GAIA — Global AI Adoption Index

GAIA (Global AI Adoption Index) is an open dataset measuring artificial intelligence exposure and adoption across occupations and countries. It combines three independent data sources into a unified, freely accessible index designed for researchers, policymakers, and development finance practitioners. Occupation-level data (923 occupations): The dataset includes three independent AI exposure measures for 923 US Standard Occupational Classification (SOC) occupations. The first is observed real-world usage from the Anthropic Economic Index (CC-BY), capturing actual Claude.ai task usage aggregated by occupation globally. The second is theoretical exposure from Eloundou et al. (Science 2024), using GPT-4 and human annotation of 18,000+ O*NET work tasks to estimate what share of each occupation's tasks a large language model could perform — both directly (E1) and with software tools (E1+E2). The third is a pre-GPT baseline from Brynjolfsson, Mitchell & Rock (AEA 2018), measuring suitability for traditional machine learning constructed in 2018 before ChatGPT, used as a falsification test in causal identification. Country-level behavioral data (138 countries): The dataset includes a country-level panel from the Anthropic Economic Index covering 138 countries with behavioral breakdowns of Claude.ai usage during the week of February 5–12, 2026. Variables include share of sessions classified as work, personal, and coursework use; collaboration styles (directive, learning, task iteration, feedback loop); and task success rates. This is the first publicly available country-level behavioral dataset on AI tool usage. Files included: gaia_occupations.csv — 923 occupations with OpenAI E1, E1+E2, human-annotated versions, SML score, composite index, major occupation group gaia_countries.csv — 138 countries with usage share, use case breakdown, collaboration style, task success, income group classification Companion paper: This dataset was built as the empirical backbone of Aghabarari, L. & Van Doornik, B. (2026). AI Exposure, Credit Markets, and Employment in Brazil. Working paper. The paper uses Brazil's 20-year credit registry to trace how AI exposure affects firm financing and labor market outcomes via a difference-in-differences design around the November 2022 generative AI shock. Data sources and licenses: Anthropic Economic Index: CC-BY 4.0. Handa et al. (2025). Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations. huggingface.co Eloundou et al. (2024). GPTs are GPTs: Labor Market Impact Potential of LLMs. Science 384(6702), 1306–1308. DOI: 10.1126/science.adj0998 Brynjolfsson, Mitchell & Rock (2018). What Can Machines Learn, and What Does It Mean for Occupations and the Economy? AEA Papers and Proceedings 108, 43–47. Replication data: ICPSR 114436, CC BY 4.0 Citation: Aghabarari, L. (2026). GAIA — Global AI Adoption Index. Zenodo. [DOI will appear after publication]