Anthropic says Australia now ranks first globally for Claude usage on a population-adjusted basis, with local users recording more than six times the usage level the country's population size would predict.
The finding comes from the latest edition of Anthropic's Economic Index, a research series that tracks how people use Claude across work and personal tasks. The latest report, titled Cadences, also found that 54.5% of Australian Claude usage falls into what Anthropic classifies as "augmentation", where users stay involved in the task and use the model as a collaborator rather than handing over the entire job. That compares with a global average of 51.4%.
Daily rhythms
The report examines AI adoption through three lenses: when and how people use Claude, what outputs those interactions produce, and how users think AI is affecting their work.
Anthropic said its updated data pipeline now captures usage at a much higher frequency, allowing it to track patterns down to the hourly level. The company also introduced a new classifier to label the output of each conversation and published more detailed data covering chat, Cowork and API usage.
One of the report's main findings is that Claude usage follows the rhythms of daily life and the working week. Anthropic found that requests for news tend to peak around 7am local time, business correspondence and email drafting rise through the workday with a mid-morning peak, recipe requests spike at 6pm, and sleep-related questions cluster in the early hours before dawn.
The report also found that work-related activity declines at weekends while personal use rises. Across chat and Cowork conversations, personal use rose from about 35% on weekdays to just under 50% on weekends during the sample period. Outside the workweek, users were more likely to ask for emotional support, medical advice and investment help, while weekday activity leaned more heavily towards business correspondence, marketing copy and slide decks.
Anthropic also observed that tax-related conversations surged around the US tax filing deadline in April, with tax-related requests on 14 April running at eight times the average level seen in May.
Work outputs
A central part of the report is Anthropic's attempt to classify what Claude actually produces for users. It refers to the main output of a session as an "artifact", covering material such as explanations, reports, documents, apps, scripts, code or presentations.
Anthropic said its classifier identified an artifact in 93% of Claude conversations. The most common outputs were explanations, which accounted for 17% of conversations, followed by documents and reports at 15%, and guidance at 11%.
For work-related conversations specifically, documents and reports were the most common output at 20%, followed by explanations at 9%, email drafts at 7%, and analyses and summaries at 6%.
The report also argues that the computational cost of a conversation tends to rise with the value of the work being done. Anthropic measured this through token usage and mapped tasks to occupations. It found that conversations tied to higher-wage occupations generally consumed more tokens, produced more output from Claude, and involved more back-and-forth from the user.
Anthropic said this pattern suggests AI is being used in a labour-augmenting way in many higher-value tasks, with people staying involved rather than stepping away from the work entirely.
Delegation split
The report also examines how much autonomy users give Claude. Anthropic measures this on a five-point scale from "none" to "extreme", with low-autonomy tasks including translation, question answering and calculations, and high-autonomy tasks including building apps, websites, games and presentations.
Across almost all output types, Anthropic found that Claude Code sessions involved more delegation than chat or Cowork sessions. It said the difference reflected both the types of tasks people bring to coding workflows and the way they interact with the model once there.
Anthropic also linked delegation style to sentiment about AI's impact on work. In a survey of about 9,700 Claude users, the company found that people who used Claude in more automated ways were more optimistic about AI's effect on future pay, job prospects and job security than those who used it more collaboratively.
At the same time, the survey found concern about job loss, especially among early-career workers. More than a third of respondents said it was likely or very likely that responsibilities would change significantly over the next 12 months, while 10% rated losing their own job as likely or very likely.
Australian picture
For Australia, Anthropic's note to media focuses on the country's high level of Claude adoption and its relatively collaborative pattern of use.
"Perhaps most interestingly, Australians use Claude more than 6x what our population would predict," Anthropic said in briefing material accompanying the report. "What also stands out is how we're using AI: 54.5% of Australian usage is 'augmentation', people staying in the loop, using Claude as a collaborator rather than handing off the whole task."
The company's broader report suggests that distinction matters. Across its global user base, Anthropic found that people often want AI to automate repetitive or low-value work while preserving human involvement in tasks that require judgement, context and creative control.
The report also found that users generally hope AI will support meaningful work rather than replace it outright, with the most common long-term aspiration being a model of work where people collaborate with AI while routine tasks are handled in the background.