A strange thing happened in the French literary world this week.
C’était ça ou mourir, the debut novel by Haitian-born Quebec writer Thélyson Orélien, had already become one of the sensations of the 2026 literary season. The novel follows Jonas Dorléon, a Haitian history and geography teacher forced to flee his country and travel across the Americas in search of safety. It had sold more than 35,000 copies in France, was being translated into more than twenty languages, appeared on the early lists of several major French literary prizes, and, on September 21, won the Prix du Roman Fnac. That prize is selected by a jury of 400 booksellers and 400 readers.1
Then came the accusation: artificial intelligence.
An anonymous account claimed that the AI detector Pangram identified substantial portions of the novel as AI-generated. Another detection company, Lucide, subsequently reported that its system classified all 35 sections it tested as AI-generated. Orélien strongly denies having used AI to write the novel, and his publisher, Grasset, has rejected the accusations.
It is important to state that clearly. At the moment, the use of AI in C’était ça ou mourir is an allegation, not an established fact.
But the outrage surrounding the possibility raises a question that will extend far beyond this particular author:
If a writer did use AI extensively while creating an excellent novel, why would that automatically make the work less creative?
That question is becoming unavoidable.
The fantasy that AI makes everyone equally talented
One of the strangest assumptions in the debate around generative AI is that it somehow eliminates talent.
The argument often sounds like this: if anyone can ask an AI system to produce a story, an image, a program or a piece of music, then the person producing it has not demonstrated meaningful ability.
Except anyone who has seriously used these systems knows how quickly that theory falls apart.
Ask an AI model to “write me a bestseller.”
See what happens.
You may receive grammatically correct sentences. You may even receive an acceptable plot. What you are unlikely to receive is a novel that hundreds of thousands of people cannot stop discussing.
The existence of AI does not mean that everyone suddenly possesses the same imagination, taste, judgment or instincts.
Give two people the same camera and they will not produce the same photograph. Give two directors the same actors, budget and screenplay and they will not produce the same film. Give two entrepreneurs identical software tools and they will not build identical companies.
Why would AI be different?
A mediocre creator with AI becomes a mediocre creator with a powerful tool.
A brilliant creator gets the same tool.
That is the part of the AI conversation that is frequently forgotten.
The comparison of the future is not “talented human versus mediocre human with AI.” It is increasingly talented human with AI versus mediocre human with AI.
And the talented human still has an enormous advantage.
AI does not eliminate taste
Suppose a novelist can generate twenty versions of a scene in an afternoon.
Which one should survive?
Should the protagonist say nothing? Should she become angry? Should the revelation happen now or three chapters later? Is that metaphor beautiful or embarrassing? Is the dialogue too explicit? Has the character suddenly behaved in a way that violates everything established about him?
The machine can produce possibilities. Someone still has to decide.
And the ability to recognize the right possibility is not trivial. It is taste.
Creative work has always involved selection. A photographer may take 2,000 photographs and exhibit twelve. A film director shoots far more footage than appears in the final film. A novelist can delete entire chapters.
Nobody argues that the existence of discarded alternatives proves that the artist did not create the finished work.
AI simply makes generating alternatives radically cheaper.
And that may make taste more important, not less.
When producing something competent becomes easy, competence stops being impressive.
The bar moves upward.
We already accept assisted authorship
There is another uncomfortable inconsistency in the outrage over AI-assisted writing.
Human beings have been assisting other human beings with authorship forever.
Politicians employ speechwriters. Celebrities release autobiographies written with ghostwriters. Novelists work with editors who can suggest substantial structural changes. Researchers have assistants. Screenplays can pass through multiple writers before filming begins.
A director receives help from cinematographers, editors, costume designers, composers, actors, set designers and hundreds of technicians. Yet we have no difficulty recognizing directing as a creative activity.
Why?
Because creativity does not necessarily mean physically executing every component of a work yourself.
It can also mean conceiving, directing, selecting, correcting and taking responsibility for the whole.
This does not mean that pressing “generate” once makes someone Tolstoy. It means that authorship is more complicated than counting who typed each sentence.
If a person develops the characters, determines the story, rejects hundreds of bad suggestions, rewrites passages, changes the structure, controls the emotional direction and ultimately decides exactly what readers encounter, dismissing that entire process because a machine participated in sentence production feels increasingly difficult to defend.
AI may reveal talents the old system never allowed us to see
There is also a deeply democratic side to this technology.
For centuries, we have confused creative potential with the ability to execute that potential using a particular technical skill.
Imagine someone with extraordinary stories in her head who struggles to write elegant prose.
Perhaps she grew up speaking another language. Perhaps she never received an elite education. Perhaps she cannot afford an editor. Perhaps she has brilliant visual ideas but cannot draw. Perhaps she understands exactly how an application should work but cannot code.
Historically, many of those people simply could not participate.
Someone with money could hire a ghostwriter, designer, developer, research assistant or editor.
Someone without money had fewer options.
AI changes that equation.
It allows people to buy, for a tiny fraction of the previous cost, capabilities that wealthy individuals and companies have always been able to purchase from other humans.
That does not guarantee brilliance.
It guarantees access.
And those are very different things.
Saving money is not cheating
There is sometimes a peculiar moral undertone in criticism of AI-generated work: if producing something becomes easier, cheaper or faster, then it somehow becomes less legitimate.
But efficiency has never worked that way.
We do not admire a filmmaker more because editing took twelve months instead of three. We do not demand that architects abandon computer-aided design because drawing everything manually requires more craftsmanship. We do not require accountants to avoid spreadsheets because arithmetic performed by hand involves more labor.
Technology repeatedly removes expensive steps from production.
Then people build more ambitious things with the resources they save.
Generative AI can do exactly that.
A small creator who previously needed a researcher, copywriter, translator, illustrator and assistant may be able to perform parts of those functions alone. A novelist can explore twenty narrative alternatives rather than two. A founder can prototype ideas before hiring a development team. A filmmaker can visualize scenes before spending money shooting them.
The savings are not merely financial.
They are also savings in time, coordination and experimentation.
And experimentation matters because great work rarely appears perfectly formed on the first attempt.
AI gives creative people more attempts.
That could make the best creators considerably better.
The detector problem makes the Orélien controversy even stranger
There is another reason to be cautious about turning C’était ça ou mourir into evidence of literary fraud.
AI detectors are not infallible.
Research published in 2025 comparing several AI-text detectors found that performance varied sharply by tool and threshold, and emphasized the policy trade-off between missed AI content and false accusations against writers.2 A 2026 analysis examining AI detection in academic writing went further, arguing that false accusations can become a serious problem depending on how detection results are interpreted.3
That does not prove Orélien wrote the novel without AI either.
It means we do not know.
And perhaps our obsession with finding out is revealing something about us.
Before the controversy, hundreds of booksellers and readers evaluated the novel for the Prix du Roman Fnac and selected it as the winner. Critics praised its language. Readers bought tens of thousands of copies. The story connected with people.
Then someone said “AI,” and suddenly the fundamental question became whether the creativity had been real.
But what changed in the book?
Nothing.
The words did not rearrange themselves overnight.
The characters did not disappear.
The readers who had been moved by the novel had still been moved by it.
What changed was our perception of who, or what, might have participated in creating it.
That is philosophically much more interesting than another argument about whether AI prose has recognizable stylistic tics.
There are legitimate lines to draw
None of this means that every use of generative AI should be celebrated.
There are serious unresolved questions around copyright and training data. Publishers may reasonably establish disclosure requirements. Readers may want transparency about how works are produced. Passing off completely automated output as the painstaking manual work of a human can raise legitimate concerns.
There is also an enormous difference between a writer spending a year directing, editing and transforming AI-assisted material and someone generating eighty novels automatically and flooding online bookstores with them.
AI spam is not creativity simply because a human clicked a button.
But these are arguments for developing better distinctions, not for declaring all AI-assisted creation fraudulent.
We should be asking how much human judgment was involved. Who made the meaningful creative decisions? Was the work transformed? Was there original thinking? Who takes responsibility for the final result? Were readers deliberately deceived about something material?
Those questions tell us far more than “Did you use AI?”
Perhaps what bothers us is that creation is becoming easier
For a very long time, we have associated artistic value with difficulty.
Years of training. Endless drafts. Expensive equipment. Technical mastery. Long nights.
There is something emotionally satisfying about knowing that a beautiful thing was difficult to produce.
But difficulty and value are not the same thing.
If technology allows someone to express an extraordinary idea in three months instead of three years, the idea does not become less extraordinary.
And if a person who previously lacked the technical ability to express her imagination can finally do so because AI fills the gap, perhaps we have not lost an artist.
Perhaps we have gained one.
The arrival of generative AI will undoubtedly produce mountains of mediocre content.
Every technology that makes publishing easier does.
But the same technology will be placed in the hands of extraordinary people.
And extraordinary people do not stop being extraordinary because their tools improve.
They become more dangerous competitors.
That may ultimately be what makes this moment uncomfortable.
AI is not lowering the ceiling of creativity.
It is lowering the floor required to participate while simultaneously giving the most talented people more leverage, more iterations and more opportunities to experiment.
The mediocre will use AI.
So will the brilliant.
And when both do, the bar does not disappear.
It rises.