Multi-Source Scoring Framework
How we score 186 economies
The AI Trajectory Index draws from 10 primary data sources to score every country across five pillars (0–20 each, total 100). Live World Bank data is our quantitative backbone — but the methodology is enriched and validated against the IMF AI Preparedness Index, Stanford HAI, Anthropic Economic Index, Oxford Insights, OECD.AI, Tortoise, WIPO GII, ITU IDI, and WEF.
World BankIMF AIPIOxford InsightsStanford HAIOECD.AIAnthropicTortoiseWIPO GIIITU IDIWEF
Data Connection Status
| SOURCE | STATUS | UPDATE FREQ. | COUNTRIES | INDICATORS |
|---|
| World Bank Open Data | LIVE API | 24h cache | 186 | 17 indicators |
| IMF DataMapper | LIVE API | 24h cache | 190+ | Govt exp, Debt % GDP |
| OECD MSTI | LIVE API | 24h cache | 38 | R&D % GDP, Researchers/1000 |
| Anthropic Economic Index | AUTO-FETCH | Weekly CSV | 150+ | AI usage intensity (0–100) |
| Oxford Insights GARI | ANNUAL | Yearly (2024) | 195 | Govt AI Readiness (0–100) |
| Stanford HAI AI Index | ANNUAL | Yearly (2025) | 75+ | Research output (0–100) |
| Tortoise Global AI | ANNUAL | Yearly (2024) | 83 | Composite AI score (0–100) |
| WIPO Global Innovation Index | ANNUAL | Yearly (2024) | 133 | Innovation score (0–100) |
| WEF Global Competitiveness | ANNUAL | Yearly (2019) | 141 | Competitiveness (0–100) |
| ITU ICT Dev. Index | ANNUAL | Yearly (2023) | 190+ | ICT Development (0–100) |
LIVE APIFetched every request, cached 24h
AUTO-FETCHPulled from public dataset automatically
ANNUALUpdated manually when report publishes
Composite Score Formula
Total Score = Infrastructure + Talent + Governance + Investment + Economic Readiness
Each pillar: 0–20 pts → Total: 0–100 · Classification: Leading (80+), Advanced (60–79), Developing (40–59), Nascent (<40)
Pillar Definitions & Sources
🔌Infrastructure
0–20 points
The physical and digital foundation for AI deployment.
Internet users as % of population — World Bank IT.NET.USER.ZS (8 pts)
Mobile cellular subscriptions per 100 people — World Bank IT.CEL.SETS.P2 (5 pts)
Electricity access as % of population — World Bank EG.ELC.ACCS.ZS (5 pts)
Infrastructure quality proxy (2 pts) — derived from ITU IDI scores and qualitative data centre assessments
SOURCES:World BankITU IDI
The human capital pipeline for AI development and deployment.
Gross tertiary enrolment ratio — World Bank SE.TER.ENRR (8 pts)
R&D expenditure as % of GDP — World Bank GB.XPD.RSDV.GD.ZS (8 pts)
Talent quality proxy (4 pts) — GII university ranking presence, Stanford HAI research output, WEF AI upskilling commitment
SOURCES:World BankStanford HAIGlobal Innovation IndexWEF Future of Jobs
Policy maturity, regulatory frameworks, and government AI strategy.
National AI strategy adoption — OECD.AI Observatory (8 pts base; +2 bonus if post-2022)
AI-specific regulation or data protection law — OECD.AI / Oxford Insights (5 pts)
OECD membership — proxy for institutional regulatory capacity (3 pts)
Digital governance proxy (2 pts) — Oxford Insights Government AI Readiness, Stanford HAI legislative activity
SOURCES:OECD.AI ObservatoryOxford InsightsStanford HAI
Capital flowing into AI — public R&D, private venture capital, and FDI.
R&D expenditure as % of GDP — World Bank GB.XPD.RSDV.GD.ZS (6 pts)
GDP per capita USD — World Bank NY.GDP.PCAP.CD (6 pts, proxy for private capital availability)
VC and ecosystem proxy (8 pts) — OECD.AI VC data, Tortoise commercial activity, GII innovation output
SOURCES:World BankOECD.AI ObservatoryTortoise Global AI IndexGlobal Innovation Index
📊Economic Readiness
0–20 points
The economy's structural capacity to adopt and commercialise AI.
GDP per capita USD — World Bank NY.GDP.PCAP.CD (8 pts)
Electricity access — World Bank EG.ELC.ACCS.ZS (4 pts)
Internet and mobile penetration — World Bank composite (4 pts)
AI adoption readiness proxy (4 pts) — Anthropic Economic Index augmentation ratio, WEF employer AI adoption rates, IMF AIPI economic integration dimension
SOURCES:World BankAnthropic Economic IndexIMF AIPIWEF Future of Jobs
Trajectory Score & 2028 Projection
The trajectory score (–10 to +10) reflects momentum. Four forward-looking components:
25%GDP growth rate (3-year avg)
World Bank NY.GDP.MKTP.KD.ZG
Structural proxy for economic capacity to fund AI investment. Compounding growth signals a broadening tax base and increasing government AI budget.
20%Internet penetration growth
World Bank IT.NET.USER.ZS (YoY delta)
Captures digital adoption momentum. Accelerating internet growth predicts faster AI tool adoption cycles within 2–3 years.
25%AI strategy recency
OECD.AI Observatory
Countries with post-2020 strategies receive a strong positive signal (reflecting active, not archival, government engagement). No strategy = negative weighting.
15%R&D spending trend
World Bank GB.XPD.RSDV.GD.ZS (YoY delta)
Year-over-year change in R&D as % of GDP. Increasing R&D signals a structural commitment to innovation that produces compounding AI capability returns.
15%AI adoption momentum proxy
Anthropic Economic Index · Stanford HAI · WEF
Incorporates AI augmentation adoption rates (Anthropic), employer upskilling commitments (WEF), and research output growth (Stanford HAI).
Projected 2028 = clamp(Total + Trajectory × 1.5, 0, 100)
Labels: Strong Positive (+6→+10) · Positive (+2→+5) · Neutral (-1→+1) · Negative (-5→-2) · Strong Negative (-10→-6)
Primary Data Sources
The index is deliberately multi-source. Each source has different country coverage, update frequency, and methodological strengths. Using them together reduces single-source bias and improves cross-country comparability.
🤖Anthropic Economic Index
Global · Country-level AI adoptionAnthropic · Updated 2025–ongoing
Visit ↗What it measures
Real-world AI usage patterns from Claude interactions — separating augmentation (humans collaborating with AI) from automation (AI replacing human tasks). Provides country-level and industry-level adoption signals, updated continuously.
How we use it
Informs the Economic Readiness pillar. Countries with high augmentation ratios signal a workforce actively integrating AI. Industry exposure patterns help weight AI adoption signals in economic readiness scoring.
Key insight
The index reveals that AI augmentation significantly outpaces automation globally — the majority of AI interactions involve humans and AI working together, not AI replacing work entirely.
INFORMS:Economic ReadinessTalent
📊IMF AI Preparedness Index (AIPI)
174 countriesInternational Monetary Fund · Updated 2023–2024
Visit ↗What it measures
Four-dimension composite: Digital Infrastructure, Human Capital & Labour Market Policies, Innovation & Economic Integration, Regulation & Ethics. Aggregates data from ILO, World Bank, WEF, and 5 other institutions.
How we use it
Primary cross-validation source. We compare our composite scores against IMF AIPI scores for countries in both datasets. Where material divergence exists (>10 pts), we review our underlying indicator weights. Freely downloadable Excel data used for static baseline calibration.
Key insight
Advanced economies score 0.6–0.7 on a 0–1 scale; emerging markets 0.3–0.5. The IMF finds governance and regulatory capacity as the greatest differentiator between similarly-wired economies.
INFORMS:InfrastructureTalentGovernanceInvestment
🏛️Oxford Insights Government AI Readiness Index
195 countriesOxford Insights · Updated 2025
Visit ↗What it measures
Public sector capacity to harness AI for public benefit. 69 indicators across 6 pillars: Policy Capacity, Governance, AI Infrastructure, Public Sector Adoption, Development & Diffusion, Resilience.
How we use it
Governance pillar calibration. Oxford Insights' detailed policy capacity and governance scores provide granular country-level data that supplements OECD policy flags. Particularly valuable for countries outside the OECD where national AI strategy quality varies.
Key insight
North America averages 81.5/100 vs Sub-Saharan Africa at 29.1 — the largest regional divide in any AI readiness index, signalling governance infrastructure as the primary bottleneck for developing economies.
INFORMS:GovernanceInfrastructure
🎓Stanford HAI AI Index Report
75+ countriesStanford Human-Centered AI Institute · Updated Annual (2025 edition)
Visit ↗What it measures
Research and development output (papers, models, citations), AI education and workforce trends, legislative and policy activity (AI mentions in 75 country parliaments), public opinion and adoption, AI model production by country.
How we use it
Talent pillar enrichment. Country-level AI research publication counts and AI model production figures supplement World Bank tertiary enrolment data. Legislative AI activity counts inform the Governance pillar recency score.
Key insight
The US produced 40 notable AI models in 2024 vs China's 15 and Europe's 3 combined — a research production gap that will compound across the five-year outlook period.
INFORMS:TalentGovernance
🌐OECD.AI Policy Observatory
70+ countries and territoriesOrganisation for Economic Co-operation and Development · Updated Live
Visit ↗What it measures
900+ national AI policies and initiatives tracked live. VC investment in AI by country, AI job postings and skills demand, software development contributions, compute capacity, national AI strategy status and quality.
How we use it
Primary source for Governance pillar: national AI strategy adoption dates, AI regulation flags, OECD membership. Also provides VC investment data that directly informs the Investment pillar for OECD members and partners. Live OECD.AI Index cross-validates Investment scores.
Key insight
Countries with comprehensive AI strategies (vs no strategy) score on average 18 points higher on total AI readiness — the single strongest binary predictor in our model.
INFORMS:GovernanceInvestment
🏦World Bank Open Data API
186 countriesWorld Bank · Updated Live (24h cache)
Visit ↗What it measures
Six live indicators: internet users (% population), mobile subscriptions per 100, tertiary enrolment ratio, R&D expenditure (% GDP), GDP per capita (USD), electricity access (% population). Two most-recent data points per indicator for trend calculation.
How we use it
Primary quantitative backbone. The most recent available indicator values drive Infrastructure, Talent, Investment, and Economic Readiness pillar scores. Year-over-year changes feed the trajectory calculation. Live data is fetched daily with a 24-hour server cache; static baseline serves as fallback.
Key insight
R&D expenditure (% GDP) has the strongest correlation (r=0.74) with total AI readiness score of all six indicators — stronger than internet penetration or GDP per capita.
INFORMS:InfrastructureTalentInvestmentEconomic Readiness
🐢Tortoise Global AI Index
83 countriesTortoise Media · Updated 2024
Visit ↗What it measures
122 indicators across Implementation (Talent, Infrastructure, Operating Environment), Innovation (Research, Development), and Investment (Government Strategy, Commercial Activity). Considered one of the most comprehensive AI-specific indices.
How we use it
Spot-check and calibration for major economies. For the 83 countries covered, we compare Tortoise rankings against our own scoring. Notable divergences flag potential data gaps or weighting differences. Saudi Arabia's government strategy ranking prompted our addition of AI-strategy recency bonus.
Key insight
Saudi Arabia tops government AI strategy rankings globally despite being outside the traditional Western AI ecosystem — a signal that sovereign AI investment can rapidly move governance scores.
INFORMS:GovernanceInvestmentTalent
💡Global Innovation Index (GII)
133 economiesWorld Intellectual Property Organization (WIPO) · Updated Annual (2024 edition)
Visit ↗What it measures
~80 indicators across innovation inputs (policy environment, education, infrastructure, market sophistication, business sophistication) and outputs (knowledge/technology outputs, creative outputs). Includes 100 top global science and technology clusters.
How we use it
Investment and Talent pillar enrichment for non-OECD countries where VC data is sparse. GII innovation input scores provide a validated proxy for the broader innovation ecosystem that supports AI investment. University ranking presence from GII supplements our static talent quality proxy.
Key insight
Switzerland, Sweden, and the US top the GII consistently — but China has been in the top 12 since 2018, confirming that sustained R&D investment can close the innovation gap within a decade.
INFORMS:TalentInvestment
📡ITU ICT Development Index (IDI)
190+ countriesInternational Telecommunication Union · Updated 2024
Visit ↗What it measures
Universal and meaningful connectivity — ability for everyone to access broadband internet at affordable cost, anywhere, anytime. Measures infrastructure quality, accessibility, affordability, and digital skills. Identifies urban-rural connectivity gaps.
How we use it
Secondary cross-validation for the Infrastructure pillar. ITU's IDI provides a connectivity-specific lens that complements World Bank internet penetration data, particularly for distinguishing between countries with similar penetration rates but very different infrastructure quality.
Key insight
Broadband quality — not just access — emerges as the critical differentiator at the 60–80% internet penetration threshold. Many countries plateau at basic connectivity but lag on the reliable, high-speed infrastructure AI applications require.
INFORMS:Infrastructure
🔮WEF Future of Jobs Report
55 countries · 1,000+ employersWorld Economic Forum · Updated 2025
Visit ↗What it measures
AI skills adoption rates by businesses, job creation and displacement projections (11M created, 9M displaced by 2030), workforce AI strategy trends, upskilling commitments (85% of employers plan AI upskilling). Sectoral breakdown across 22 industry clusters.
How we use it
Talent pillar future-weighting. Countries where employer surveys show high planned AI upskilling investment receive a forward-looking boost to their talent trajectory score. IT sector AI adoption rates by country calibrate industry-level AI maturity signals.
Key insight
85% of employers globally plan to upskill workers for AI collaboration — but only 50% plan to retrain displaced workers. The talent gap is growing faster than the training pipeline can fill it.
INFORMS:TalentEconomic Readiness
Source Coverage by Pillar
| SOURCE | INFRA | TALENT | GOV | INVEST | ECON READY | TRAJ. |
|---|
| World Bank API | ● | ● | ○ | ● | ● | ● |
| OECD.AI Observatory | ○ | ● | ● | ● | ○ | ● |
| IMF AIPI | ● | ● | ● | ● | ○ | ○ |
| Oxford Insights | ● | ○ | ● | ○ | ○ | ○ |
| Stanford HAI | ○ | ● | ● | ○ | ○ | ● |
| Anthropic Economic Index | ○ | ● | ○ | ○ | ● | ● |
| Tortoise Global AI Index | ● | ● | ● | ● | ○ | ○ |
| WIPO GII | ○ | ● | ○ | ● | ○ | ○ |
| ITU IDI | ● | ○ | ○ | ○ | ○ | ○ |
| WEF Future of Jobs | ○ | ● | ○ | ○ | ● | ● |
Normalisation & Cross-Validation
Each live World Bank indicator is normalised against a global benchmarking range using min-max scaling. For example, internet penetration uses a 0–95% practical ceiling. GDP per capita uses a logarithmic scale.
The composite is cross-validated against IMF AIPI scores (174 countries) and Oxford Insights rankings (195 countries). Where our score diverges by more than 10 points from IMF AIPI on the same country, we flag for manual review and adjust static proxy weights accordingly.
Countries missing World Bank data fall back to the static 2024 baseline. The data_source flag on each country response indicates whether scores are live-calculated or from the static baseline.
Limitations & Caveats
⚠World Bank data lags by 1–3 years for many countries. The most recent available point is used, which may not reflect the current situation.
⚠The Anthropic Economic Index reflects Claude usage patterns — not all AI adoption globally. It should be read as a directional signal, not a census of AI use.
⚠Stanford HAI AI Index country coverage varies by metric: research output covers 75+ countries but some public opinion data covers only 17 countries.
⚠Any composite index involves trade-offs in weighting. Our weights reflect the authors' best judgement and cross-validation against IMF AIPI — not a peer-reviewed consensus.
⚠The governance pillar heavily rewards formal AI strategies. Countries with strong strategies but weak implementation may be overstated.
⚠Countries with populations under 1M (small island states, city-states) may have extreme indicator values. Singapore and Luxembourg are legitimate outliers; Maldives and Seychelles should be interpreted with caution.
⚠AI safety, ethics oversight, and bias mitigation are not scored — critical dimensions omitted due to data availability constraints.
Methodology Changelog
March 2026CURRENT
- Removed R&D expenditure from Talent pillar — was double-counted with Investment.
- Removed GDP per capita from Investment pillar — was double-counted with Economic Readiness.
- Removed electricity/internet/mobile from Economic Readiness — was triple-counted with Infrastructure.
- Switched GDP per capita to PPP-adjusted values (NY.GDP.PCAP.PP.KD) for fairer cross-country comparisons.
- Added WGI estimates (Rule of Law, Govt Effectiveness, Regulatory Quality) to Governance pillar.
- Added high-tech exports trend and labor productivity trend to Trajectory calculation.
- Total live World Bank indicators expanded from 6 to 17.
February 2026PREVIOUS
- Replaced fully static scoring with live World Bank API integration (6 indicators, 24h cache).
- Added trajectory score (–10 to +10) based on year-over-year WB indicator deltas.
- Introduced projected_score_2028 = clamp(total + trajectory × 1.5, 0, 100).
- Added static fallback for countries with missing World Bank data.
- Launched /map page with Readiness and Adoption heat-map lenses.
January 2026INITIAL
- Static baseline scores for 186 countries across 5 pillars (0–20 each).
- Scores cross-validated against IMF AIPI (174 countries) and Oxford Insights (195 countries).
- Policy flags (AI strategy, regulation, OECD membership) from OECD.AI Observatory.
- Qualitative evidence strings per pillar sourced from Stanford HAI, Tortoise, WIPO GII.
About the Builder
This index was built by Ankit Mishra as an independent research tool to support work at the intersection of AI governance, emerging markets, and technology policy. Ankit is Commercial Portfolio Director at a leading African climatetech venture fund, a member of the Schwartz Reisman Institute AI & Trust Working Group at the University of Toronto, and a Forbes contributor with 50+ articles reaching 200,000+ readers.
🚀AI Adoption Scorecard
METHODOLOGY · 2026
The Adoption Scorecard measures whether countries are actively deploying and using AI — distinct from the Readiness Index which measures capacity. A country can score highly on readiness but deploy AI slowly due to cultural, regulatory, or economic friction. Conversely, some countries deploy AI rapidly through mobile-first channels despite lower readiness scores. This gap between readiness and adoption is one of the most important insights the combined platform reveals.
The gap is calculated as: adoption_score − readiness_score.
A positive gap means the country is adopting faster than its capacity predicts. A negative gap means it has untapped AI capacity not yet being utilised.
🏛️ Government Deployment (0–20)
Active AI in public services: e-government AI integration, AI in healthcare delivery, algorithmic public administration, smart city deployments, government AI procurement.
Sources: Oxford Insights Government AI Readiness Index, UNDP e-government surveys, national AI strategy implementation reports
🏢 Enterprise Adoption (0–20)
Business AI usage: World Bank Enterprise Survey digital adoption rates, OECD business AI data, fintech AI penetration, manufacturing and retail AI adoption.
Sources: OECD Business AI Adoption Survey, World Bank Enterprise Surveys, McKinsey Global AI Survey
💼 Talent Demand (0–20)
Labour market AI demand: AI/ML job postings as % of total, YoY growth rate, AI skills salary premium, AI startups per million population.
Sources: OECD.AI job postings data, LinkedIn Economic Graph, Indeed AI jobs tracker
📱 Consumer Usage (0–20)
Everyday AI adoption: smartphone AI assistant penetration, AI-powered mobile payments, voice assistant adoption, AI in e-commerce and healthcare apps. Mobile-first fintech (M-Pesa, GCash, bKash) is a key signal for emerging markets.
Sources: GSMA Mobile Economy, Statista AI consumer surveys, fintech adoption data
🔬 R&D Pipeline (0–20)
Research-to-deployment velocity: AI patent filings per million (WIPO), university-to-industry AI transfer rate, AI unicorn and soonicorn count per capita, AI accelerator density.
Sources: WIPO Patent Database, Crunchbase AI unicorn tracking, Stanford HAI AI Index
THE LEAPFROGGING EFFECT
Several Sub-Saharan African and Southeast Asian economies score higher on adoption than their readiness would predict. This is the mobile-first leapfrogging effect: countries without legacy banking or desktop internet infrastructure have adopted mobile-first AI tools (M-Pesa in Kenya, GCash in the Philippines, bKash in Bangladesh) at scale, embedding AI into daily financial life ahead of their overall digital infrastructure development.