arXiv Is Rate Limiting Submissions Because It Can’t Keep up With AI Slop

Research submission queues may now be longer for everyone, not just those using AI.

arXiv, a leading open-access repository for research preprints, is implementing submission rate limits due to a surge in low-quality papers, largely attributed to AI-generated content. Researchers can now submit only two papers per month, with a maximum of three active submissions at any time. The repository's volunteer moderators are overwhelmed by the sheer volume of submissions, which have doubled in the last two years and sextupled in computer science. While arXiv permits AI as a research tool if disclosed and compliant with standards, many new submissions are "thin papers of narrow scope" or "salami papers," often dense and AI-written, failing to meet scholarly requirements. This measure follows previous actions, including restricting review articles and requiring author endorsements, to combat AI-driven spam and maintain the integrity of the research archive.

The strain on arXiv highlights a critical challenge for academic publishing: distinguishing genuine research from AI-generated spam at scale. As AI tools lower the barrier to content creation, repositories face an escalating arms race to develop effective detection and moderation methods without stifling legitimate research. This situation forces a re-evaluation of submission review processes and the sustainability of volunteer-based moderation in the face of technological acceleration.

This move signals a growing effort across academic platforms to curb AI-generated content that degrades the quality and discoverability of research. It will likely prompt other preprint servers and journals to consider similar limitations, potentially increasing friction for all researchers. The increased scrutiny may also lead to more sophisticated AI detection tools being integrated into submission workflows.