Can scientists reliably tell if a novel is written using AI?
Tuhin Chakrabarty, a computer scientist at Stony Brook University, studies how to identify AI-written text. His team analyzed 200 AI-flagged books, 200 self-published bestsellers, and 200 literary fiction titles from Amazon. They found AI-generated books contained the most verbatim overlaps with existing works, often repeating rare expressions or narrative beats from prior books. Chakrabarty…
Key points
- Stony Brook researchers found AI-generated books repeat rare word sequences from prior works more than human-written ones
- AI struggles with originality in literary fiction but mimics genre tropes and clichés in self-published books
- Chakrabarty’s methods flagged *Daggermouth* as partially AI-written, but the author denies AI use and publishers defend the book
Chakrabarty emphasizes transparency: disclosing AI use in writing could clarify creative contributions. He contrasts AI’s role in plagiarism with its potential for creative collaboration, citing examples like Vauhini Vara’s Searches (2025) and Sheila Heti’s New Yorker story. His research highlights a growing challenge: distinguishing AI-assisted work from fully human creation, particularly as AI tools become more integrated into creative processes.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Society & Work
All →- Elon Musk says he experiences ‘cyber psychosis’ over AI · 1 src
- ChatGPT AI predicts BTC peak by 2027 · 1 src
- Hacker wins $50k bounty for PS5 Linux exploit, sparks backlash from jailbreakers · 1 src
- Author fakes optical soliton paper with AI tools, IEEE catches it · 1 src
- AI art and props invade haunted attractions, sparking backlash · 1 src
Comments
via GitHub Discussions