The Worst Thing about AI Writing
· The Atlantic
![]()
In 1968, Yambo Ouologuem was the toast of Paris. His first novel, Le devoir de violence, which traces the rise and fall of an African dynasty across centuries, was a best seller and won the Prix Renaudot, France’s second-most-prestigious literary prize. Ouologuem was the first African writer to receive the honor. The book continued to be celebrated until 1972, when the Times Literary Supplement published an extract of Le devoir de violence—in English, “Bound to Violence”—alongside a remarkably similar passage from Graham Greene’s 1934 novel, It’s a Battlefield. Given that certain scene descriptions and dialogue appeared nearly word-for-word in each novel, Ouologuem’s astonishing debut seemed at least partially plagiarized.
The story quickly became global news. Ouologuem’s American publisher pulped thousands of copies of the novel’s English translation. The head of Le Seuil, Ouologuem’s French publisher, called the accusations of plagiarism “exaggerated,” but Ouologuem accused Le Seuil of launching a smear-filled “press campaign” to avoid paying royalties. Communications with his publisher soon ceased, and Ouologuem returned to his native Mali. He continued to write privately, his son later said, but he never published again.
Visit betsport.cv for more information.
This past September, a Haitian Canadian author named Thélyson Orélien came to Paris as the toast of what the French call the rentrée littéraire, or “fall publishing season.” Orélien, who had been on the margins of the literary scene in Quebec, had quickly become a star in France, and he was now living what one critic called a “conte de fées”—a “fairy tale.”
C’était ça ou mourir—in English, “It Was Either That or Die”—is the story of a Haitian schoolteacher who, driven from his native country by gang violence, attempts to make his way north to Canada, a journey that is often dangerous and degrading. It’s much less stylish and thought-provoking than Bound to Violence, but it became a best seller in France; won the Prix du Roman Fnac, the first major award of the rentrée; and was in the running for the Renaudot and the even more prestigious Prix Goncourt. Then, on September 21—the same day that Orélien won the Fnac—a small, anonymous account on X posted an analysis of several chapters of the novel by the AI-detection tool Pangram, which claimed that 100 percent of the text analyzed was written by AI. Soon afterward, the newsmagazine Le Point reported on several instances of alleged plagiarism in Orélien’s early work, which the author later admitted to and apologized for. The Académie Goncourt disqualified him. Late last month, the Quebecois publisher of Orélien’s book announced that he had in fact used AI, despite his denials, and ended their relationship; his French publisher, Éditions Grasset, continues to sell the book. (Orélien, who in one of his latest public statements called the attacks he was experiencing a “media lynching,” did not respond to multiple requests for comment via e-mail and through his publishers.)
[Read: Yet another literary prizewinner seems to be AI]
The scandal was a reminder of how much—and how little—has changed in the French literary world since the uproar over Ouologuem’s work more than five decades ago. Although some of Orélien’s detractors voiced an earnest interest in authorial integrity, anyone following the online backlash against him could see that some of it was tinged with racism. Several commentators, for example, attacked Orélien’s command of the French language, suggesting that he lacked the ability to write his own work—a malevolent criticism that has long been marshaled against Black writers in France, including Ouologuem. And some of Orélien’s supporters, like Ouologuem’s, seemed determined to overlook the serious evidence against him out of loyalty to the narrative that he advanced.
Above the chaos is a question: Why did this novel, thematically rote and stylistically bland as it is, receive such acclaim in the first place?
Samuel Fitoussi, a young columnist for Le Figaro who is known for satirical columns that often target progressive ideals, is a member of the group behind Balance ton Claude, the X account that first alleged that Orélien had used AI. Its first posts argue that some influencers on X, including on the political right, were using AI without disclosing it.
Fitoussi, who also works at a venture-capital firm that has invested in AI, looked into Orélien’s novel after a friend read it and suspected AI use. When we met in Paris in October, Fitoussi told me that he saw sharing Balance ton Claude’s findings as a matter of fairness. He worries that writers who “finish second in those prizes will not get noticed,” making victory by an AI-aided text more troubling and “discouraging humans to write.”
But Fitoussi didn’t think that Orélien should be condemned wholesale over his apparent AI use. In a post for the Substack Persuasion, the Balance ton Claude collective wrote, “What matters here isn’t Orélien’s race or nationality, or our political views (or the political views that are sometimes attributed to us)” but rather “whether we are willing to have our literary and public discourse overwhelmed by AI writing—and whether we really will deprive ourselves of the discernment to tell the difference.”
[Read: The human skill that eludes AI]
That effort to depoliticize the incident came too late: Commentators on both the left and the right had already retreated to their respective corners. The left seized on the fact that allegations of AI use in fiction have focused disproportionately on Black writers, critiqued Fitoussi’s personal politics, and claimed that AI-detection technology is not reliable. (Although Pangram remains susceptible to false negatives—meaning that it can incorrectly flag AI writing as human—some research suggests that its rate of false positives is very low.)
Meanwhile, the right harped on the hollowness of literary institutions that it claimed were more interested in ideology than in literature. In Valeurs Actuelles, a far-right magazine, the critic Alix Avril argued that the novel’s subject matter had made it verboten to critique on literary grounds. “The hero is a migrant,” she wrote. “Consequently, the book is irreproachable. To criticize it is to risk being relegated to the camp of the salauds”—an insult that roughly translates to “assholes.”
At the same time, a third camp warned against judging the political value of the novel by how it was allegedly written. The 2021 Prix Goncourt winner, Mohamed Mbougar Sarr, writing in French, argued that someone testing immigrant writers with an AI detector doesn’t automatically make them a xenophobe, but also noted that stories about immigration are “the obsession of the fachosphère”—a term for the nationalist far right. It would be a mistake to discount the novel’s humanity, he added, just because the text appeared to have been generated by machines. “I did not say to myself, after finishing it, that it was an AI-generated book,” Sarr wrote. “I attributed its weaknesses to the limits and clumsiness of the writer.”
The narrator of C’était ça ou mourir, 29-year-old Jonas Dorléon, is portrayed as being something close to an angel in disguise. He manages to survive an extraordinarily hostile migrant’s journey without once taking advantage of someone else, or even considering the possibility of doing so. His fellow migrants are docile and possessed of Job-like patience. No one complains or raises their voice. Individually almost indistinguishable, they appear to exist merely to deliver proverbs in portentous tones. “Do you know why I would like to die before you?” a Peruvian migrant asks Jonas. “Because if I die before, at least you could tell the tale.”
Afterward, the narrator uses a “not X, but Y” formulation that is typical of AI writing, and that appears dozens of times throughout the book. “He did not say this emphatically,” Jonas says, “not like a phrase destined to leave a mark. He said it softly, almost like one asks for the time.”
[Read: The biggest tell that something was written by AI]
The strongest element of the book is its unhurried exploration of each stage of Jonas’s winding and perilous passage, which takes him from Haiti through the Dominican Republic, Brazil, the Darién Gap, and Mexico to the U.S. border. Occasionally, the stories Jonas encounters during this crossing can be gripping, including one about a migrant woman who does not notice that the infant child she’s carrying has stopped breathing. “She carried him on her back until the next camp,” Orélien writes, “then dug a grave for him with a teaspoon.”
Orélien largely struggles, however, to bring his characters to life, relying instead on stereotypes and caricatures. His noble, suffering migrants advance the idea that refugees must be perfect to be worthy of interest or compassion. The villains, too, are one-dimensional; most notable is a MAGA-hat-wearing Border Patrol officer who spits in Jonas’s face: “I don’t know if he had ever read a book in his life,” Jonas reflects, “but his eye had the hardness of someone who believes he knows without ever having seen.”
In an interview with Le Figaro that appeared before the scandal, Orélien said that the book was inspired by the stories of Haitian refugees whose Temporary Protected Status was revoked in 2017, during the first Trump administration. According to the interview, Orélien originally considered writing nonfiction but worried that the effect of a reported piece would be fleeting. “I told myself,” he said, “that literature could be a tool for creating something that lasts.”
As I read the book, I thought of a passage from Sarr’s Prix Goncourt–winning La plus secrète mémoire des hommes—a sexy, funny, pulpy, and erudite novel inspired by Ouologuem’s life, in which the narrator mocks the publishing world’s desire to praise the right kind of Black writing. “Western (let’s just say it: white) readers,” Sarr writes, want a particular kind of African writing: “Those Africans skilled at weaving tales in the moonlight, those Africans who don’t complicate things, those Africans who can still pull at your heartstrings with their moving stories.”
[Read: The paradox of African writing and AI]
Although Orélien isn’t African, the pressures he faced as a Haitian trying to write a novel that would sell well in many markets are similar—and didn’t end with the allegations against him. I spoke off the record with several people in the French literary world who separately told me they had hoped that Orélien would reveal himself to be a kind of situationist mastermind. Perhaps he was sending up the pretensions of the literary world in the manner of Percival Everett’s Erasure, in which a bourgeois Black professor finally achieves literary success by writing a work of outrageously stereotyped fiction under a pseudonym.
To summarize these people’s thinking: It might have been better if the artist had been playing a cerebral, sophisticated joke than a man who felt that he had an important story to tell and made some mistakes in telling it. But Orélien’s project was clearly quite earnest. “The idea,” he said, in an interview with the newspaper Libération, “was to reflect on the causes that push us to leave, the problem of insecurity, and to find solutions so that people can avoid fleeing their own homeland.”
Amid these debates, one irony is that the novel that did end up winning the Goncourt used AI—albeit in the satirical mode that my French sources had imagined Orélien using. In a radio interview last week, Philippe Jaenada said he made light use of ChatGPT while writing his novel L’inconnue du quai de Javel, including by hiding a single unattributed sentence written by the technology somewhere in the lengthy book. He seemed to be treating his own use of the technology as a joke, while also arguing that “we cannot push AI away, we cannot act as if it did not exist. We have to use it, we have to acknowledge it, we have to tame it and not let ourselves be replaced.”
One thing that AI hasn’t changed but has highlighted: Readers have always been able to convince themselves that a righteous cause is a proper substitute for a good story. In her 1965 Atlantic essay, “Must the Novelist Crusade?,” the author Eudora Welty argued that “the zeal to reform, which quite properly inspires the editorial, has never done fiction much good.” In place of the ambiguities and contradictions that define individual experience, she argued, the “crusader-novelist” substitutes crude generalities and just-so arguments. “If human beings are to be comprehended as real,” Welty wrote, “then they have to be treated as real, with minds, hearts, memories, habits, hopes, with passions and capacities like ours.”
Welty’s essay helps clarify a problem with some identitarian defenses of Orélien: Isn’t it a form of condescension to think that a novel should be subjected to less serious intellectual and artistic standards if it stands on the right side of history?
[Read: Writing about Blackness in its fullness]
A more moving defense of the book came from Lydia Vázquez, the novel’s Spanish-language translator, who has also worked with writers including the Nobel Prize winner Annie Ernaux. For Vázquez, what Orélien’s detractors labeled as clunky AI writing appeared to be a metaphor- and image-rich style that Orélien had forged in collections of poetry published well before the advent of ChatGPT.
I asked Vázquez about a specific hallmark of AI writing that appears in C’était ça ou mourir: In a scene with three characters, only two will speak at any given time. Vázquez told me that she had noted the same pattern but had a different opinion of it. “I thought that it was a metaphorical method to talk about the impossibility of communication,” she said. “If there was a collective affinity between the characters, there would be a rebellion, a revolution!”
After Orélien’s publisher decided to cut ties, Vázquez was the first to text me the news. But her interpretation stayed with me because it captured her honest emotional reaction to the book—and the opportunity it had briefly represented to expand the bounds of the French literary world. “I was very happy to see French readers give a good kick in the nose to the elitists who run the Goncourt and the other prizes in France,” she said, “because they’re tired of the French intelligentsia.” This comment, I felt, explained part of the novel’s appeal: It had been so successfully publicized as anti-elitist that it could be seen as a blow to the very elitists who had elevated it.
I read C’était ça ou mourir alongside Ouologuem’s Bound to Violence. I found the latter novel to be arrestingly strange and powerful, and it made me wish, despite his alleged plagiarism, that he had published more work. The Saïfs, a grand and often rapacious African dynasty that slaughters and enslaves other African peoples on its way to power, are at the center of a challenging narrative full of rich, complicated characters. Its vision of the world struck me as original, even if some of the language used to create it wasn’t Ouologuem’s. If he stole work, the act deserves condemnation, but he did create something genuinely new, and fully human.
In the end, the real problem may not be whether books written with the assistance of AI are published, or even whether they are given awards. It may be that the flattening ease AI brings to the writing process will drown out the ambitious, imperfect work that utters the truth no one wants to hear. At least for now, that is an exclusively human ability.