Guide19 min read27 sources
arXiv's two-submissions-per-month limit: what counts, and why October 2026 still set a record
Since 1 October 2026 arXiv lets each submitter make two new submissions a month. What counts, what arXiv leaves open, and what October's record 42,152 can and can't show.
Since 1 October 2026, arXiv has limited each submitter to two new submissions per calendar month, with no more than three active submissions at any one time. The limit follows the person who submits: a paper submitted on behalf of several co-authors counts only against the submitter. Rejected submissions count, submissions to different subject areas count together, and a paper deleted before it is announced does not count. arXiv calls the rule "a stopgap".1 It did not stop the growth in its first month. arXiv's own monthly figures, downloaded on 17 November 2026, show 42,152 new submissions in October 2026, a record, after 40,363 in September.
What is not established is whether the limit works. One month cannot show it, and we found no figures from arXiv on declined papers or moderation times since the rule began. The public monthly count also appears to include only papers that were posted, so it cannot show whether low-quality submissions fell. arXiv's pages are silent on several cases authors ask about, among them cross-lists added later, submissions made through conference proxies and the time zone that ends a month, and they describe no formal way to ask for an exception.
Below: what counts, why arXiv acted, the numbers and why October still set a record, a timeline of arXiv's recent rules, how to plan a busy month, and what is still open. Every source was checked on 17 November 2026.
Key terms#
A submitter is the arXiv account holder who uploads and completes a submission. A new submission is a paper not yet on arXiv; a replacement is a new version of a paper already announced. A paper is announced when it appears in arXiv's public listing, and only then does it receive a permanent identifier. A submission on hold has been stopped by arXiv's automated or human checks for a closer look. arXiv does not define an active submission; its FAQ treats a paper on hold as active.
What is arXiv's two-submissions-per-month limit?#
arXiv announced the rule on its blog on 1 October 2026, in a post by Kat Boboris titled "Fair Moderation, Equitable Access, and AI". Its central sentence: "arXiv now limits submitters to up to two submissions per calendar month, with a limit of three total active submissions at any given time."1
arXiv's moderation help page3 now has a section on submission rate: "Each author may submit up to two new submissions per calendar month with a limit of three active submissions." For replacements, it says "there is a one per week rate limit after version 5", and it adds: "arXiv may request that a particular author further limit their submission rate." The blog speaks of submitters and the help page of each author; the blog's FAQ settles the question of co-authors in favour of the submitter, as shown below. The help page's wording also makes clear that the monthly count is about new papers.
The rule replaced a looser one. On 30 September 2026, the same help page had a section headed "Excessive submission rate". It said "There is a practical limit to the rate at which appropriate, independent submissions can be produced by any one person" and asked authors with a back catalogue to "submit no more than three papers per day".4 The new FAQ adds: "Since 2024, arXiv has had a limit of 3 active submissions at any time, and this limit remains in place."1 We found no 2024 announcement of that limit on arXiv's blog or help pages.
arXiv describes the rule as temporary in purpose but gives no end date. It is in place "as a stopgap while we determine what may be the new best practice for authors employing more and more advanced AI tools, and to give us time to improve our moderation tools and procedures accordingly", and arXiv will "continue to monitor the effect of this update on authors, readers, and the arXiv corpus, and modify as needed."1 The text has not changed since launch: the post we read on 17 November matches the Internet Archive's copy2 taken at 20:13 UTC on 1 October word for word.
What counts towards the limit, and what does not?#
The blog's FAQ answers five cases, the help pages one more. For the rest, arXiv's documents are silent, and a search of its user-support knowledge base on 17 November found no article on the rate limit.
| Case | What arXiv says | Where |
|---|---|---|
| A submission that is declined | Counts: "whether they are accepted or rejected, you will have to wait until the next month" | Blog FAQ |
| Papers in different subject areas | Count together: "across all categories" | Blog FAQ |
| A paper submitted for several co-authors | Counts for the submitter only | Blog FAQ |
| A paper deleted before it is announced | Counts towards neither limit, "whatever the reason" | Blog FAQ |
| A paper still on hold when the next month starts | Does not count against the next month's two; still one of the three active | Blog FAQ |
| A new version of an announced paper | The help page limits new submissions and sets a separate rule for replacements: one a week after version 5 | Moderation and replacement help |
| A paper unsubmitted (returned to incomplete) before announcement | Not addressed; the FAQ speaks only of deleting | None |
| A cross-list added after announcement, or a withdrawal | Not addressed | None |
| Submissions by a proxy, such as a conference or journal | Not addressed for the monthly limit; a proxy account may make no more than 10 submissions a day without prior approval | Proxy help |
| The time zone that ends a month | Not addressed | None |
| Exceptions | No process published; no exemption mentioned for endorsers, journals or conferences | None |
The rejected case comes with a reason: "this rate limit is for submissions, not announced papers, because it is submissions that consume moderator time."1
Readers asked about new versions, overlay journals and conference proceedings in comments under the blog post on 2 October, and about the time zone on 31 October; no reply from arXiv appears. arXiv's daily cut-off is in US Eastern Time: "New submissions received by 14:00 (Eastern Daylight/Standard Time Zone) are generally made available at 20:00 (Eastern)".15 Whether the month also ends in Eastern Time, arXiv does not say.
For exceptions, the pages offer only the general channel: "All communication between authors and moderators or about moderation decisions must be addressed through our user support portal".3 Steinn Sigurdsson, arXiv's scientific director, told Times Higher Education (THE) on 3 November: "If you think you're a greater scientist than Erdős and deserve to be publishing at a higher rate, then come talk to us."24 That is the nearest thing to an exception route we found; arXiv's pages publish no procedure. The limit can also go lower: THE reported that on 20 July 2026 arXiv's support team asked one author, who had been making an estimated three or four AI-assisted submissions a month, to limit himself to one.
Does arXiv's submission limit apply to co-authors?#
No. The FAQ is direct: "All arXiv rate limit policies apply only to submitters. If you submit a paper with several co-authors, that only counts towards your submission rate, not theirs. It is important for co-authors to coordinate their submissions accordingly."1
So a group can post more than two papers a month between them, as long as no single submitter makes more than two new submissions. Each submitter must be endorsed for the category (our endorsement guide covers how), and "It is the responsibility of the submitter to ensure that all authors consent to the article being submitted".16 Each person may have one account: "Each user may have only one registered arXiv account."17
Spreading papers across submitters does not change what moderators look for: "If we notice that an author has made multiple similar submissions, or that a new submission appears to be a revision of a recent submission, we may request that the submissions be consolidated or versioned."3
On 11 November, THE reported an unpublished analysis by the founder of Originality.ai, who said there appeared to have been a "shift towards sharing submission duties" in October.25 We could not see the analysis or check it. One mathematician named in it told THE: "Being a co-author on several papers does not by itself show that someone is working around a submission cap."
Why did arXiv limit submissions?#
arXiv gives three reasons: the volume, the kind of papers arriving, and fairness to other authors.
Volume. "In September of 2024, arXiv received 20,569 submissions. This September, arXiv received 40,363 submissions, which in turn generated almost 9,000 support tickets for arXiv staff and moderators."1 A chart in the post shows cs.AI, the artificial intelligence category, with an "over 6X increase" in two years.
The kind of papers. "Our moderators are observing an increase in thin papers of narrow scope, as well as 'salami' papers, where a single work is broken up and submitted as a set of smaller papers. There is also a marked increase in dense, AI-written papers."
Fairness. The post quotes Thomas Dietterich, chair of the arXiv Editorial Advisory Council: "a relatively small proportion of authors are submitting a large number of low-quality papers and consuming a disproportionate fraction of the moderators' time. This is unfair to authors who continue to submit quality papers". The rule is meant "to equitably distribute our volunteer moderators' time across arXiv authors, and protect the arXiv corpus from a sharp increase in inappropriate submissions."
Sigurdsson called the cap "an expedient measure to buy us some breathing room" and said the team had thought about how to "minimise the impact on good faith authors".24 Journals face the same pressure; our post on the probability journals that paused submissions covers their new AI rules.
The limit does not ban AI. The blog says: "arXiv policy permits authors to employ AI as a tool in support of their research as long as it is disclosed and the research meets arXiv's standards of scholarly interest and advances the field of research."1 The moderation page asks authors to report significant use of tools, including "text-to-text generative AI", and reminds them that "by signing their name as an author of a paper, they each individually take full responsibility for all its contents, irrespective of how the contents were generated." AI tools "should not be listed as an author".3 Our guide to writing an AI disclosure statement shows how to word one.
How many papers did arXiv receive in October 2026?#
arXiv publishes a monthly count on its statistics page5, which describes it as "the number of new submissions received during each month since August 1991". We downloaded the page's CSV file6 on 17 November 2026. It gives 40,363 for September 2026 and 42,152 for October 2026, the same figures as arXiv's blog and THE.
| Month | 2024 | 2025 | 2026 | 2026 against 2025 |
|---|---|---|---|---|
| September | 20,569 | 26,646 | 40,363 | +51.5% |
| October | 24,226 | 27,692 | 42,152 | +52.2% |
| Change from September to October | +17.8% | +3.9% | +4.4% |
From the same file: the first month above 20,000 was May 2023 (20,092), as arXiv itself noted in 20247. The first above 30,000 was March 2026 (30,045), and the first above 40,000 was September 2026 (40,363). January to October 2026 adds up to 312,802 submissions, 32.6% more than the same months of 2025 and more than the whole of 2025 (284,486). The November row read 17,345, with the page's totals given as of 16 November, so that partial month cannot be compared.
What the count measures. arXiv calls these figures submissions received. They behave like a count of new papers that received an arXiv identifier in that month, and a submission gets its identifier only when it goes public: "Once a submission is made public, it receives a permanent arXiv identifier".8 On 17 November, arXiv:2609.40363 existed and arXiv:2609.40364 did not; arXiv:2610.4215210 existed and arXiv:2610.42153 did not. The highest number in each month equals that month's total. The page's totals agree: the CSV's rows add up to 3,252,491, the page's total of 3,252,335 plus the 156 deleted articles it excludes, and adding its 2,431 migrated articles gives exactly the 3,254,766 articles it says are available.5
Two things follow, on our reading. A declined submission never receives an identifier, so it is missing from the count, although it counts towards an author's limit. And a paper counts in the month it is posted, which can be long after it was submitted: arXiv:2610.000019, the first October number, has a first version dated 8 May 2026, a date stamp "recorded as the time the submitter clicks" to submit. arXiv's pages do not describe the count in these words.
Why did arXiv submissions keep rising under the cap?#
THE's headline on 11 November was "Submissions to preprint server keep rising despite author cap".25 The public numbers fit several explanations and cannot separate them.
The cap binds only above two. arXiv's post says a disproportionate share of moderator time goes to "a relatively small proportion of authors". If most submitters send one or two papers a month, a limit of two trims only that minority, and the total can still rise. arXiv has not published how many submitters were above two a month before October.
Papers can move to co-authors. The rule allows it. The analysis THE reported is the only evidence we found that this happened in October, and we could not see it.
The count includes papers submitted earlier. On our reading of the count, a paper held in September and posted in October counts in October.
The trend was already steep, and October is a heavy month. arXiv wrote in 2024 that "October has historically been a heavy submission month for arXiv".7 In the CSV, August 2026 was already 42.8% above August 2025, before the cap; September was 51.5% up and October 52.2%. The rise from September to October 2026, 4.4%, was close to 2025's 3.9% and well below 2024's 17.8%.
One month cannot show what a rule does. arXiv's own reasons point to other measures, moderator time spent on low-value papers and how long good papers wait, and neither is in the public statistics.
How have arXiv's rules for authors changed since 2024?#
Each row comes from arXiv's own pages, except the May 2026 penalty, which we found only in an announcement on X and its report in the press.
| Date | Change | In the source's words |
|---|---|---|
| 2024, no exact date given | Limit of three active submissions | "Since 2024, arXiv has had a limit of 3 active submissions at any time" (2026 FAQ; we found no 2024 text) |
| 31 October 202521 | Review and position papers in Computer Science | They "must now be accepted at a journal or a conference and complete successful peer review" |
| December 2025 (Mathematics), 21 January 202620 (all categories) | Endorsement | An institutional email address is no longer "the sole qualifier of endorsement for new authors" |
| 14 May 202622 | Penalty for unchecked LLM output, announced on X by Thomas Dietterich | A one-year ban "followed by the requirement that subsequent arXiv submissions must first be accepted at a reputable peer-reviewed venue", as quoted by THE |
| 1 October 20261 | Rate limit | "up to two submissions per calendar month, with a limit of three total active submissions" |
| 8 October to 6 November 202626 | Four blog posts: two holiday schedules, an update on arXiv's move to the cloud, a database migration | None changes the rules for submissions |
The May penalty needs care. Dietterich, then described by THE as chair of arXiv's computing section, opened his thread: "Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents were generated."22 THE quoted him from the same announcement: "If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can't trust anything in the paper."23 On 17 November we searched arXiv's help pages and blog again, including the blog's search for "ban" and "hallucinated", and found no arXiv page that states the one-year ban. The moderation page holds authors responsible for "incorrect references" produced by AI, and the Code of Conduct enforcement page18 lists, among general actions, "Suspending submission privileges: the user may not be allowed to submit papers for a designated period." Our guide to checking for hallucinated references covers how to avoid the problem.
How do you plan a month with more than two papers?#
The FAQ's advice for anyone with a backlog or a conference deadline is "to plan their submissions carefully so that the monthly submission rate limits and active submission rate limits will not negatively impact them."1 In practice:
- List the month's papers and their submitters. Each submitter has two new submissions a month and three active at once.
- Agree who submits each paper. A co-author may submit, and the paper then counts only for that person, who must be endorsed for the category and have every author's consent.
- Check each paper before it goes in. A declined paper still uses a slot. Check completeness, category, references and the AI-use statement first.
- Delete early if you spot a mistake. A paper deleted before announcement counts towards neither limit; submitting it again is a new submission.
- Leave room for holds near deadlines. A paper on hold stays among your three active until it is resolved, which "may take several days".
- Avoid the last day of the month. arXiv does not say which time zone ends the month, and the end of the year "has many US federal holidays in quick succession" that move announcement dates.27
- Put new versions in as replacements. The help page limits new submissions; replacements have their own rule.
What does a rejected or on-hold submission cost you?#
A rejected submission costs one of your two slots for that month, and it can cost time later: "Submitters who have had works previously delayed or declined by arXiv should anticipate closer scrutiny on future submissions."3 You can appeal through the user support portal, but "decisions upon appeal are final"; our endorsement guide's section on declined papers covers the appeal steps.
A submission on hold does not use next month's allowance, but it stays among your three active submissions until arXiv resolves it. arXiv's status page says: "Due to the large volume of submissions the process may take several days." And: "Please do not make a duplicate submission while your work is on hold. Deleting and resubmitting your work will lead to further delays."13 An older knowledge-base article, last edited in April 2021, still says "Usually most submissions that are on Hold will be resolved within 48 hours."19
Should you replace an arXiv paper or submit it as new?#
Replace it. A new version of a paper already on arXiv is a replacement, and the monthly limit is on new submissions. arXiv's proxy help page states the principle: "If a work already exists on arXiv then any new versions should be replacements of the existing arXiv version."12
Replacements have their own limits. The replacement page11 says: "We ask that articles be replaced no more than once per week. After version 5 replacements are limited to no more than once per week." Each replacement is a new public version, and "Previous versions of your article will remain publicly available." Withdrawal is no shortcut: "It is not appropriate to withdraw a paper because it is being updated."14
Where can you post while you wait for next month?#
Other repositories and preprint servers post papers under their own rules. Our endorsement guide compares nine of them in one table, checked in October 2026. If you plan to post the same paper on arXiv later, read each server's terms on licences and versions first.
What should you keep on file?#
Keep each version of the manuscript as submitted, with its date; the submission identifiers and the emails arXiv sends; who submitted which paper, and each co-author's agreement; and the AI-use statement each version carried, with tool and model names as the app or service reported them. An appeal or a support ticket will need them.
What is still unknown about arXiv's rate limit?#
- How long it lasts. Sigurdsson told THE: "We're not done yet. We are trying to be responsive and adapt our policies as necessary", and "There's much more coming in the near future."24 No end date or review date is published.
- How it applies across accounts. arXiv allows one account per person and says nothing more on this.
- Whether it reduced low-quality submissions. We found no figures from arXiv on declined papers, holds or support tickets after September, and the public count appears to leave declined papers out.
- The open cases. Cross-lists added later, withdrawals, unsubmitted papers, proxy and conference submissions, the time zone that ends a month, and any exception process.
- The analysis THE reported. It is unpublished, so its method and figures cannot be checked.
What this means for a research record#
A monthly limit makes the dates and people around a paper part of its history: who submitted which version, on which day, and what it said about AI use.
On Papex, an author can sign one exact version of a paper with its AI-use statement, and the signed version carries its date. The record shows who signed which version, when, and what they said the tools did, with model names as declared. A signature and a date are not peer review, and they do not show that a paper is correct; an AI is never listed as an author.
Sources#
- Fair Moderation, Equitable Access, and AI: arXiv's Updated Rate Limit Policy (arXiv blog, Kat Boboris, 1 October 2026), checked 17 November 2026 a b c d e f g h i j
- The same post, Internet Archive copy of 1 October 2026 (Internet Archive), checked 17 November 2026 ↩
- arXiv moderation, including the submission rate and the policy for authors' use of generative AI language tools (arXiv help), checked 17 November 2026 a b c d e
- arXiv moderation, Internet Archive copy of 30 September 2026 with the earlier section on excessive submission rate (Internet Archive), checked 17 November 2026 ↩
- Monthly Submissions, with totals as of 16 November 2026 (arXiv statistics), checked 17 November 2026 a b
- Monthly submissions data file, CSV (arXiv statistics), checked 17 November 2026 ↩
- arXiv sets new record for monthly submissions (again)! (arXiv blog, 4 November 2024), checked 17 November 2026 a b
- Version Availability (arXiv help), checked 17 November 2026 ↩
- arXiv:2610.00001, abstract page with submission history (arXiv), checked 17 November 2026 ↩
- arXiv:2610.42152, the highest October 2026 identifier on 17 November (arXiv), checked 17 November 2026 ↩
- Submit a new version of a work (arXiv help), checked 17 November 2026 ↩
- Proxy / Third Party Submission (arXiv help), checked 17 November 2026 ↩
- Status Information (arXiv help), checked 17 November 2026 ↩
- Withdraw / Retract a Submission (arXiv help), checked 17 November 2026 ↩
- Submission Overview (arXiv help), checked 17 November 2026 ↩
- Metadata for Required and Optional Fields (arXiv help), checked 17 November 2026 ↩
- Identity and Affiliation (arXiv help), checked 17 November 2026 ↩
- arXiv Code of Conduct Enforcement (arXiv help), checked 17 November 2026 ↩
- Why is my submission on Hold? (arXiv User Support Knowledge Base, last edited 7 April 2021), checked 17 November 2026 ↩
- Attention Authors: updated endorsement policy (arXiv blog, 21 January 2026), checked 17 November 2026 ↩
- Attention Authors: Updated Practice for Review Articles and Position Papers in arXiv CS Category (arXiv blog, 31 October 2025), checked 17 November 2026 ↩
- Thomas G. Dietterich, first post of a thread to arXiv authors (X, 14 May 2026), checked 17 November 2026 (read through X's public embed service, which shows the first post only) a b
- Ban for authors submitting AI content 'welcome but unenforceable' (Times Higher Education, 20 May 2026), checked 17 November 2026 ↩
- How can research platforms deal with the 'AI slop' deluge? (Times Higher Education, Frances Jones, 3 November 2026), checked 17 November 2026 a b c
- Submissions to preprint server keep rising despite author cap (Times Higher Education, 11 November 2026), checked 17 November 2026 a b
- News from arXiv, list of posts (arXiv blog), checked 17 November 2026 ↩
- Attention Authors: Temporary changes to announcement schedule due to upcoming US federal holiday (arXiv blog, 6 November 2026), checked 17 November 2026 ↩