(Co-authored with Lekha Chakraborty)

Artificial intelligence (AI) is no longer a distant promise—it is reshaping economies, labour markets, and societies at unprecedented speed. Yet, reliably measuring its spread remains challenging for policymakers. How do we capture not just frontier breakthroughs but also the broader diffusion that determines who benefits?

Existing indices offer valuable insights, but a balanced gauge combining innovation, investment, and structural readiness can provide a clearer picture. Our study introduces the AI Composite Index, a measure designed to track AI diffusion across countries in a policy-relevant way ( Dubey and Chakraborty (forthcoming).

The literature on measuring AI has grown rapidly, reflecting the technology’s rising importance. Early efforts focused on specific dimensions, such as patent activity as a proxy for innovation (WIPO, various years). More comprehensive approaches emerged with the Stanford Human-Centered Artificial Intelligence (HAI) AI Index, which since 2017 has tracked trends in research, performance, investment, and ethics. The 2025 edition highlights record private AI investment exceeding $250 billion globally in 2024, with generative AI attracting strong momentum and the United States dominating funding (Stanford HAI, 2025).

A key contribution comes from the International Monetary Fund. In their analysis of generative AI and work, Cazzaniga et al (2024) introduced the AI Preparedness Index (AIPI), scoring 174 economies on digital infrastructure, human capital, labour policies, innovation, and regulation. Advanced economies average high readiness, while emerging markets and low-income countries lag, underscoring divides in capturing AI gains.

Venture capital trends feature prominently in OECD analyses, though recent data from broader sources reveal explosive growth. While OECD reports noted steady increases through 2020, with the United States and China absorbing most funds (OECD, various years), updated figures show global AI private investment hitting new highs—over $250 billion in 2024 alone, driven by megadeals (Stanford HAI, 2025; various reports, 2025).

Building on these foundations, the AI composite Index integrates three core pillars to capture both supply-side leadership and enabling conditions for broad adoption.

First, AI patent registrations per million population from the World Intellectual Property Organization (WIPO). Patents signal frontier innovation, with generative AI filings surging from fewer than a thousand annually in the mid-2010s to over 14,000 by 2023 (WIPO, 2024). Normalised by population, this indicator highlights leaders like South Korea and China while adjusting for scale.

Second, venture capital investment in AI as a share of GDP, drawing on OECD and complementary sources. This reflects private-sector confidence and commercialisation potential. Recent booms—U.S. AI funding alone reaching over $100 billion in 2024—show concentration in key players, but scaling by GDP allows fairer comparisons across economies (Stanford HAI, 2025).

Third, the IMF’s AI Preparedness Index (Cazzaniga et al., 2024). This broader gauge covers structural preconditions essential for effective diffusion, including broadband access, STEM skills, regulatory adaptability, and labour mobility. It bridges innovation metrics with real-world enablers. To construct the index, skewed distributions in patents and investment are addressed through logarithmic transformation for better balance. Each component is then standardised to have comparable spread, ensuring no single pillar dominates due to measurement differences. The index takes a simple average of the three, with equal weights for transparency and to avoid strong prior judgments on relative importance. Finally, scores are rescaled from zero to one for easy interpretation, without changing country rankings. Robustness checks confirm stability.

Applying principal component analysis extracts a main factor with balanced contributions from all pillars, yielding rankings nearly identical to the baseline (correlations above 0.99). Alternative approaches, like min-max scaling, produce similarly consistent results across specifications.

The resulting index reveals sharp global divides. Advanced economies with strong innovation ecosystems—high patents, surging venture flows, and solid preparedness—score toward the top, positioning them to reap productivity and growth benefits sooner. Emerging markets often excel in one area, such as patent volume in China, but lag in preparedness, risking slower or uneven adoption. Lower-scoring nations face barriers in infrastructure and skills, potentially widening gaps. The IMF’s preparedness focus echoes this, suggesting advanced economies prioritise regulation and reskilling, while others invest in foundations (Cazzaniga et al., 2024).

For policymakers, the index offers actionable insights. Tracking these dimensions helps identify bottlenecks—whether low investment signaling weak entrepreneurial confidence or preparedness gaps hindering broad uptake. Future enhancements could incorporate user adoption metrics or gender-disaggregated data to better capture inclusive diffusion.

In a fast-evolving field, systematic measurement is vital. By blending established indicators into a transparent composite, the AI Composite Index provides a practical tool for monitoring progress and designing policies that make AI benefits widely shared.
References

Cazzaniga, Mauro, Florence Jaumotte, Longji Li, Giovanni Melina, Augustus J. Panton, Carlo Pizzinelli, Emma Rockall, and Marina Mendes Tavares. 2024. Gen-AI: Artificial Intelligence and the Future of Work. IMF Staff Discussion Notes 2024/001. Washington, DC: International Monetary Fund.

Dubey, Rohan and Lekha Chakraborty (forthcoming). Measuring AI for its impacts on human capital: Empirical evidence from G20 countries, mimeo, National Institute of Public Finance and Policy, New Delhi: NIPFP 

Organisation for Economic Co-operation and Development (OECD). Various years. Venture Capital Investments in Artificial Intelligence. Paris: OECD. https://oecd.ai/en/vc.

Stanford Institute for Human-Centered Artificial Intelligence (Stanford HAI). 2025. AI Index Report 2025. Stanford, CA: Stanford University.

World Intellectual Property Organization (WIPO). 2024. Patent Landscape Report: Generative Artificial Intelligence. Geneva: WIPO.

World Intellectual Property Organization (WIPO). Various years. IP Statistics Data Center. Geneva: WIPO. https://www.wipo.int/ipstats/.