French Publisher Gallimard to Screen Manuscripts for AI From January 2027

Madrigall, the French publishing group chaired by Antoine Gallimard, will begin running its incoming manuscripts through artificial intelligence detection software from January 2027, a source close to the matter said on Thursday. Antoine Gallimard has emailed the heads of the group's publishing houses to tell them that manuscripts will be screened from next year, according to the same source. The move makes the group, the fourth largest in French publishing, one of the first to put automated AI checks into its regular submission process.

The instruction covers titles from imprints including Denoël, Folio, J'ai lu, Casterman, Flammarion, P.O.L. and Minuit, among others. The group did not say which detection tool it will use or what happens to a manuscript that a screen flags.

The decision follows a controversy that has divided French letters since September. The Canadian publisher Boréal said on Wednesday that it was ending its collaboration with the author Thélyson Orélien, whose novel C'était ça ou mourir was published under its imprint. The house had been reviewing whether the book was written with artificial intelligence.

A Prize Withdrawn and a Reputation in Question

The book had won the FNAC prize in early September before the author was accused of cheating, and he was removed from the shortlist for the Goncourt, France's top literary award, according to an account of the case given by three technology entrepreneurs who have launched a prize of their own. The author, who is Canadian and Haitian, denies the allegations.

Until now, French houses have guarded against machine-written submissions almost entirely through the experience of the people who read the slush pile. In the vast majority of cases, publishers do not use software to detect possible AI use in the texts they receive, and the primacy stays with human reading. Albin Michel already uses a detection program, but a source said its output is not decisive when deciding whether to publish a book.

The stakes go beyond one disputed novel. A survey of readers in the United States found that 28 percent do not want artificial intelligence to play any role in creating books. Labeling books as written by humans is also gaining ground in the publishing industry, while the organizers of established literary prizes are rewriting their rules to exclude AI-generated submissions.

The controversy has also produced a counter move. Three entrepreneurs, Philippe Rodriguez, Albin Serviant and Gilles Mentre, said on Wednesday that they had launched what they describe as the first literary prize for books written with artificial intelligence. Named the Prix Pascal, it has no jury yet and no declared prize money. In a statement, the founders said the contest is open to anyone writing in French anywhere in the French-speaking world, with no requirement of nationality or residence and no publisher needed. Submissions will be accepted until November 30, and the prize is due to be handed out in mid-January.

What Workers in Publishing Face

For editors, scouts and agents, the shift adds a step to a job that has long rested on reading judgment. Detection software is not settled science, and a false accusation carries professional risk for authors and the editors who back them. The debate also raises questions the publishing industry has not resolved. Will authors be allowed to use services to research a book or seek writing advice? At what point does assisted research become assisted writing? Those questions are likely to shape contracts, submission guidelines and prize rules before any detector answers them.

Gallimard's group has not said whether a positive screen would halt a submission or simply trigger a second human read. Boréal's decision to cut ties with its author, and the continuing fallout over the novel's withdrawn prize eligibility, will be watched as early tests of how far houses will go.