arXiv, the preprint server that much of the AI and physics research community uses to publish work before peer review, now limits each author to two submissions per calendar month, it said in a blog post Thursday. The rule, effective Oct. 1, also caps active submissions at three at a time and responds to a surge in low-quality papers that arXiv links to AI writing tools.

What Happened

The new limits apply across every arXiv category. Both accepted and rejected submissions count toward the monthly cap, so a paper that moderators turn down still uses up one of the two slots. A paper submitted on behalf of co-authors counts only against the person who submits it.

Some submissions are exempt from the monthly count. Papers deleted before they are announced do not count, and a paper that was submitted in an earlier month and is still on hold does not count against the new month's limit, although it does count toward the cap of three active submissions.

RelatedOpenAI Says Its Models Broke Into Two Outside Data Sites and Left It Out of Their Answers

The blog post includes support from Thomas Dietterich, a distinguished professor emeritus at Oregon State University who chairs arXiv's Editorial Advisory Council.

Why It Matters

The numbers arXiv published show how quickly volume has grown. The service received 9,869 submissions in September 2016 and 20,569 in September 2024. In September 2026 it received 40,363, which generated roughly 9,000 support tickets. Overall volume doubled in two years, and the cs.AI category alone grew more than sixfold since 2024.

arXiv points to three kinds of submissions behind the strain: low-quality papers, "salami" papers that split a single piece of work into several thin ones, and dense text generated with AI tools. It says a relatively small share of authors submit a large number of these papers and take up a disproportionate amount of moderators' time.

That moderation burden is the core issue. arXiv does not peer review papers, but moderators screen each submission for basic fit and quality before it is announced. A flood of marginal papers makes that screening slower for everyone, including researchers posting legitimate results.

Context

The rate limit follows earlier steps to control AI-era volume. In late 2025, arXiv's computer science section began requiring review articles and position papers to have passed peer review elsewhere before it would accept them, citing a similar flood of low-effort submissions.

A blanket cap is a blunter tool. It does not try to judge whether a paper was written with AI help, which is hard to prove, and instead limits how much any one submitter can push through the system each month.

What to Watch

Large labs and prolific authors who post more than two papers a month will feel the cap first. Because a paper counts only against its submitter, groups with many co-authors have room to spread submissions among members. Watch whether support tickets and moderation backlogs fall in the coming months, and whether arXiv adjusts the limits or adds exceptions in response to feedback.