L'IA peut-elle écrire de la fanfiction dans la voix de n'importe quel auteur ?

AO3 a dû repenser la modération — à la fois en raison du volume et parce que les pastiches sont si bons qu'il est difficile de les distinguer de ceux écrits par des humains.

Background

Fan fiction communities like Archive of Our Own (AO3) have grappled with moderation challenges—not just due to sheer volume, but because some AI-generated stories are so finely crafted that they resemble human-written works indistinguishable from those by real authors (AO3 has had to rethink moderation — both because of the volume and because of pastiche so good it's hard to tell from human-written). AI systems have made significant progress in generating creative writing, including fan fiction, in various authors' voices. These models learn to mimic an author's style and tone by analyzing their existing works and using this information to generate new content. The quality and coherence of AI-generated fan fiction varies depending on the complexity of the author's style and the size of the training dataset. While current AI models can produce impressive results, they still struggle to fully capture the nuances and subtleties of human authors (AI systems have made significant progress in generating creative writing, including fan fiction, in various authors' voices. These models can learn to mimic the style and tone of a particular author by analyzing their existing works and using this information to generate new content. However, the quality and coherence of the generated fan fiction can vary greatly depending on the complexity of the author's style and the size of the training dataset. Current AI models can produce impressive results, but they still struggle to fully capture the nuances and subtleties of human authors. — Enriched May 9, 2026 · Source: The Guardian).