PARIS: The French literary world has been rocked by an AI scandal, unfolding in waves. Haitian–Canadian author Thélyson Orélien had his novel removed from the shortlist of the Prix Goncourt, France’s most prestigious literary prize, last Friday – soon after an anonymous account on X claimed it was written “almost entirely by artificial intelligence”.
It Was Either That or Die (C'était ça ou mourir), traces a Haitian refugee's perilous journey towards Canada through Brazil, Mexico and the US. It has topped French bestseller charts, selling roughly 35,000 copies, and recently won the Fnac novel prize alongside glowing critical reviews.
Orélien, 38, who emigrated to Canada in 2010 after a devastating earthquake in Haiti flattened his home, firmly denies using AI tools to compose his work.
The Goncourt Academy, which oversees the prize, withdrew the book citing “consistent findings of analyses by credible researchers and journalists” that suggested the work was, in all likelihood, largely produced by artificial intelligence.
Orélien’s French publisher said he began writing the novel in 2017 and finished a draft in 2019 – long before generative AI was commercially available. Orélien, who is Black, said he had been the victim of “numerous racist attacks” and is “currently compiling a dossier with [his] publishers in France and Quebec to prove [his] claims”.
Samuel Fitoussi, a columnist for conservative daily Le Figaro and author of Woke Fiction, later admitted to being part of the group project behind the anonymous X account. However, he denied any political motivation.
The controversy deepened when Quebec newspaper La Presse published allegations of past plagiarism against Orélien involving opinion pieces written between 2012 and 2013.
A subsequent investigation by Radio-Canada revealed that at least five of his earlier columns contained passages copied directly from other writers.
La Presse also found that Orélien’s short story Où va le monde, which won an award in 2012, had “at least four identical sentences” to Louis Pauwels’ 1960 book, Le matin des magiciens. Orélien was contacted by La Presse, but didn’t respond.
The scandal puts the spotlight on the reliability of such detectors. It also questions our expectations of writers, with some French authors arguing AI is now part of our lives – and that authorship is evolving.
Defending his style, Orélien maintained: “The images, the repetitions, the rhythms are living speech, they are mine. My traditions are Haitian and Caribbean.”
To test whether non-standard phrasing or oral storytelling traditions could lead Pangram into error, French AI expert Benoît Raphaël submitted chapters from four Haitian novels into the detector. All were correctly identified as human-authored.
Raphaël then entered a sample of Orélien’s poetry, short prose and critical writings published before 2023, when generative AI was not sufficiently developed to produce such text. These, too, were identified as human created.
Raphaël said before this scandal broke, he had written to another AI expert with his suspicions Orélien’s novel was written with AI assistance. He had spotted the telltale signs of AI-generated prose, including the overreliance on certain rhetorical constructions.
While studies show Pangram achieves high accuracy rates exceeding 95% with virtually no false positives, errors can still occur. The tool previously raised allegations against Jamir Nazir’s Commonwealth Writers' Prize-winning short story before the author was officially cleared.
Appearing on French television, iconic author Virginie Despentes forcefully rejected holding writers accountable for using AI. Defending Orélien, she insisted authors should write however they wish. Despentes praised the novel's brief sentences, striking metaphors and episodic rhythm as authentic markers of modern creative writing.
Author Raphaël Doan, who co-created an alternative history novel using ChatGPT in 2023, voiced strong scepticism. Doan argued that Orélien's novel exhibited formulaic AI traits, cliché imagery and a disjointed storyline, reflecting the difficulty AI models face in sustaining long-form narrative coherence.
Alexandre Gefen, editor of Culture History of AI, compared Orélien to a Renaissance master populating his studio with assistants working under his direction, noting that authorship has always evolved alongside changing writing technologies.
The commercial success of the book demonstrates that the author created a compelling voice for readers. If AI was used, it required extensive, iterative prompting. The effective use of AI may eventually become a recognised literary skill, distinguished by genuine ingenuity or uninspired shortcuts.