This piece is an analysis of public materials: arXiv's announcement and its FAQ, plus press coverage. We have not observed how the limit is displayed inside the submission interface.
Who this affects
Everyone who posts to arXiv is covered, but the change bites hardest for three groups: a single lab member or PI who uploads the whole group's papers, machine-learning researchers who release several preprints right after a conference deadline, and high-volume authors. Researchers who post one or two papers a month may not notice any difference.
What is confirmed
arXiv's October 1 blog post, written by Kat Boboris, states: "arXiv now limits submitters to up to two submissions per calendar month, with a limit of three total active submissions at any given time." The FAQ fills in the mechanics.
Monthly cap 2 submissions per submitter per calendar month (new)
Active cap 3 concurrent active submissions (since 2024, unchanged)
Scope All arXiv subject areas
Volume 40,363 submissions in Sept. 2026; 20,569 in Sept. 2024; 9,869 in Sept. 2016
Support load Almost 9,000 tickets in Sept. 2026
| Scenario | How arXiv says it is handled |
|---|---|
| Co-authored paper | Counts only toward the submitter's limit, not the co-authors' |
| Rejected submission | Counts — "this rate limit is for submissions, not announced papers, because it is submissions that consume moderator time" |
| Deleted before announcement | Does not count toward either the monthly or the active limit |
| On hold from a prior month | Not charged to the new month's quota, but still counts toward the three active slots |
| Conference deadlines or backlogs | No exception; authors are told to "plan their submissions carefully" |
| Replacements (new versions) and cross-lists | Not addressed — unconfirmed |
The stated reason is an "exponential increase" in submissions that doubled volume in two years. arXiv points to thin papers of narrow scope, "salami" papers that split one piece of work into several, and dense AI-written papers as the drivers of moderator workload, which in turn slows review of solid work. It says it will monitor the effect "and modify as needed."
Not the first tightening
In October 2025, arXiv's computer science section began requiring review articles and position papers to have passed peer review at a journal or conference before they would be accepted (arXiv blog). That rule targeted specific paper types in one field. The new cap applies to every submitter in every field.
Reaction is split
Times Higher Education reports a divided response. Rutgers mathematician Alex Kontorovich backed the policy: "I'm not going to read your 42 papers this month. Pick two, your very best two." Harvard's Scott Kominers argued it "confuses" economizing on moderation with economizing on scholarship and called for better attention-allocation mechanisms instead of quotas. Terence Tao flagged the change to his readers on his blog.
Some coverage describes the limit as "per author." arXiv's own FAQ is explicit that it applies to the person who submits; this article follows the official wording.
Unconfirmed and outlook (kept separate from facts)
- Which time zone defines a "calendar month": not stated. Whether it follows US Eastern time, arXiv's usual operating clock, is unconfirmed, so submissions made right around midnight on the 1st carry some ambiguity.
- Replacements and cross-lists: not mentioned in the announcement; unconfirmed.
- Exceptions process: none is described.
- Outlook (our interpretation): because only the submitter is charged, many groups will likely spread submitting duties across co-authors. Whether that blunts the policy's goal, and whether arXiv responds with further changes, is not yet known.
What it means for you
arXiv remains free, so there is no new cost; the impact is on timing. A US or European lab that funnels every preprint through one account will hit the ceiling after the second paper and wait until the next month. Because rejections count, a miscategorized or poorly formatted upload now has a real cost. Teams posting camera-ready versions after a major ML conference deadline should map out who submits what before the deadline week, not after.
1. List this month's planned preprints and assign a submitting author per paper; only that person's quota is used.
2. Treat every submission as final: check category, formatting, and abstract first, since a rejection still uses a slot.
3. If you upload something by mistake, delete it before announcement so it does not count.
4. Check whether on-hold papers are occupying your three active slots.
5. Avoid relying on submissions made right at a month boundary until the time-zone rule is clarified.
Limitations
This article relies solely on arXiv's announcement and public reporting. We could not confirm how the submission interface enforces or displays the cap, or how new versions and cross-lists are treated. arXiv says the policy may be adjusted, so re-check the official post before a deadline.
Last verified 2026-10-03.
· [Primary] arXiv blog — Updated rate limit policy and FAQ (Oct 1, 2026)
· [Primary] arXiv blog — CS review and position paper practice update (Oct 31, 2025)
· [Press/Analysis] Times Higher Education — arXiv's new preprint submission cap divides opinion (Oct 2, 2026)
· [Press/Analysis] Terence Tao — arXiv updates its rate limiting policy (Oct 1, 2026)
- arXiv now caps each submitter at two submissions per calendar month in all fields, effective Oct. 1, 2026
- The three-active-submission limit (since 2024) remains; rejected submissions count
- Co-authored papers count only against the submitter; deleting before announcement frees the slot
- September 2026 brought 40,363 submissions, double two years earlier, and nearly 9,000 support tickets
- Month-boundary time zone and treatment of new versions remain unconfirmed; plan who submits what