arXiv Caps Monthly Submissions at Two as AI-Written Papers Strain Moderators

Researchers submitting to arXiv now face a fixed monthly ceiling, not just a moderator’s judgment. Effective October 1, 2026, the repository limits each submitter to two submissions per calendar month across all subject areas. arXiv says the change protects volunteer moderators as AI tools help drive a surge in low-value papers.

The limit follows the submitter

The monthly allowance applies to the person uploading a paper, not every name on it. A submission with several co-authors uses only the submitter’s allowance. Submitting in different subject areas does not create separate allowances; arXiv asks co-authors to coordinate their submissions.

A separate ceiling allows three active submissions at once. That restriction has existed since 2024; the October update does not introduce it. Earlier rate limiting was primarily left to individual moderators’ discretion. The new monthly rule applies across submitters and categories.

The policy’s FAQ spells out how the two ceilings interact. The key distinction is between work submitted for moderation and papers that ultimately appear on arXiv:

• Rejected papers count. After two submissions, accepted or rejected, a submitter must wait until the next calendar month.
• Papers held over from an earlier month do not consume the new month’s allowance, but still count toward the three-active-submission ceiling.
• Deleting a submitted paper before it is announced removes it from both the monthly and active-submission counts.

For authors with a backlog or a conference deadline, arXiv’s guidance is to plan submissions around both limits. It explains that even rejected submissions consume moderator time—the resource the monthly ceiling is designed to conserve.

Faster production, slower screening

September’s submissions generated almost 9,000 support tickets for staff and moderators, arXiv says. Growth was particularly steep in cs.AI, its artificial-intelligence category: submissions increased more than sixfold over two years. That figure describes the AI category, not computer science as a whole.

Moderators are seeing more narrow-scope papers, dense AI-written papers and single studies split into several smaller submissions, according to arXiv. Thomas Dietterich, chair of its editorial advisory council, says a small share of authors consumes disproportionate moderator time, delaying quality papers by days or weeks.

This is not a ban on AI-assisted research. arXiv permits AI use when disclosed and when the research meets its standards for scholarly interest and advancing the field. Its complaint is that many incoming papers fail those requirements, even as AI makes research tasks faster and easier.

arXiv calls the blanket limit a stopgap while it develops better moderation tools and procedures and considers best practices for increasingly advanced AI use. It says it will monitor the effects on authors, readers and the repository, and modify the policy as needed.