- AI research now accounts for over 57% of all computer science publications
- China leads with more than 83,000 AI papers, followed by the US and India
- India shows the fastest growth in AI research volume globally
- AI publication share continues to climb, reaching record highs in 2025
- Data sourced from OpenAlex and filtered using the CSO Classifier to isolate AI-related research.
Read below for the methodology. You are allowed to use the data. Please consider citing this page.
The number of AI publications has climbed from around 106,000 to over 264,000 in recent years. After a period of steady output, growth accelerated sharply, especially from 2017 onward. This reflects increasing global investment and interest in artificial intelligence across both academia and industry.
China leads the world in AI research output, publishing over 83,000 papers, more than twice as many as any other country. The United States follows with a steady volume of publications, while India ranks third and shows the fastest growth among the top contributors. India’s AI research volume has more than tripled in recent years, reflecting its expanding focus on artificial intelligence across academia and industry.
Other countries such as the United Kingdom, Germany, Japan, and South Korea also play a significant role in global AI research. Their consistent publication levels indicate strong institutional investment and long-term engagement with AI. Together, these countries highlight the geographic concentration of AI research across Asia, North America, and Europe.
The share of AI publications within computer science has grown from 36 percent to 58 percent. After years of stability, the shift began around 2017 and has accelerated since. AI now represents the majority of all CS research, showing how central it has become to the field.
Methodology
This analysis draws on structured metadata from OpenAlex, covering research outputs between 2008 and 2025. We focused specifically on works categorized under the “Computer Science” concept to ensure consistency across disciplines. To maintain data quality and relevance, only journal articles with an available abstract were included. This helps exclude short formats or incomplete records that could distort topic classification.
To identify which of these publications were related to artificial intelligence, we applied the CSO Classifier. This tool uses the Computer Science Ontology (CSO), a domain-specific taxonomy developed to detect AI and related subfields through semantic analysis of abstracts and titles. By matching research papers to AI-specific terms in the CSO, we generated a consistent and scalable estimate of AI’s footprint within the broader field of computer science.
For your questions, you add me on Linkedin or email at enes at caracal.news
