Career Risk · 6 min read
What this page shows
Which countries actually use AI the most, per person
This page ranks every country and US state by per-capita usage of Claude, the AI assistant built by Anthropic, using Anthropic's own Economic Index data. The reason to look at per-capita usage rather than total volume is simple: a country with a hundred million people will always show more total AI conversations than a country with five million, even if the smaller country is far more AI-forward. Per-capita usage strips out population size and shows you which economies are actually leaning into AI tools in everyday work.
The headline metric, the Anthropic AI Usage Index (AUI), is a ratio. A score of 1.0 means a country's share of Claude usage is proportional to its share of the global working-age population. A score of 4.0 means the country uses AI four times as much as population alone would predict. We stitch three Anthropic snapshots together (August 2025, November 2025, and February 2026) so each row also carries a trend arrow showing whether per-capita adoption is rising or falling in that country. Monaco currently leads at roughly ten times the baseline; the United States is at 3.8×; the global picture is diverging, with top-adopting countries pulling further ahead while many emerging markets remain near 1.0×.
Countries ranked
173
With ≥ 200 conversations in the reporting window
US states ranked
51
With ≥ 100 conversations in the reporting window
Top country AUI
9.58×
Monaco leads per-capita usage
Top US state AUI
4.31×
District of Columbia leads per-capita usage
All 173 countries by per-capita Claude usage
AUI = per-capita usage relative to global baseline. 1.0× is proportional to working-age population; higher values mean more usage per capita than population alone would predict.
Sort:
How people use AI
Automation vs augmentation by country
Anthropic classifies every conversation by collaboration pattern. Automation covers directive use and feedback loops, where Claude completes the task on the user's behalf. Augmentation covers validation, iteration, and learning, where Claude assists a person doing the work themselves. Most countries sit close to a 50/50 mix, but the lean varies, and the lean tracks the work being done. Coding-heavy economies skew slightly more toward automation. Writing-heavy and education-heavy economies skew slightly more toward augmentation.
Monaco
0.0% / 100.0%
Cayman Islands
56.9% / 43.1%
Andorra
25.9% / 74.1%
Singapore
39.9% / 60.1%
Israel
41.3% / 58.7%
Luxembourg
43.2% / 56.8%
Switzerland
40.0% / 60.0%
United States
41.4% / 58.6%
Canada
39.7% / 60.3%
Australia
40.1% / 59.9%
The Netherlands
38.6% / 61.4%
Malta
41.1% / 58.9%
Bar shows automation share of classified collaboration patterns. Right column reads automation / augmentation.
What people use AI for
Dominant occupational work mix in the top AI-adopting countries
For each country we surface the top three SOC occupational groups represented in Claude.ai conversations, mapped from O*NET task classifications. Computer and Mathematical work dominates almost everywhere; Educational Instruction and Office and Administrative Support typically round out the top three. This is where the country-level adoption signal connects to job-level AI exposure.
Monaco
+4.66
AUI 9.58×
Not enough classified observations.
Work 51.0%Personal 42.9%
Cayman Islands
+2.51
AUI 4.92×
Not enough classified observations.
Work 47.2%Personal 37.5%
Andorra
+2.17
AUI 4.76×
Not enough classified observations.
Work 56.8%Personal 41.9%
Singapore
+0.07
AUI 4.64×
Computer and Mathematical26.9%
Educational Instruction and Library8.3%
Office and Administrative Support4.9%
Work 40.3%Personal 40.0%Coursework 19.6%
Liechtenstein
+1.43
AUI 4.51×
Not enough classified observations.
Work 51.6%
Israel
-2.64
AUI 4.37×
Computer and Mathematical23.5%
Educational Instruction and Library11.6%
Office and Administrative Support6.1%
Work 46.3%Personal 42.9%Coursework 10.7%
Luxembourg
+1.51
AUI 4.25×
Not enough classified observations.
Work 56.0%Personal 39.8%Coursework 4.2%
Switzerland
+1.26
AUI 4.07×
Computer and Mathematical23.4%
Educational Instruction and Library6.4%
Office and Administrative Support5.5%
Work 45.9%Personal 47.5%Coursework 6.6%
United States
+0.22
AUI 3.84×
Computer and Mathematical26.1%
Educational Instruction and Library10.7%
Arts, Design, Entertainment, Sports, and Media9.1%
Work 40.4%Personal 47.0%Coursework 12.5%
Methodology
How the AUI is computed
The Anthropic AI Usage Index (AUI) is a per-capita concentration ratio. For each geography, Anthropic divides that geography's share of Claude.ai conversations by its share of global working-age population. A score of 1.0 means proportional usage. A score above 1.0 means usage is over-represented relative to population.
Working-age population is the 15-64 age band, sourced from the World Bank (plus the National Development Council for Taiwan). GDP per working-age capita is sourced from the IMF World Economic Outlook. US state populations come from the US Census Bureau.
Privacy and signal floors: countries with fewer than 200 conversations and US states with fewer than 100 conversations in the reporting window are excluded by Anthropic before the file is published.
Source dataset: Anthropic Economic Index, geographic insights release (September 2025). Reporting window: 2026-02-05 → 2026-02-12. Platform: Claude.ai (Free, Pro, and Max).
See also our role-level AI adoption data, which uses a different cut of the same Anthropic Economic Index: the SOC-level job exposure file powers the "AI Adoption" metric on every role page and the AI Career Risk Rankings.
The AUI is a geographic concentration ratio, not a role-risk score. It tells you where Claude is being used relative to population, not which jobs inside a country are most exposed. The AI Career Index applies a separate deterministic structural model to the role-level question, scoring an occupation across five weighted dimensions against O*NET task data and BLS wage data. For the formula, the dimensions, and how the projection math works, see the AI Career Index methodology.
FAQ
Common questions about country AI adoption
"Used this before a performance review and it completely changed my framing. Walked in with a structural argument for scope expansion instead of the usual list of accomplishments. The conversation shifted immediately. I got the authority expansion I had been trying to earn for eighteen months."
Jonathan Pierce
Director of Client Strategy · New York, USA
"My CEI was low which I thought meant I could relax. The Drift Path projection showed my position would still erode meaningfully over 24 months without deliberate action. Protection is its own strategy. I would not have seen that without the tool."
Saoirse O'Donnell
Principal Architect · Dublin, Ireland
"I am the kind of professional who usually dismisses career tools. This one is different because the outputs are reproducible and the formulas are disclosed. It passed the test I apply to everything: would I trust this enough to base a real decision on it? The answer was yes."
Rafael Costa
Head of Operations · Lisbon, Portugal