arXiv's AI-Generated Papers Problem: Why It Now Caps You at Two a Month, and How to Spot Slop

In September 2026, arXiv, the free preprint server where most new AI research first appears, received 40,363 submissions. Two years earlier it was 20,569. The flood of AI-generated research papers is a big part of why, and on 1 October 2026 arXiv started limiting every submitter to two new papers a month.

That matters beyond universities. Journalists, investors and product teams treat "a new paper on arXiv shows..." as evidence, so if more papers carry fake references or unchecked results, everyone downstream inherits the problem.

Below: what arXiv changed, what the research says about hallucinated citations and AI reviewing AI, and a ten-minute checklist for judging any paper. If your team builds products on new AI methods, our LLM integration work starts with exactly this kind of check.

arXiv submission limit: two new papers a month from 1 October 2026

Since 1 October 2026, arXiv lets each submitter make two new submissions per calendar month, on top of its limit of three active submissions at once. Rejected papers count, co-authors aren't limited, and AI-assisted papers are still allowed if the AI use is disclosed. arXiv calls it a stopgap against a flood of submissions it can't moderate.

The details, from arXiv's blog post of 1 October 2026 (all submitters, all categories):

• Two new submissions per calendar month, on top of the 2024 limit of three active submissions at once.
• Rejected papers count. Papers deleted before announcement don't.
• Only the submitter is limited. Co-authors can still submit their own papers.
• AI use is still allowed if it is disclosed and the work meets arXiv's standards.

arXiv calls it a "stopgap" while it improves its moderation tools. The reason is volume: submissions have doubled in two years, and the cs.AI category grew more than six times over the same period.

September's record month produced almost 9,000 support tickets for arXiv's staff and moderators. According to Times Higher Education, arXiv works with more than 170 volunteer moderators.

arXiv reports more thin papers of narrow scope and "salami" papers that slice one study into several. Thomas Dietterich, who chairs arXiv's Editorial Advisory Council, told 404 Media that "a relatively small proportion of authors are submitting a large number of low-quality papers", which he called "unfair to authors who continue to submit quality papers."

Not everyone agrees. Harvard Business School's Scott Kominers argued that arXiv should "economise on moderation" rather than on scholarship. Rutgers mathematician Alex Kontorovich welcomed it and asked authors to pick their best two papers.

Am I affected? What this means for you

Run through these four questions:

1. Do you post to arXiv? Plan around two new papers a month, and remember a rejection uses up one of them.
2. Do you read or share "new paper shows" posts? You're relying on work that, on arXiv, has not been peer reviewed.
3. Does your product or roadmap rest on a recent paper's method or benchmark? You need to check it holds on your data before you build on it.
4. Do you train or fine-tune models on scraped research text? Fabricated references and AI-written text are now in that data.

If you answered yes to any of them, the 10-point checklist further down is for you. It takes about ten minutes per paper.

How arXiv has tightened on AI-generated papers: five moves in 12 months

The rate limit is the latest step, not the first.

Hallucinated citations: rare, but growing fast

A hallucinated citation points to a paper that doesn't exist, or pairs a real title with invented authors, venue or year. It's the clearest fingerprint of unchecked AI writing. Four 2026 studies measure it:

• Across 2.5 million papers. Zhao and colleagues audited 111 million references on arXiv, bioRxiv, SSRN and PubMed Central and estimated, conservatively, 146,932 hallucinated citations in 2025 alone. One co-author is Paul Ginsparg, who founded arXiv.
• In biomedicine. A Columbia team led by Maxim Topaz, writing in The Lancet, checked 97 million citations in more than 2 million papers. Papers with at least one fabricated reference went from 1 in 2,828 in 2023 to 1 in 458 in 2025, and 1 in 277 in the first seven weeks of 2026, as STAT reported.
• At top AI and security conferences. Russinovich, Kumar and Salem found that in 2025 roughly one in twenty NeurIPS and USENIX Security papers had at least two likely hallucinated references, even after peer review. Their automated audit cost about $0.04 per paper.
• In computing education. Denny and colleagues (arXiv 2609.16574, accepted to SIGCSE 2027) checked 24,751 papers. Verified hallucinated references at the SIGCSE Technical Symposium rose from 3 in 2025 to 17 in 2026, in 2.3% of 2026 proceedings papers.

The Denny paper shows what fakes look like. Of 30 confirmed hallucinated references, 13 were entirely invented and 17 attached a real title to the wrong authors, venue or year (one paper had seven).

So "the title exists" is not enough: check the authors too.

A reference list is the cheapest part of a paper to check, which is why a fake one tells you so much. If the authors didn't check the part that takes thirty seconds to verify, assume they didn't check the part that takes thirty days.

AI writes the paper, AI reviews it: the arms race

A survey posted on 7 September 2026, "The Emerging AI Paper-Review Arms Race" (arXiv 2609.07713), draws on 230 publications and records to map the cycle: cheaper papers strain review, review gets automated, authors game the automated reviewer, institutions add safeguards. Some of its evidence:

• Hidden instructions. A July 2025 audit found 18 arXiv preprints with concealed text asking AI reviewers for a favourable assessment.
• Gaming works. One study found a full-paper manipulation attack succeeded 75.1% of the time across three AI reviewer models, adding 1.21 points on a ten-point scale.
• AI reviewers miss real flaws. In one experiment, edits that broke a paper's soundness produced no significant change in AI review scores.
• Official AI reviews are here. At AAAI-26, all 22,977 main-track papers that entered full review received an AI-generated review.

Human reviewers use AI too. Nature reported in November 2025 that the detection company Pangram Labs judged about 21% of the 75,800 reviews for ICLR 2026 to be fully AI-generated.

ICML 2026 hid instructions in each submission PDF that only an AI would follow, asking it to include two phrases from a 170,000-phrase dictionary. That caught 795 reviews (about 1%) from 506 reviewers, each case checked by a person.

The survey's authors are candid: the evidence is strongest for scaled production, manipulation and policy responses, and weakest for long-term effects on the scientific record.

Why this matters if you never read a paper

• News. A viral post saying "a new paper shows X" may be summarising a preprint nobody has reviewed. The fix: open the paper (step 1 below).
• AI training data. Models learn from the open literature, so fabricated references and recycled text become tomorrow's training data.
• Products built on "the latest paper". If a reported score came from a benchmark the model had already seen, the gain may vanish on your data. When researchers built GSM1k, a fresh copy of a popular maths benchmark, some models scored up to 8% lower than on the original, a sign they had partly memorised it.

In "AI and the Enclosure of the Scientific Commons" (arXiv 2610.27251, 19 October 2026), P. K. G. Williams of the Center for Astrophysics, Harvard and Smithsonian, argues that AI both extracts value from shared knowledge and floods the systems that maintain it. His bluntest line: "It is now clear that if you give people generative AI tools and a network, they will spam slop."

How to tell if a research paper is AI generated (or just not solid): a 10-point checklist

You don't need a PhD, just ten minutes and a search engine. Use this on any paper, or any post that claims "a new study shows".

1. Open the actual paper. Find the arXiv or journal link, not a screenshot or a thread. If you can't find it, stop there.
2. Check three to five citations exist. Search each title in Google Scholar or follow its DOI. Then check the authors and year match, since half-fakes use real titles.
3. Check peer review status. arXiv states that its moderation "is not a peer-review process". Look at the Comments line for "Accepted to..." and confirm it in that venue's proceedings.
4. Look for code and data. Does the repository link work, and does it contain more than a README? Results nobody can rerun are claims, not findings.
5. Look up the authors. Click their names on arXiv. A real track record in the field is reassuring. Dozens of papers in a few months, across unrelated topics, is not.
6. Compare the claims with the results. Does the abstract's headline match the tables? Watch for one benchmark, one run, tiny gains and no error bars.
7. Ask about benchmark contamination. Is the test set old and public? Did the authors test on fresh or held-out data?
8. Scan for leftover AI text. Phrases like "Certainly, here is" or "as of my last update", or a note offering to revise the section, are exactly what arXiv's ban targets.
9. Check the version history. A v3 that quietly changes the main result, or a withdrawal notice, tells you a lot.
10. Don't treat an AI detector as a verdict. In one NeurIPS 2026 audit of 969 position papers, a default detector flagged 28.2% at its maximum AI score. A finer-grained analysis cut that to 12.7%.

A paper that passes most of these may still be wrong, but it's the kind of wrong science knows how to fix. For teams adopting a method, the last step is your own test: see how to evaluate a RAG chatbot, and OpenAI's math proofs explained for how verification works when claims are big.

AI-assisted writing done right: disclosure, not denial

None of this means researchers shouldn't use AI. arXiv explicitly allows it. Its rules ask three things:

• Disclose significant use of generative AI tools, in line with your field's norms for reporting methods.
• Take responsibility for everything, "irrespective of how the contents were generated", including errors, bias and fabricated references.
• Don't list an AI tool as an author.

In practice:

• Say what the tool did: "used to edit language in sections 2 and 4" beats a vague line.
• Verify every reference yourself against a DOI or the publisher's page.
• Keep your drafts and data, so you can show your work.
• Follow the venue's review rules. At ICML, breaking them cost reviewers their own papers.
• Publish one solid paper rather than three thin ones.

The same rules suit a company white paper or technical blog. For where AI fits in your own workflow, see RAG vs fine-tuning and AI agents for small business.

Frequently asked questions

The bottom line

arXiv didn't ban AI-assisted papers. It started rationing moderators' attention, because a few authors were using AI to submit more than anyone could check. Fake references are still rare per paper but rising fast, and AI reviewers can be gamed.

The fix for readers is unglamorous: open the paper, check a few references, find the code, and test it on your own data before you trust it. Start with step 2 on the next paper you're about to share.

Building a product on a method from a recent paper? Book a free consultation and we'll help you test whether its results hold up on your data before you commit engineering time to it.

Sources

• arXiv blog, Fair Moderation, Equitable Access, and AI: arXiv's Updated Rate Limit Policy (1 October 2026): https://blog.arxiv.org/2026/10/01/updated-rate-limit-policy/
• arXiv blog, Attention Authors: Updated Practice for Review Articles and Position Papers in arXiv CS Category (31 October 2025): https://blog.arxiv.org/2025/10/31/attention-authors-updated-practice-for-review-articles-and-position-papers-in-arxiv-cs-category/
• arXiv blog, Attention Authors: updated endorsement policy (21 January 2026): https://blog.arxiv.org/2026/01/21/attention-authors-updated-endorsement-policy/
• arXiv, Content moderation (accessed 26 October 2026): https://info.arxiv.org/help/moderation/index.html
• 404 Media, arXiv Is Rate Limiting Submissions Because It Can't Keep up With AI Slop (5 October 2026): https://404media.co/arxiv-is-rate-limiting-submissions-because-it-cant-keep-up-with-ai-slop
• Times Higher Education, arXiv's new preprint submission cap divides opinion (2 October 2026): https://www.timeshighereducation.com/node/745300
• Times Higher Education, Ban for authors submitting AI content 'welcome but unenforceable' (20 May 2026): https://www.timeshighereducation.com/news/ban-authors-submitting-ai-content-welcome-unenforceable
• Denny et al., Testing Our Foundations: Citation Trends, Errors, and Emerging Hallucinations in the Computing Education Literature, arXiv 2609.16574 (15 September 2026): https://arxiv.org/abs/2609.16574
• Wang et al., The Emerging AI Paper-Review Arms Race: Adversarial Co-Evolution in Scholarly Publishing, arXiv 2609.07713 (7 September 2026): https://arxiv.org/abs/2609.07713
• Williams, AI and The Enclosure of the Scientific Commons, arXiv 2610.27251 (19 October 2026): https://arxiv.org/abs/2610.27251
• Zhao et al., LLM hallucinations in the wild: Large-scale evidence from non-existent citations, arXiv 2605.07723 (8 May 2026): https://arxiv.org/abs/2605.07723
• Russinovich, Kumar and Salem, Phantom References: Hallucinated Citations That Survive Peer Review at Top-Tier Conferences, arXiv 2607.00738 (1 July 2026): https://arxiv.org/abs/2607.00738
• STAT, Fraudulent citations, blamed on AI hallucinations, are becoming more common in research papers (7 May 2026): https://www.statnews.com/2026/05/07/lancet-study-finds-steep-rise-fraudulent-citations-academic-papers/
• ICML blog, On Violations of LLM Review Policies (18 March 2026, updated 26 March 2026): https://blog.icml.cc/2026/03/18/
• Nature, Major AI conference flooded with peer reviews written fully by AI (27 November 2025): https://www.nature.com/articles/d41586-025-03506-6
• Zhang et al., A Careful Examination of Large Language Model Performance on Grade School Arithmetic, arXiv 2405.00332 (1 May 2024): https://arxiv.org/abs/2405.00332