Claude Usage Diversifies, Learning Curves Emerge

Anthropic's latest look at AI's economic footprint, the Anthropic Economic Index report for February 2026, reveals a subtle but significant shift in how Claude is being deployed. The report, utilizing privacy-preserving data analysis, tracks AI integration and its economic impact, aiming to equip researchers and policymakers with early insights. This latest analysis builds on previous findings, focusing on usage patterns three months after the release of Claude Opus 4.5 and coinciding with the debut of Claude Opus 4.6.

Usage Diversification and Shifting Task Value

The data indicates a slight increase in 'augmentation' – collaborative interactions where Claude complements user abilities – across both Claude.ai and API traffic. Notably, on Claude.ai, the concentration of top tasks has decreased, suggesting a broader range of applications. This diversification has led to a marginal decrease in the average 'wage value' of tasks performed, as more personal queries around sports, product comparisons, and home maintenance are now being logged.

Coding tasks continue their migration from Claude.ai to the more automated workflows of Anthropic's first-party API. While coding remains a dominant use case, its share within Claude.ai's top 10 tasks has fallen from 24% in November 2025 to 19% in February 2026. Conversely, personal use cases saw a rise on Claude.ai, accounting for 42% of conversations, up from 35% previously.

The average economic value of work done via Claude.ai has seen a slight dip, from $49.3 to $47.9 per hour, a trend attributed to the rise in simpler queries and the continued shift of coding to the API.

Learning Curves and User Experience

A key finding highlights the emergence of learning curves in Claude adoption. Users with higher 'tenure' – those who have been using the platform longer – are demonstrating more effective engagement. These experienced users not only attempt higher-value tasks but also achieve higher success rates in their conversations, a phenomenon not explained by task selection or origin.

Evidence suggests that seasoned users develop strategies to better harness Claude's capabilities. For instance, they are more likely to select the most advanced Opus model for complex, higher-wage tasks. This 'learning-by-doing' effect appears to be a significant factor in maximizing the benefits derived from AI tools.

Global and Domestic Adoption Trends

Inequality in global Claude usage persists, with a slight increase in concentration among the top 20 countries, now accounting for 48% of per capita usage. However, within the United States, usage is converging, with the top 10 states decreasing their share of per capita usage from 40% to 38%.

This convergence trend within the US is slowing, suggesting that full parity in usage per capita may take between five to nine years at the current pace.