Australia is using AI for work less than you think. Forty-seven percent of conversations on Claude.ai from Australian users are classified as personal, putting Australia near the top of the distribution for personal use among English-speaking countries, according to Anthropic's research page on how Australia uses Claude. Work conversations sit at 46 percent, coursework at 7 percent. That near-parity between personal and work use is a useful corrective to the enterprise-first framing that dominates most AI adoption research.
Australia, a country of 26 million people and no frontier AI labs of its own, is using Claude at a rate that places it seventh globally on a per-capita basis, according to the same Anthropic data. Its Anthropic AI Usage Index, a measure of usage relative to working-age population, stands at 4.1, meaning Australians use the model over four times more than their demographics would predict.
The distribution within the country is uneven. New South Wales and Victoria, Australia's two most populous states, account for 37.2 and 30.8 percent of conversations respectively, with Queensland at 17.7 percent and the remaining states and territories splitting 14.3 percent. Only NSW (AUI 1.20) and Victoria (AUI 1.19) exceed the national baseline of 1.0. Western Australia, home to the mining sector, scores 0.68. Tasmania sits at 0.32 and the Northern Territory, sparse and remote, at 0.12. The gap between the territory's adoption rate and NSW's is tenfold.
The autonomy score for Australia is 3.38 on a 1-to-5 scale, below the global average. Australians, in other words, are less likely to use AI as a delegating tool and more likely to use it as a collaborative one: asking questions, working through problems, keeping the human in the loop. That framing fits with the task complexity data. Australian prompts require the equivalent of 11.9 years of schooling to process, and the tasks they're applied to would take an estimated 2.7 hours without AI assistance, versus a 3.3-hour global average. Australians are using Claude for work that is genuinely time-consuming and genuinely complex.
The use case distribution is also more diverse than in comparable countries. The top 100 task categories account for 47.3 percent of Australian usage, compared to 52.3 percent globally, and below the US, UK, and Canada. The standout sector difference: Computer and Mathematical tasks run 8 percentage points below the global baseline, a larger gap than any other category. The offsetting positives are spread across Management, Office and Administrative Support, and Life, Physical, and Social Science occupations. In plain terms: Australians are not using Claude primarily to write code. They are using it to manage, analyze, and reason.
Australia accounts for 1.6 percent of global Claude.ai traffic, ranking eleventh among all countries, according to Anthropic. Seventh per-capita, eleventh in absolute terms. That gap is population, not a ceiling.
On March 31, Anthropic CEO Dario Amodei met with Prime Minister Anthony Albanese in Canberra to formalize a memorandum of understanding on AI safety and economic data tracking, Reuters reported. The deal mirrors agreements with safety institutes in the United States, Britain, and Japan. Australia currently has no specific AI legislation, and the centre-left Labor government has said it would rely on existing laws to manage emerging AI risks while introducing voluntary guidelines. Anthropic's release of detailed Australian usage data alongside the MOU is not a coincidence. It is the company's evidence that its model is being used productively and broadly in a country with no mandatory AI regulations, and no particular reason to favor one frontier lab over another except the quality of the product.
The natural experiment is running. Australia, a middle-power English-speaking country with a relatively small population, a common-law legal tradition, and no frontier AI labs of its own, is using Claude at rates that put it in the top tier globally. The per-capita ranking is the more informative number. When you adjust for population, a country of 26 million people punches well above where its absolute traffic numbers suggest it should. What happens to that usage pattern as regulatory frameworks develop, as they eventually will, is the open question the MOU begins to address.
The dataset Anthropic published for Australia follows the same approach it has used for the US, UK, Canada, Germany, and France. GNews covered the Anthropic research release, though the company's own research page is the primary source for the underlying data.
