A researcher publishing one AI paper in 2014 had real company. About 1,200 others did the same that year, across the field’s top conferences. By 2024, that number hit over 12,000. A tenfold jump in a decade.
How Many AI Papers Get Published Now?
The scale is hard to overstate. Researchers tracking arXiv found 170,927 AI papers posted between the start of 2025 and mid-2026 alone, across just four core machine-learning categories.
That period alone showed roughly 25% growth in paper volume. And this is one platform, not the entire research world.
Conference publishing tells a similar story over a longer stretch. Across five leading AI and ML venues, AAAI, ICLR, ICML, IJCAI, and NeurIPS, publications grew from 1,206 papers in 2014 to 12,026 in 2024.
AI Research Papers by the Numbers
| Metric | Figure |
| AI papers on arXiv, Jan 2025–June 2026 (core categories) | 170,927 |
| Growth in paper volume over that window | ~25% |
| Top-5 AI conference papers, 2014 | 1,206 |
| Top-5 AI conference papers, 2024 | 12,026 |
| Growth in top conference output, 2014–2024 | 10x |
| Growth in arXiv AI papers, 2010–2019 | 20x+ |
| Share of all peer-reviewed papers that were AI-related (2018) | ~3% |
| Growth in machine learning arXiv output, 2015–2020 | 10x |
| Growth in robotics arXiv output, 2015–2020 | 11x |
| Mentions of “agentic workflows” in papers, 2025 | 4,585 → 10,496 |
| Growth in “long-horizon planning” papers, 2025 | +510% |
Different trackers measure different slices of the field, conferences, arXiv, or specific subfields. The exact count varies. The growth curve doesn’t.
Where the Growth Is Actually Coming From
Academia Still Leads, But Industry Is Catching Up Fast
Academic output keeps growing every year, and it hasn’t slowed. But industry publishing has accelerated even harder in recent years, at points growing faster year-over-year than academic output itself.
Mixed academic-industry collaborations are growing too, a sign that the line between a university lab and a corporate research team keeps blurring.
The Subfields Driving the Boom
Machine learning and computer vision remain the two largest categories on arXiv, together accounting for the majority of AI-related submissions in recent years.
The fastest-growing corner right now is agentic AI research specifically. Mentions of agentic workflows in papers more than doubled in a single year. The sub-topic of long-horizon planning, getting a model to pursue a goal across many steps without losing track, grew faster than anything else measured, over 500% in one year alone.
That trend lines up closely with what’s happening outside research papers too. A closer look at agentic AI covers how these systems are already moving from academic papers into real, deployed enterprise tools.
A Complication Worth Knowing: AI-Written Papers
Here’s a wrinkle most coverage skips. As AI tools got better at writing, the number of survey papers, review-style papers summarizing a research area, exploded on arXiv.
Researchers studying this trend flagged something specific. A growing share of these survey papers show strong signs of being AI-generated or AI-assisted themselves, based on content-detection analysis. Some authors were found submitting more than three survey papers in a single month with fewer than two collaborators, a pattern researchers describe as a strong red flag for automated output.
That doesn’t mean every new paper is suspect. It does mean paper volume alone is becoming a less reliable signal of real research progress than it used to be.
Where to Actually Find AI Research Papers
arXiv.org remains the default home for AI preprints, free, fast, and searchable by category. Top conference proceedings, NeurIPS, ICML, ICLR, and similar venues, carry peer-reviewed work with a higher editorial bar than a preprint server. Semantic Scholar and Google Scholar both index across sources and make citation tracking easier than browsing a single repository. For a broader, standardized view of where the field stands each year, the Stanford AI Index Report remains one of the most widely cited annual benchmarks tracking research output, funding, and adoption trends across the entire industry.
Reading AI Papers Without a PhD
A few habits make dense research papers more approachable for a non-specialist.
Read the abstract and conclusion first. They usually summarize the actual finding without the methodology detail. Check the institution and author history. A well-established lab or a highly cited author is a reasonable, if imperfect, quality signal. Look for a plain-language summary. Many papers now include one, and some platforms generate them automatically. Cross-check big claims. A single paper rarely settles a debate on its own, however confident its abstract sounds.
Publication Trends by Region
Research output has also shifted geographically over the past decade. US and China-based institutions dominate the most prolific AI research organizations, and collaboration between the two remains active despite broader competitive tension between them. Universities across Asia have taken on a larger share of academic publishing in recent years, a shift that’s changed the balance of who’s actually driving the field forward.
Final Thoughts
AI research has gone from a specialized academic niche to one of the fastest-growing publishing categories in modern science. Conference output alone is up tenfold in a decade. Preprint volume keeps climbing even faster.
The honest caveat is that raw paper counts don’t tell the whole story anymore. As AI tools help write more of the papers themselves, judging real progress takes more than just counting how many show up each month.
The field isn’t just studying AI faster. It’s increasingly using AI to study itself, and that shift is worth watching as closely as the growth numbers themselves.
