Integrated AI evidence

Seven source groups, kept distinct by population, time and method. These static pages work without scripts.

Sources and coverage

SourcePublishedObservedCoverageLimit
Microsoft AI Economy Institute2025–2026 report seriesH1 2025–Q2 2026; US Q1 2026147 economies; 50 states + DC, 35 metros, 3,143 countiesAdjusted population telemetry and modeled local estimates. Usage is not proficiency or productivity.
Anthropic26 June 2026Geography: May 2026; output analysis: 10 Apr–10 Jun 2026Claude chat and Cowork; country publication thresholds applyProvider-specific conversations; classifier-inferred purpose. The per-capita index is not a user percentage.
European Commission JRC / EurostatJRC: 14 October 20262021–2025; Eurostat snapshot updated 15 June 2026EU27; eligible enterprises with 10+ persons employedBroad AI, not only GenAI. Strategic-sector coverage varies; the country map uses the broader enterprise series.
McKinsey25 August 20264 May–8 June 20261,719 respondents across 97 countriesSelf-reports, GDP weighting, non-census sample. Reported financial effects are not causal estimates.
Gallup / JFF / Families & Workers Fund6 October 202626 January–24 March 202615,482 US employees; AI-effect questions: 7,845 AI usersEmployee assessments; frequency and benefit questions have different bases. Job quality association is not causation.
HEPI / Kortext12 March 2026December 2025 for the 2026 edition1,054 full-time UK undergraduatesWeighted self-report survey. Earlier waves differ; 2026 adds wellbeing uses. Tool provision is student-reported.
National Literacy Trust18 August 2026Early 2026; pupils and teachers January–March40,543 pupils aged 13–18; 2,567 teachers; 3,073 parentsSchool recruitment; pupil sample largely England. Questions use different routed bases; findings are not national prevalence estimates.

Reuse and attribution