Measuring how the world adopts AI — monthly, across 129 economies
Last updated: Jul 21, 2026 · Pilot edition
27.5% of working-age people in the Global North used AI in Q1 2026 versus 15.4% in the Global South. The top 20 economies now account for 48% of per-capita frontier-assistant usage — up from 45% one wave earlier.
The US Census (all firms, any use) reads 19.8%. Ramp's spend panel (tech-forward firms, paid use) reads 54.2%. Both are right — which is exactly why this index publishes definitions next to every number.
Jordan, Vietnam and Lebanon score far above their income-group average; Russia, Kuwait and Saudi Arabia score far below. Structural readiness is not destiny — policy and diaspora networks move the needle.
Pilot composite = population AI usage (Microsoft diffusion telemetry, 40) + frontier-assistant intensity (Anthropic AUI, 25) + structural readiness (IMF AIPI, 35). Countries need ≥2 of 3 core indicators to be ranked. Click any row for its full scorecard.
Confidence grades: A = all three core indicators plus enterprise/labor module data · B = all three core indicators · C = two of three. 18 further economies are tracked on a single indicator and listed unranked in the data appendix.
Readiness indexes largely reproduce GDP per capita. The more interesting question is the residual: each economy's GAAI score minus its World Bank income-group average. Positive bars punch above their weight.
No harmonized global country-by-sector statistic exists — so this module shows the three best measured lenses side by side, mapped to common ISIC-style sections: the US Census BTOS survey (all US firms, any use, May 2026), Eurostat's enterprise survey (EU firms 10+ employees, 2025), and Ramp's spend panel (tech-forward US firms, paid AI, June 2026).
Between annual surveys, three high-frequency series carry the signal. GAAI's monthly cadence rests on these plus quarterly telemetry waves — each shown in its own frame, on its own scale.
The field splits four ways: slow-but-representative official surveys, fast-but-biased vendor telemetry, static income-correlated readiness composites, and self-selected executive polls. Nothing occupies the cell this index targets: monthly, multi-country, sector-aware, depth-graded, and honest about uncertainty.
✓ = published dimension · — = absent. “Actual use” means observed usage or representative adoption surveys, as opposed to readiness inputs or opt-in executive self-reports. Full critiques of each source are in the accompanying research paper.
The Skills Readiness Index (SRI) is GAAI's companion measure of the workforce's capacity to supply AI-era skills — the human-capital side of adoption. It blends three components: applied skill proficiency (technology & data-science competency, weight 45), foundational human capital (35), and tertiary attainment (20), each scaled 0–100. An economy is scored when at least two of the three are observed. 126 economies covered.
Click any column header to sort. Prof = applied skill proficiency (technology & data-science competency percentiles) · HumCap = foundational human capital (0–100) · Tert = tertiary attainment (gross enrolment, capped at 100). Bands: Very high ≥ 82 · High 68–82 · Moderate 52–68 · Developing 38–52 · Low < 38. SRI correlates 0.86 with the GAAI adoption composite — coherent, but a distinct construct: some economies read AI intensively without a deep skills base, and vice versa.
Click any column header to sort. MS = Microsoft diffusion (% working-age population, Q1 2026) · AUI = Anthropic AI Usage Index (Nov 2025) · AIPI = IMF AI Preparedness (2023) · Ent = % firms using AI (official surveys) · Lab = AI share of job postings (Lightcast 2025). Unranked economies carry a single core indicator.
The official GAAI monthly series since inception — March 2025, the first Microsoft diffusion wave — for all 111 ranked economies. Between published waves, sub-indicators are linearly interpolated; outside their observed anchors they are carried constant (never extrapolated), so every monthly value is bounded by real observations. July 2026 equals the published edition scores exactly. Economies with a single wave per source show a flat series until their history accumulates.
Wave timing: Microsoft H1 ’25 → Mar, H2 ’25 → Sep, Q1 ’26 → Feb · Anthropic AUI Aug ’25 / Nov ’25 / Feb ’26. Per-economy wave charts live in Country profiles.
Means over ranked economies only. Income groups: World Bank FY2026. Economy types: IMF WEO. GAAI scores 0–100.
Six machine-readable datasets, generated from this edition's frozen vintage. Free to reuse with attribution; third-party source data remain © their publishers.