A rate limit answers a capacity problem with a number. It does not answer whether the work behind that number is valuable. From 1 October, arXiv allows each submitting researcher up to two submissions per calendar month across all categories, while retaining a maximum of three active submissions at once. A rejection still uses one of the monthly slots because it still consumes moderation work.[1]
The immediate logic is queue management. arXiv reported 40,363 submissions in September 2026, compared with 20,569 in September 2024 and 9,869 a decade earlier. The latest month generated almost 9,000 support tickets. Nature reported that nearly 300 volunteer moderators were collectively checking about 1,000 manuscripts a day and that thousands of submissions were waiting to appear.[1][2]
What the rule actually says
| Measure | Value | Context |
|---|---|---|
| Monthly allowance | 2 submissions per submitter | Shared across all arXiv categories and reset each calendar month.Source 1 |
| Active pipeline limit | 3 submissions at one time | The pre-existing cap remains and includes papers still on hold.Source 1 |
| Rejected submission | Counts toward the monthly allowance | The moderation work has already been incurred.Source 1 |
| Co-authors | Only the submitting account uses its allowance | The policy attaches to the submitter rather than every listed author.Source 1 |
| Status | Stopgap | arXiv says it will monitor the effect and modify the policy as needed.Source 1 |
The limit changes access, not the standard
A strong paper submitted third in the month can be delayed. A weak paper submitted first can still enter moderation. The numeric cap is therefore a prior restraint on throughput, not a verdict on the manuscript. arXiv’s scholarly requirements—originality, relevance, self-containment and interest to the field—still have to be applied separately by moderators.
Four problems that are easy to confuse
| Problem | What a monthly cap can do | What it cannot do |
|---|---|---|
| Moderator overload | Reduce the maximum volume from one submitting account | Create more volunteer expertise or eliminate backlogs immediatelySource 1Source 2 |
| Low-quality or fragmented papers | Raise the cost of sending many submissions quickly | Determine whether a particular paper is thin, duplicated or significant |
| AI-assisted writing | Slow high-volume output from one account | Distinguish responsible assistance from fabricated or unchecked research |
| Fair access | Prevent a small number of submitters from occupying unlimited queue capacity | Guarantee equal impact across fields, teams and conference schedules |
Why “AI slop” is an incomplete diagnosis
arXiv says moderators are seeing more narrow papers, fragmented “salami” submissions and dense AI-written manuscripts. Nature quotes arXiv’s scientific director attributing much of the recent surge to AI tools. Those observations matter, but the published totals do not identify the proportion of September’s 40,363 submissions that used AI, failed standards or came from unusually prolific accounts.[1][2]
That gap matters because AI can change several things at once. It can lower the cost of writing a bad paper, help a competent researcher prepare a legitimate one, translate work into clearer English or generate plausible references that were never checked. A policy aimed at volume cannot tell those cases apart. Its advantage is administrability; its weakness is bluntness.
The obvious workaround is also a governance signal
Because the allowance belongs to the submitter, collaborators can rotate who uploads a paper. That may be entirely legitimate in a large lab, but it also means determined high-volume groups can distribute their queue usage. arXiv can respond with stronger account-level or author-group rules, yet each additional linkage increases privacy, identity and false-positive complexity.
The archive is infrastructure, not merely a website
Preprint servers sit between private research and formal publication. They make results visible quickly, establish a public timestamp and let communities inspect work before journal review. When moderation capacity becomes scarce, a delay can affect priority claims, hiring, conference attention and the pace of discussion. That is why an apparently simple queue limit has scientific consequences.
The two-paper rule may be a sensible emergency valve. It should not become an excuse to avoid the harder work: better moderation tools, clearer AI-use disclosures, sanctions for fabricated references, staffing models that match the archive’s scale and evidence about which submission patterns consume disproportionate effort. A rate limit can buy time. It cannot decide what science is worth sharing.
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