CtrlK
The Anthropic Economic Index is a research initiative that tracks how Claude is used across occupations, tasks, and geographies in order to study AI's economic effects. Anthropic launched the initiative in February 2025 and has published recurring reports based on privacy-preserving analysis of Claude usage data.[1][2]
The project is not a benchmark in the usual model-evaluation sense. It is an economic measurement program built to show where AI is being adopted, which tasks are being augmented or automated, and how usage patterns change over time.[1][3]
Anthropic's March 2025 report explains that the company uses a privacy-preserving analysis system called Clio to map conversations to tasks in the U.S. Department of Labor's ONET database. In that report, Anthropic said the taxonomy covered about 17,000 ONET tasks and that the company analyzed one million anonymized Claude.ai conversations after the release of Claude 3.7 Sonnet.[2]
| Date | Report or milestone | Main focus |
|---|---|---|
| February 2025 | Initiative launch | Introduced the Economic Index program |
| March 27, 2025 | "Insights from Claude 3.7 Sonnet" | Changes in usage after Claude 3.7 Sonnet and extended thinking |
| September 15, 2025 | "Tracking AI's role in the US and global economy" | Geographic and occupational adoption patterns |
| March 24, 2026 | "Learning curves" | Adoption over time, task diversification, and evidence consistent with learning by doing |
The March 2025 report said Claude 3.7 Sonnet increased the share of usage devoted to coding, education, science, and health-related tasks. It also found that extended thinking mode was used disproportionately for technical work such as software development and research.[2]
The March 2026 report said Claude.ai usage had diversified relative to earlier data, while more experienced users appeared to bring more complex and more collaborative tasks to Claude. Anthropic argued that the results were consistent with learning by doing, though the report also discussed survivorship bias and cohort effects as competing explanations.[3]
The Anthropic Economic Index has become one of the more prominent company-run attempts to study real-world AI adoption using observed product usage rather than surveys alone. For researchers and policymakers, its value is that it tracks actual AI-assisted work as deployment changes over time.[1][2][3]