Tracking AI’s economic impact

Just a year or two ago, AI was seen primarily as a tool for answering questions or speeding up research. Countless queries across far-flung corners of the internet, involving far too many open tabs, led to lots of clicking, skimming and summarising—and AI made that easier. But today, it's becoming much more than that: from classrooms to science labs, AI is starting to function as a true collaborator—helping people to write, code, create, plan and problem-solve in entirely new ways. This shift is not just about convenience or efficiency; it marks a deeper transformation in how knowledge is produced, how work gets done, and how value is created.

What had proved elusive, however, was a way to measure AI’s real-world economic footprint. To fill that gap, Anthropic—the creator of Claude—built the Anthropic Economic Index, a resource that offers policymakers, the public and businesses a data-driven look at where and how AI is being used in the real world, and what that means for society and the global economy. The Index tracks patterns of AI adoption—whether people are using AI to collaborate on tasks or delegate them—and usage patterns across industries and geographies, offering a near real-time picture of AI’s economic impact. Launched in February 2025, primarily in the US, it scaled globally later in the year. 

The Anthropic Economic Index covers more than 150 countries, revealing how AI adoption is unfolding across geographies. It also breaks down how AI is used in each US state, thereby identifying at state level the industries in which it is most widely adopted. The Index maps Claude usage to the US Department of Labor’s O*NET-SOC taxonomy, a classification system that organises job roles in the American workforce, with analysis that enables it to track structured patterns of task automation as well as the more creative, unconventional ways in which people are using AI. This dual approach enables the Index to surface quantifiable usage patterns as well as activities that would elude conventional labour taxonomies. The Index data are publicly available and open source, so academics, policymakers and the public can use the information to conduct their own research.

Mapping AI’s geography, and its gaps

The Index uncovers where early adopters are, what they use AI for and how those patterns develop over time, offering a window into the technology’s economic diffusion. The Anthropic AI Usage Index (AUI) measures whether Claude usage is over- or under-represented in an economy relative to its working-age population. Currently, results show that higher-income countries display far greater levels of AI activity. Advanced economies such as Singapore and Canada rank among the world’s most intensive users per capita, while emerging markets such as Indonesia, India and Nigeria have lower rates of adoption. The creation of this research tool allows Anthropic to measure and track diffusion and examine how usage develops in different parts of the world.

In higher-adoption regions, we find that people use AI for more diverse applications than simply coding, such as education, arts, and physical and social sciences. Countries with lower AI adoption, in contrast, tend to use Claude to delegate work—such as coding—whereas countries with higher AI adoption use AI as more of a collaborative partner.

Linking AI to local economies

In the US, adoption closely mirrors the make-up of local economies, notes Ms Heck. Washington, DC leads the nation, with per-capita usage at 3.8 times its population share, followed closely by Utah at 3.7 times. The data suggest that regional patterns reflect distinctive economic specialisations. Utah’s strong showing for AI usage in computing and mathematical tasks reflects the state’s so-called Silicon Slopes ecosystem, where there is a concentrated tech workforce and a thriving culture of innovation. In Washington, DC, meanwhile, AI is most often used for career support and document editing.

“For the most part, you’re seeing a trend of professions and industries using AI,” says Ms Heck. “If you go to Utah, you’re seeing things that are the core to the economy of Utah using AI in outsized ways. Even though diffusion is slow, concentration is in industries based on geography.”

Such uneven diffusion, she adds, could soon spill into politics. “I do think it's going to be a campaign issue in 2028, for sure,” she says, alluding to the next US presidential election. “I think in certain places in this country, it will probably be a 2026 agenda item too, because diffusion is not equal across the US and topics related to AI are not equal.” The economic impacts of AI, Ms Heck points out, “are kitchen table issues for people”.

Widening or narrowing the gap?

The implications go beyond cartography. If AI adoption today mirrors wealth, tomorrow it could reinforce or reshape it, predicts Anthropic in its report. Its data show that for every 1% rise in GDP per capita, AI use climbs roughly 0.7%, suggesting that wealthier economies are pulling further ahead in digital capability. This uneven adoption is something to watch.

“This technology is very transformative and does have immense productivity benefits,” Mr McCrory notes. “To the extent that those productivity benefits concentrate in places where adoption is high, this could lead to divergence in outcomes.” Ensuring equitable diffusion, he adds, depends on investing in digital infrastructure and ensuring that the benefits are broadly available.

Yet the same data hint at opportunity as well as risk. “Access is available around the world,” Mr McCrory adds, “and perhaps this technology can be a great boon to lower-income economies that can maybe follow a different trajectory… and get to the frontier faster by adopting this technology more quickly.”

How work itself is changing

Anthropic’s data show how AI is changing not just where people work, but how. Coding dominates Claude usage (36%), followed by education (12.4%) and science (7.2%). More telling is the rise of what Anthropic describes as directive use, where users assign tasks to Claude and await completion, which has climbed from 27% to 39% in just eight months.

The Index, Ms Heck believes, gives policymakers the means to act. Rather than merely observing the spread of AI, they can shape it, by using reliable data that can inform policy decisions. By visualising usage across sectors and occupations, the Index offers a data-backed map of where intervention, investment or training might be most effective to ensure digital inclusion and workforce planning.

The road ahead

Anthropic’s ambitions for the Index extend well beyond dashboards and charts. The company plans to expand international visualisations and support external researchers studying AI’s socioeconomic impact.

The Anthropic Economic Index is part map, part mirror—a reflection of how humans are learning to work alongside intelligence of a different kind.