Cliquez ici pour lire en français
The row over C’était ça ou mourir (“It Was That or Die”), a novel by Haitian-Canadian writer Thélyson Orélien, has pushed a very technical question into everyday conversation: can software really tell you whether a text was written by artificial intelligence? Since the accusations against the book, AI detectors have been pulled into the spotlight too. In Cameroon, students, teachers, journalists and users of these tools are far from convinced they can be trusted.
A novel at the centre of an AI controversy 📚🤖
Published in Canada in spring 2026 and in France in August, the novel follows a history-geography teacher from Carrefour-Feuilles, in Port-au-Prince, who is forced to flee violence and cross several countries before reaching Montréal. It quickly drew major recognition, including the 2026 Prix du Roman Fnac, awarded by France’s biggest bookseller chain, and a place on the first selection for the Goncourt, France’s most prestigious literary prize.
Then, on September 21, a post by an anonymous X account called “Balance ton Claude” upended that trajectory. It accused the novel of having been written with the help of AI. The analyses shared online lean in particular on tools that estimate the probability that a text was AI-generated.
Orélien disputes the claims. He says he began the book in autumn 2017 and finished a first draft in 2019. In his view, stylistic quirks pointed out in the text cannot, on their own, prove it was produced by AI.
A few days later, on September 25, the Académie Goncourt withdrew the novel from its first selection. The Académie cited allegations of plagiarism and converging analyses that, in its view, point to extensive use of AI. Beyond the fate of one book, the affair raises a question well outside literature: how much is a detector’s score really worth?
What does a “90%” score actually mean?🤔
In a classroom, the question gets concrete fast. Picture a student handing in an assignment. The teacher suspects ChatGPT and runs the text through a detector. A percentage appears on screen, supposedly showing how far the text was produced by AI. For some Cameroonians, that number isn’t enough.
“If software tells me my text is 90% written by an AI, I’d like to know how it got there. Can it be wrong too? What is this tool actually basing its conclusion on?” asks Leila Mbango, a student.
That mistrust has taken on a new dimension with the Orélien affair. On social media, the debate has shifted to the tools used to analyse the text.
For some, the detector has become a new “inspector” 🕵🏾♂️
Among some teachers and supervisors, these tools are seen as a useful aid. The arrival of ChatGPT and other generative tools has changed how assignments, reports and other work get produced. Against that backdrop, detectors can look like one more way to check a document.
“I can understand a teacher using this kind of tool to get a first signal. But I don’t think a student should be condemned straight away on the basis of the result,” says Charles Atabgana, a teacher.
That nuance comes up again and again. A detector could be used to raise a flag, but not necessarily to settle a case. Some users argue that a software result should trigger a check rather than count as final proof.
Instead of a percentage, look at how the person works ✍🏾
For others, the best way to know whether someone really produced a text is still to follow their work. A teacher can ask a student to explain their reasoning, an editor can compare an article with a journalist’s earlier output, and a supervisor can ask for the steps behind a piece of work.
“If a student writes normally with me in class and their assignment looks different, I can ask questions. I find that more useful than looking only at a percentage. We all know each of our students’ intellectual level. I confront them as soon as I sense that what they’ve written doesn’t fit their usual reasoning,” explains Laure Abomo, a teacher.
Cameroonian students between trust and worry 🎓
In universities and schools, reactions are mixed. Some students see detectors as necessary to curb AI abuse. Others fear being wrongly accused.
“These days, if you write well, someone can say it’s ChatGPT. If you write badly, they can also say you used AI. You do need proof, but how can you be sure when even old books are indexed?” says Belinga, a student, wryly.
For people who regularly use writing assistants, things get even murkier. They may ask an AI to rephrase a sentence, fix a mistake or improve a text’s structure without ever asking it to produce the whole document. In those cases, the final text may well be the work of a person using a tool.
“Just because I use ChatGPT doesn’t mean all my work is written by ChatGPT. Otherwise, what are these apps even for? I think we’re not nuanced enough about this,” says Ludovic Aroga, a young professional.
For him, a detector that merely says a text “looks like” AI output can’t explain the process that led to the final result.
But some Cameroonians still want a way to check 🔎
Distrust of detectors doesn’t mean everyone wants to drop them. On the contrary, some users argue that tools for spotting fully AI-generated work are needed, notably in schools, competitive exams and some professional settings.
“There are people who hand their whole assignment to ChatGPT and present it as their own. Even on social media, you read a text and you can feel it’s pure AI. We have to find a way to check,” says Arnol Mba, a recent graduate.
For these users, then, the problem isn’t that detectors exist. It’s how their results are read. A score could be a signal, but not necessarily a conviction.
This is where the novel affair goes well beyond literature. Imagine a student sanctioned on the basis of a wrong result, or a journalist accused of using AI on an article they wrote themselves. Worse still, a writer whose work is called into question by an automated analysis. For some Cameroonians asked about it, the real issue is trust.
“A machine can help us check, but it must not become the judge,” sums up Armand Nga, a user.
That line captures a worry bigger than text detectors. As AI moves into schools, businesses, media and creative work, the tools built to detect it are gaining weight too.
Will detectors become unavoidable ? 🤖
Given the controversies online, it’s hard to say. What is certain is that AI has created a new need: working out whether content is human-made, machine-generated, or produced with a bit of both.
The C’était ça ou mourir affair has had an unexpected side effect. Beyond the accusations around the novel, it is pushing the public to look harder at software whose verdicts can fit in a single percentage. It’s a number that can reassure or alarm, but one many Cameroonians already refuse to accept blindly.
Where do you stand?
Are AI text detectors reliable enough to settle a case on their own, or should a human always check the result? Tell us in the comments
📱 Get our latest updates every day on WhatsApp, directly in the “Updates” tab by subscribing to our channel here ➡️ TechGriot WhatsApp Channel Link 😉
