Leading preprint server clamps down on ‘AI slop’

Back to news list

Source: Science Magazine

Original: https://www.science.org/doi/abs/10.1126/science.aef8896?af=R...

Published: 2026-01-29T07:00:02Z

In October 2025, the preprint server arXiv introduced measures against "AI slop", i.e. low-quality contributions generated by artificial intelligence. It will no longer accept reviews or position papers in computer science unless they have passed the previous review. Since early 2025, the number of such submissions has risen exponentially, first in computer science and then in other fields, with rejection rates rising from an average of 4% to 10-12%. Most AI-generated content comes from first-time authors, often young researchers who may not know about moderation. Other servers, such as OSF, stopped new submissions in August 2025 due to the majority of low-quality files, and PsyArXiv went into pre-moderation. Moderators expect that improving language models will make it harder to distinguish fake from legitimate work.[1]