How Canada uses Claude: Findings from the Anthropic Economic Index

Adoption of Claude is high in Canada. Based on a sample of Claude.ai conversations in February 2026, 2.6% of global traffic is in Canada. Adjusting for population, its Anthropic AI Usage Index (AUI) is 4.4, which implies that usage per capita is more than four times higher than would be expected based on its working-age population. Among the top ten countries that lead in terms of overall Claude usage volume, Canada has the second-highest AUI, just behind the United States (Figure 1).

In part, this reflects the fact that Canada is a high-income country. Among advanced economies, as defined by the IMF, there is a clear link between usage per capita and GDP per working-age capita. But Canada’s adoption exceeds what would be expected based on its income (Figure 2). In this sense, Canada appears further along on its AI adoption curve as compared to peer countries, perhaps reflecting its highly educated workforce and proximity to the US technology frontier.

Just as global adoption of Claude is disproportionately concentrated among a small number of countries, usage within Canada is unevenly distributed across provinces. Ontario accounts for 43.9% of conversations. Another 50% of usage within Canada comes from Quebec (20.8%), British Columbia (18.9%), and Alberta (10.2%). Adjusting for population, the ordering shifts: British Columbia has 1.4x more Claude usage than expected based on its working age population, followed by Ontario (1.1x). The remaining provinces have below-parity rates of adoption. The province of Newfoundland and Labrador has an AUI of 0.2.

While broad usage patterns are consistent across provinces, we nevertheless find evidence that particular use cases tend to coincide with local economic characteristics. For example, we find that use of Claude for translation and editing requests is systematically higher in parts of the country with a larger share of public administration employment—perhaps reflecting Canada’s commitment to official bilingualism in the public sector, which generates demand for English–French translation wherever government employment is concentrated. Indeed, New Brunswick, Nova Scotia, and Quebec each have the highest rates of public administration employment and the largest share of conversations related to translation assistance.

Particular use cases also reflect the structure of the Canadian economy. Translation is more common in provinces with more public administration employment, consistent with the official-languages requirements that apply across much of Canada's public sector—and document translation is the single most distinctive Canadian use of Claude relative to its Anglosphere peers. More broadly, Canadian usage tilts toward education and labor-market entry, with academic coursework, coding assistance, and resume drafting all overrepresented relative to Australia, the United Kingdom, and the United States.

What explains the pattern of regional adoption? We explore this question in Figure 4. In contrast to the global relationship between income and usage, we find that provincial usage per capita is mostly uncorrelated with income. Instead, we find that the industrial composition of regional economies is a key determinant of usage intensity: regions with larger professional, scientific, and technical service sectors—as measured by the share of employment—have systematically higher usage per capita.

Evidently, workforce composition—rather than income per se—is a key determinant of within-Canada patterns of adoption. This evidence is consistent with our earlier findings concerning adoption diffusion within the United States. It appears that AI adoption in high-income countries is primarily shaped by how well matched model capabilities are to the structure of the local economy.

Despite the large regional variation in rates of Claude usage, people across Canada tend to use Claude in broadly similar ways. In every province, personal use—looking up health information, researching products, getting help with recipes or home repairs—accounts for 44% to 51% of conversations. Work-related use, where Claude is used to troubleshoot software, draft work emails, and build business applications, accounts for around 34% to 40%. What remains is using Claude to assist with coursework-related tasks (Figure 5).