The only free, publicly ranked dataset of AI automation risk for every country's workforce. Built on ILO, Eurostat, BLS, ONS and OECD official employment data. Updated 2026.
Key findings
Luxembourg ranks first globally with a workforce AI exposure score of 5.87/10, followed by Macao (5.74/10) and the Netherlands (5.44/10). High-income countries with large professional, financial and service sectors dominate the top 25 -- their workforces are concentrated in cognitive, administrative and analytical roles that current AI systems can most readily augment or replace.
Tanzania ranks last at 2.85/10, ahead of Rwanda (2.91/10) and Madagascar (2.91/10). Countries where agriculture employs the majority of the workforce score lowest on AI exposure because physical, outdoor labour in unstructured environments remains difficult for current AI systems. These same countries tend to score higher on robotics risk.
The global average AI exposure is 4.11/10 across 206 countries. 25 countries score 5.0 or above -- the threshold where a majority of typical workforce tasks are within reach of AI tools available today. 99 countries score below 4.0, reflecting workforces still dominated by manual, physical and agricultural labour. Germany ranks 9th globally at 5.3/10 with an unusually fast disruption timeline (9.6/10 velocity) given its industrial base.
5.87
Highest (Luxembourg)
2.85
Lowest (Tanzania)
4.11
Global average
206
Countries ranked
2.9B
Workers covered
25
High AI risk
Countries scoring 5.0+/10
82
Medium AI risk
Countries scoring 4.0-5.0/10
99
Lower AI risk
Countries scoring below 4.0/10
4.39
Avg robotics risk
Global average across 206 countries
2.97
Avg WFH potential
Global average, varies widely by income
2.22
Avg offshoring risk
Global average across all countries
Full global rankings
All 206 countries sorted by AI exposure score. Click any country name for the full occupation breakdown, wages and disruption timeline. Click any column header to re-sort.
206 countries
#▼
Country↕
Data
AI Exposure▼
Robotics↕
WFH Potential↕
Offshoring↕
Workforce↕
No countries match your search.
Frequently asked questions
Which country has the highest AI job risk in the world?
Luxembourg has the highest AI job exposure score globally at 5.87/10, followed by Macao (5.74/10) and the Netherlands (5.44/10). Luxembourg's workforce is heavily concentrated in financial services, professional services and administration -- all occupation categories with high AI substitution potential. Its small size means a relatively small shift in employment structure has outsized effects on the national average.
Which country has the lowest AI job risk?
Tanzania has the lowest AI job exposure score at 2.85/10, followed by Rwanda and Madagascar at 2.91/10. These countries have workforces dominated by smallholder farming, subsistence agriculture and informal manual labour -- activities that require physical presence, contextual judgment and adaptability in outdoor environments that current AI systems cannot automate at scale. The primary automation threat in these countries is robotics in agriculture, not AI in offices.
Is AI risk the same as robotics risk?
No, they are separate dimensions measured independently. AI exposure measures how much of an occupation's cognitive, administrative and communicative tasks can be performed or augmented by AI software (language models, image recognition, decision systems). Robotics risk measures how much physical task displacement is likely from industrial robots, agricultural machines and service robots. A country like Luxembourg scores high on AI (5.87) but moderate on robotics (2.86) because its workforce is office-based. A country like Burundi scores low on AI (3.05) but high on robotics (6.16) because its workforce is in agriculture where mechanisation is the primary automation threat.
What does the score actually measure?
Each score is a workforce-weighted average of occupation-level risk scores. For example, if 30% of a country's workforce is in clerical roles (AI score 8.5) and 70% is in farming (AI score 1.5), the country's weighted average AI score is (0.3 x 8.5) + (0.7 x 1.5) = 3.6. The occupation-level scores (0-10) reflect how much of that occupation's core tasks current AI or robotics systems can perform or significantly augment, based on research from Frey-Osborne (Oxford), OECD, and IMF. A score of 8 means roughly 80% of the occupation's core tasks are within reach of today's AI tools -- not that jobs disappear overnight, but that the pressure for automation investment is high.
What is Tier A vs Tier B vs Tier C?
Data quality tiers reflect how much occupation-level detail is available. Tier A (21 countries) means full employment data by detailed occupation group plus wage data by occupation -- the richest dataset. Tier B (19 countries) means employment by occupation group but limited or suppressed wage data. Tier C (166 countries) means employment by major ISCO-08 group only (9 broad categories) sourced from ILO ILOSTAT. The AI exposure and other scores are valid for all tiers -- the tier affects how granular the occupation breakdown is on the individual country page, not whether the country-level scores are reliable.
Where does this data come from?
Employment data by occupation comes from: ILO ILOSTAT (most countries, ISCO-08 major groups), Eurostat Labour Force Survey (EU countries, detailed groups), US Bureau of Labor Statistics OES (US, 341 detailed occupations), UK Office for National Statistics ASHE (UK, occupation-level wage data), and national statistical offices. Wage data comes from OECD Average Annual Wages, Eurostat Structure of Earnings Survey, and BLS. AI exposure and other risk scores are research-based estimates per ISCO-08 occupation group, informed by Frey and Osborne (Oxford 2013), OECD (2019, 2023), and IMF World Economic Outlook (2024). All underlying data is from official government or intergovernmental sources.
How often is this data updated?
The employment and wage data reflects the most recent available from each source -- primarily 2024-2025 releases covering 2022-2024 reference years. ILO ILOSTAT publishes annual updates typically with a 1-2 year lag. Eurostat and BLS publish annually. The AI exposure scores are updated when significant new research or capability shifts warrant revision. The last full dataset update was June 2025.
What each country record contains
Every country JSON at /data/countries/{code}.json includes the following field groups. A machine-readable OpenAPI 3.1 spec is also available.
9
ISCO-08 groups
AI, robotics, WFH, offshoring scores per group
8
WB economy fields
GDP, unemployment, Gini, poverty, literacy - 198 countries
6
HDR fields
HDI, rank, GNI PPP, schooling, life expectancy - 177 countries
127
Countries with wages
Median annual wage USD per ISCO group
4
Risk scores
Velocity, resilience, demographic alignment, sovereign buffer
Free
API access
No key, no auth - fetch directly from browser or server
What World Bank fields are included?
Each country JSON includes a wb_economy object with: GDP per capita (USD), total population, labour force participation rate (total and female), unemployment rate, poverty headcount at $3/day, Gini inequality coefficient, adult literacy rate, and life expectancy. All figures are the most recent available year per indicator from World Bank Open Data (CC BY 4.0). 198 of 206 countries have at least one World Bank field. The 8 missing are small territories not covered by the World Bank dataset.
What UNDP HDR fields are included?
Each country JSON includes an hdr_data object with: Human Development Index (HDI), global HDI rank, GNI per capita in PPP dollars, life expectancy at birth, expected years of schooling, and mean years of schooling. Source: UNDP Human Development Report 2023/24 (CC BY 3.0 IGO). 177 of 206 countries have HDR data. Small territories and disputed areas not included in the UNDP report will have a null hdr_data object.
How do I access the data programmatically?
All data is served as static JSON files with CORS open and no authentication. Fetch https://www.worldjobsdata.com/data/countries/us.json for the US, replacing us with any lowercase ISO 3166-1 alpha-2 code. A country list with summary scores is at /data/meta.json. ISCO-08 scores are at /data/isco_scores.json. Full endpoint reference and an OpenAPI 3.1 spec are at worldjobsdata.com/api.
Data sources: ILO ILOSTAT (CC BY 4.0), Eurostat Labour Force Survey (Eurostat open dissemination policy), US Bureau of Labor Statistics OES, UK ONS ASHE, OECD Average Annual Wages (2024, USD PPP), World Bank Open Data (CC BY 4.0), UNDP Human Development Report 2023/24 (CC BY 3.0 IGO). AI exposure scores are research-based estimates informed by Frey-Osborne (2013), OECD (2019, 2023), IMF (2024). Last updated: August 2026.
Explore individual countries: worldjobsdata.com/countries | Interactive tool: worldjobsdata.com/explore | API docs: worldjobsdata.com/api