Preprint server arXiv struggles under flood of AI-generated papers, impacting research accessibility

The arXiv preprint repository has been overwhelmed by an exponential surge in AI-generated papers, with submissions tripling from approximately 20,500 in September 2024 to over 40,000 in September 2026, threatening to render the database unusable for legitimate scientific research. Mathematicians report that the flood of low-quality AI-generated submissions has slowed peer review processes, making it difficult for genuine papers to be published promptly, and has forced researchers to rely on AI tools themselves to identify which papers are worth reading. In response to the crisis, arXiv has implemented submission limits to stem the tide of low-value contributions, though concerns remain about the long-term viability of the service.

The submission surge to arXiv represents a dramatic acceleration in the scale of the problem. Over a two-year period, monthly submissions more than doubled, with the growth rate itself increasing sharply in the final year. This exponential trajectory suggests the database's capacity management systems were designed for a previous era of scientific publishing. The core issue extends beyond volume: AI-generated papers often lack the clarity and pedagogical quality that human researchers expect, even when mathematically sound. Researchers must now dedicate resources to filtering rather than discovering, inverting the original purpose of the preprint service.