arXiv caps submissions as AI‑assisted papers overwhelm volunteer moderators

arXiv announced a new rate‑limit policy that will take effect on 1 October 2026. Under the rule, any individual can submit no more than two papers in a calendar month across all categories, and may have at most three papers pending review or revision at any given time.

The restriction applies to every submission, regardless of whether the paper is eventually accepted. A rejected manuscript still consumes one of the two monthly slots, while a paper that is deleted before it is announced does not count toward the limit.

Why the cap was introduced

arXiv’s moderation volunteers have reported a dramatic increase in the volume of AI‑assisted submissions. The repository recorded 9,869 submissions in September 2016, 20,569 in September 2024 and a staggering 40,363 in September 2026, a rise that coincided with nearly 9,000 support tickets generated in the same month.

According to arXiv, advanced AI tools are enabling authors to produce thin, narrowly scoped papers that are densely written by algorithms and deliberately fragmented into multiple “salami” slices. These papers demand disproportionate time from volunteers who must verify originality, significance and scholarly relevance.

What is still allowed

The repository has not banned AI‑assisted research. Authors may continue to use AI tools provided the use is disclosed and the final work satisfies arXiv’s standards for originality, significance, self‑contained research and scholarly interest.

The policy is described by arXiv as a temporary stopgap while the organization upgrades its moderation tools and works out longer‑term practices for AI‑generated content. It is explicitly not a ban on AI‑assisted papers.

How the limits work in practice

Co‑authors do not each consume a slot when one of them submits on their behalf. This encourages coordinated submission strategies among collaborators, as only the submitting author’s quota is affected.

Papers that remain in the system from a previous month do not reduce the number of new submissions allowed in the current month, but they do count toward the three‑paper active limit. Consequently, an author who carries over two active papers can only submit one new manuscript until one of the older papers is withdrawn or accepted.

• Maximum two submissions per calendar month per author
• Maximum three active submissions at any time
• Rejected papers still count toward the monthly quota
• Deleted, unannounced papers are exempt from the quota

arXiv says it will monitor the impact of the new limits and will revise the policy if necessary. The organization has not yet measured how the caps affect different disciplines or early‑career researchers, leaving open questions about equity and access.

What this means for English‑language institutions is that research groups will need to plan submissions more carefully, especially when multiple team members are working on AI‑heavy projects. Coordination will become essential to avoid exhausting monthly slots, and teams may need to prioritize the most impactful manuscripts for early submission while holding back less critical work for later months.

Potential long‑term implications

If the caps prove effective, arXiv may consider more nuanced controls, such as category‑specific limits or weighted quotas based on citation impact, to balance the need for rapid dissemination with sustainable moderation.

Conversely, critics warn that strict limits could discourage rapid sharing of incremental results, particularly in fast‑moving fields like machine learning, where timely pre‑prints are a key part of the research ecosystem.

The policy also raises questions about how non‑English speaking communities will adapt, given that many AI‑generated papers are produced in English and may face additional scrutiny when translated.

In the short term, authors are advised to consolidate related findings into single, more comprehensive manuscripts rather than fragmenting them, a practice that could improve overall scholarly quality.

Finally, arXiv’s leadership emphasizes that the limits are provisional and that feedback from the global research community will shape any future adjustments, underscoring the collaborative nature of the repository’s governance.