Archive of Our Own faces grassroots AI-detection campaign with accuracy concerns

A new movement among fanfiction writers aims to identify and flag stories written with Claude, ChatGPT, or other generative AI tools on Archive of Our Own, the community's largest repository hosting over 12 million fanworks. The campaign launched this week, driven by longstanding opposition to AI use in creative writing communities. Participants are deploying detection methods to scan uploaded works, though the tools' reliability remains contested.

The effort has divided the fanfic community. Critics warn that current AI detection software produces false positives at rates high enough to wrongly accuse human authors, particularly those whose prose is formulaic or non-native English speakers. Writers whose styles lean toward simple sentence structures, repetitive phrasing, or non-idiomatic English face disproportionate risk of being flagged. The detection tools in question—many of them the same commercial classifiers used in academic plagiarism cases—have documented accuracy problems when applied to creative fiction. Archive of Our Own does not ban AI-generated content outright, leaving enforcement to volunteer moderators and community norms. The site's terms of service require that uploaded works be transformative fan creations, but they do not explicitly address authorship by large language models. That ambiguity has left the door open for grassroots policing, with some users now running suspected fics through third-party detectors and reporting results in comment threads and social media. The dispute escalated quickly, with accusations already circulating based on inconclusive scans. As of early July 2026, the Organization for Transformative Works, which governs Archive of Our Own, has announced no formal policy change.