Since 1 October, arXiv has limited each submitter to two submissions per calendar month across all subject categories. A separate ceiling of three active submissions remains. The platform calls the monthly cap a stopgap: a way to protect volunteer moderation while it works on better procedures for a research environment in which advanced AI can make plausible manuscripts much faster to produce.[1][2]
The volume that triggered the rule
| Measure | Value | Context |
|---|---|---|
| September 2016 submissions | 9,869 | A ten-year comparison point.Source 1 |
| September 2024 submissions | 20,569 | The same month two years before the new cap.Source 1 |
| September 2026 submissions | 40,363 | A record month reported by arXiv.Source 1 |
| September 2026 support tickets | Almost 9,000 | Tickets add a second workload beyond screening manuscripts.Source 1 |
The September total was 96.2% higher than September 2024 and about 4.09 times the September 2016 level. Almost 9,000 support tickets is roughly 22 tickets for every 100 submissions, although a ticket is not necessarily tied to one paper. The point is not that one manuscript creates one ticket. It is that the archive’s human workload rose on two fronts at once.
The limit follows the submitter, not every co-author
How the new rule works
| Situation | How arXiv counts it | Practical consequence |
|---|---|---|
| Two papers submitted in one month | Both use the monthly allowance whether accepted or rejected. | A rejected submission still consumes moderator time and still uses a slot.Source 1 |
| A multi-author paper | It counts against the person who makes the submission, not all co-authors. | Labs must coordinate who submits, but rotating submitters does not remove the underlying quality obligation.Source 1 |
| A paper remains on hold into the next month | It does not use the new month’s two-paper allowance, but it still occupies one of three active slots. | Backlogs can constrain a submitter even after the calendar resets.Source 1 |
| A submission is deleted before announcement | It does not count toward the monthly or active limits. | An upload corrected before release can be removed without permanently spending the slot.Source 1 |
This is a moderation policy, not an AI ban
arXiv permits authors to use AI tools when that use is disclosed and the work still meets its scholarly standards. Its complaint is about the economics of volume. Moderators report more narrow papers, “salami” papers that split one body of work into several submissions, and dense AI-written manuscripts that do not provide enough scholarly value. A small group of high-volume submitters can therefore delay stronger papers for days or weeks.[1][2]
The policy exposes a broader shift. Before generative systems, the labour required to produce a credible-looking paper acted as an informal rate limit. Once drafting, coding and formatting become cheaper, screening becomes the scarce input. arXiv’s first response is not a smarter scientific judge; it is a queue-control rule that rations the attention of human experts.
What changes for researchers and readers
- Research groups need a submission calendar, not merely a publication calendar: a conference deadline does not create an exception.
- Authors have a stronger incentive to combine related results into one substantial paper instead of slicing work into the smallest publishable units.
- Readers and literature-search systems should not treat presence on arXiv as evidence that a paper has passed peer review or a correctness check.
- The success of the cap should be judged by moderation delay, inappropriate-submission rates and access for ordinary authors—not only by a lower paper count.
Sources & further reading
- Fair Moderation, Equitable Access, and AI: arXiv’s Updated Rate Limit Policy arXiv Blog · 1 Oct 2026 · Accessed 16 Oct 2026
- ArXiv versus AI slop: will capping preprint submissions work? Nature · 15 Oct 2026 · Accessed 16 Oct 2026