[Digital Today reporter Jinju Hong (홍진주)] Anthropic released a report analysing actual usage patterns among Claude users.

On June 29 (local time), online media outlet Gigazine reported that the study compiled and analysed usage data for the chat service Claude.ai, the developer tool Claude Code, and APIs. It summarised differences by time of day and day of week, purpose of use and job category.

The most notable finding was a clear split between weekday and weekend purposes. Claude.ai had a higher share of personal use than Claude Code or APIs. All three services showed an increase in personal use on weekends compared with work use. Anthropic explained that a common pattern appeared across usage types, with the share of personal use rising on weekends.

Differences by time of day were also clear. Everyday conversation increased from night into the morning, then declined during the daytime. By contrast, work-related use increased from morning through the afternoon.

By topic, news-related questions were most frequent in the morning. Sleep-related consultations surged after midnight and peaked around 5 a.m. Searches for cooking recipes were most active at 6 p.m., showing a tendency to align with dinner preparation time.

The report also confirmed cases where real-world schedules directly affect AI usage volumes. In the United States, tax-related questions surged just before the April 15 tax filing deadline. The increase among U.S. users was much larger than among overseas users, showing that actual social schedules are reflected directly in AI usage patterns.

Differences also appeared by gender. Men accounted for a higher share of work-purpose use, use of Claude Code and automation-related use. Women, however, spent more total time using the services, showing different characteristics between usage frequency and time spent.

Usage behaviour also differed by wage level. Lower-income users (Q1, Q2) had a higher share of weekday usage, while higher-income users (Q3, Q4) tended to keep using Claude steadily even on weekends.

Characteristics also varied clearly by purpose. Recipe searches, writing and drafting various guides were more than 80 percent personal use. Marketing work, blog writing and generating database queries, among others, had work-purpose shares of more than 80 percent.

In education, use was most active for solving math problems and reading academic papers, the report found. The job-category distribution also differed from the general labour market.

Usage shares among workers in computer and mathematics fields and among managers were higher than their overall shares of employment in the United States. By contrast, transportation, food processing and hospitality services showed relatively lower shares of AI usage.

Anthropic said the report is meaningful because it goes beyond a simple count of usage volume and shows when and for what purposes AI is used in real life and at work. It said that by analysing Claude.ai together with Claude Code and APIs, it was able to confirm in detail how personal conversational AI use and work and automation use differ by time of day, job category and income level.