Before ChatGPT became a verb, machines were already writing. The difference is that almost nobody expected automated literature to become a normal consumer interaction quite this quickly.
Authentiques Victimes began in 2016. Inside the novel were yellow ROBOTIQUES® kiosks: machines capable of manufacturing books on demand from profiles, keywords and stylistic preferences. At the time, it was deliberately uncanny — a publishing system in which a reader could increasingly request the book they wanted instead of waiting for an author to write it.
That idea did not emerge in a vacuum. Computer-generated literature is much older than generative AI. In 2013, NaNoGenMo — National Novel Generation Month — began challenging programmers to write code that generates a novel of at least 50,000 words. Its output ranged from procedural experiments to surreal machine literature.
But those systems mostly produced artifacts. Large language models changed the relationship: the reader can now become prompter, editor, commissioner and sometimes co-author, all inside the same conversation.
The uncomfortable part is not quality
A 2026 study by Sydney Sears and Deena Skolnick Weisberg found that participants rated AI-generated short stories as more absorbing and higher quality than the human-written stories used in the experiment. At the same time, the exact same kind of evaluation improved when people believed a story had been written by a human. In follow-up studies, participants were no better than chance at reliably identifying which stories were machine-generated.
That gap between perceived quality and perceived authenticity is more interesting than a simple human-versus-machine contest. We may enjoy a text and still care deeply about who — or what — made it.
Research on real-world AI fiction use is already pointing toward another shift: people use conversational models not only to request finished stories, but to repeatedly extend, remix and personalize narrative worlds. The boundary between reader and writer becomes less stable.
A fiction that had to be rewritten by reality
Authentiques Victimes was fully rewritten in 2026, ten years after it began. That creates an odd loop. A novel that imagined industrialized, personalized machine writing now exists in a world where the underlying mechanism has become ordinary enough to fit inside a browser tab.
The interesting question for the book was never simply whether software could produce sentences. It was about authorship, imitation, memory and the possibility that an original text might already exist under someone else’s name.
Generative AI has not resolved those questions. It has made them less hypothetical.