahoo News.

The account also added a contrast: Four texts Orélien wrote in 2014 scored 0 percent AI on the same tool (The Epoch Times).

CBC News repeated part of the experiment with other excerpts from the book and got the same result. That makes this a case where two independent actors report the same detector outcome – but it still does not resolve the underlying question of how reliable such tools are.

On Friday, the French essayist Samuel Fitoussi, who writes for the right-wing paper Le Figaro, came forward saying he was part of the collective behind the account. He justified the intervention with the fear that "after several millennia of civilization built on writing, AI will lead to reading and writing disappearing" (France 24/AFP).

The author's and publishers' response

Orélien rejects having used AI. "I have always loved creating and writing. I don't see why I would use a machine," he told the French newspaper Libération, according to France 24/AFP. He is in the process of compiling a documentation file together with his publishers to prove he wrote the book himself.

In a statement conveyed through the Canadian publisher Éditions du Boréal and quoted by CBC News, he argued in literary terms: "One can debate my style, but treating its techniques as evidence of artificial intelligence has nothing to do with literary criticism." To Yahoo News, he tied the style to tradition: "My traditions are Haitian and Caribbean. The images, the repetitions and the rhythms are the living word, they are my own" (Yahoo News).

Boréal has suspended all promotion of the book, but is not disowning the author. "Unfortunately, at this stage it is impossible for us to confirm or refute the accusations against the author," the publisher said, while urging people not to "jump to conclusions" (France 24/AFP).

The publisher has also presented a manuscript history: Orélien is said to have completed the first draft several years ago, and the text "then matured through many rounds with the editor, as the successive versions of the manuscript up to the final phase prove." "This editorial work was carried out with the utmost thoroughness, without any intervention from AI," Boréal wrote (The Epoch Times).

Orélien's French publisher has given an even more precise timeline: He is said to have begun the novel in 2017 and completed a draft in 2019 – years before commercial generative AI programs existed (The New York Times). It is worth noting that this timeline comes from the publisher itself and has not been independently verified; the two publishers' versions are compatible, but the accuracy is not independently confirmed.

The plagiarism accusations – a separate thread

Alongside the AI accusations, Orélien's earlier texts have been scrutinized. According to Reuters, citing the Canadian newspaper La Presse, a 2012 short story allegedly plagiarized passages from a book by the French author Louis Pauwels, while a 2013 opinion piece was allegedly copied from the Philippine newspaper The Philippine Star.

On 23 September, Radio-Canada reported that "at least five op-eds and opinion pieces" published by Orélien in Canadian and French media in 2012 and 2013 were partly copied from other writers' texts (Radio-Canada, via The Epoch Times).

These accusations are dated and documented in the press, but Orélien had not responded to them as of the reporting – Reuters states that he did not answer an inquiry. They are also logically separate from the AI question: Any plagiarism of texts from 2012–2013 says nothing directly about how the 2026 novel came into being.

What the case reveals about AI detectors

The case illustrates a larger knot: AI detectors are contested tools. They can produce false positives, their results vary with text length and genre, and different tools can reach different conclusions about the same text. That an anonymous account and CBC both got 100 percent AI scores on excerpts from the book is a consistent result on the same tool – not independent confirmation that the book is actually AI-generated.

The Canadian literary prize Writers' Trust of Canada has discussed how it should approach such tools, and expresses caution. "It's a very fast-moving subject," executive director David Leonard told CBC News. "We don't want to be in a situation where we get false positives, and we also don't want to end up using a tool that isn't effective."

The French debate has already reached television and radio, where writers and editors are discussing whether AI means "the beginning of the end of literature." Publishers are weighing the limits of detection tools, what kind of AI use is acceptable – if any – and what they should disclose to readers (The New York Times).

What remains open

A few limitations in the picture should be noted: All statements from the Académie, the publishers and the X account have been conveyed through secondary sources such as AFP, Reuters, CBC, The New York Times, BBC, La Presse and Radio-Canada; no primary documents underlie this article. That is worth keeping in mind when weighing the parties' claims against one another.

The unanswered question remains: Was the novel written by a machine? The Académie treats the detector findings as probable; the author and both publishers deny it and point to a manuscript allegedly dating from 2017–2019. Orélien promises documentation. Whether it arrives, and what it contains, is not yet known. The Académie will not comment on the case before 6 October, when the shortlist is announced – a moment that will show whether the case is closed, or reopened.