Methodology
What each pillar means
- Government
- Whether the public sector has the vision, governance, regulation and internal capacity to adopt AI responsibly — strategies, ethics frameworks, procurement readiness and digital capacity.
- Technology
- The maturity and innovation capacity of the country's technology sector — AI firms, research output, venture funding and the wider business environment for building AI products.
- Data & Infrastructure
- The availability, representativeness and openness of data, plus the physical and digital infrastructure (connectivity, compute, cloud) that AI systems depend on.
- Human Capital
- The supply of AI-relevant skills — STEM graduates, digital skills in the workforce, and the education and labour-market policies that develop them.
Where the numbers come from
Every figure rendered on this site carries a sourceId that resolves to one of the providers below, and its citation is shown next to the data. Providers are never blended: switching source switches the entire dataset, and each provider's values stay on its native scale in the data files. For display we normalise to 0–100 using the source's declared scale; the native value is always available in tooltips and score cards.
AI Readiness Index — Yaqtin
Our own index, not a third party's, published as an annual series from 2010 to 2023. Thirteen indicators across the four pillars, all rates, ratios or bounded indices — no raw counts, which would make the index a proxy for country size. COMPARABILITY ACROSS YEARS: each indicator's scaling bounds (the 2.5th and 97.5th percentiles) are computed ONCE across every country-year and reused for all years, so a 2010 score and a 2023 score are measured with the same ruler. Scaling each year against its own distribution would have measured relative position only — if every country improves, every score stays flat and real global progress disappears. On this shared ruler the world mean rises from 0.393 in 2010 to 0.494 in 2023. A pillar is the unweighted mean of its available indicators (at least 2 required), and the overall score is the unweighted mean of ALL FOUR pillars — a country missing any pillar is left unscored rather than averaged over the rest, because averaging over whichever pillars a country happens to report flatters those with sparse statistics. Equal weighting is a deliberate choice, mirroring the IMF's stated approach, not a finding. EACH YEAR USES ONLY FIGURES REPORTED FOR THAT YEAR: no carry-forward and no interpolation, so a country that skipped a reporting year shows a gap. Between 109 and 137 countries are scored per year. The series ends at 2023 because national statistics arrive with a lag — 2024 currently has about 58 reporting countries and 2025 almost none, and publishing those would show a cliff that is a reporting artefact rather than a fall in readiness.
Native scale: 0–1 (higher is better) · https://yaqtin.net/ai-readiness/methodology
| Government | Regulatory Quality, Rule of Law and Government Effectiveness (Worldwide Governance Indicators) |
|---|---|
| Technology | R&D expenditure (% of GDP), researchers per million people, ICT service exports (% of service exports), high-technology exports (% of manufactured exports) |
| Data & Infrastructure | Internet users (% of population), fixed broadband and mobile subscriptions per 100 people |
| Human Capital | Tertiary enrolment (% gross), government education spending (% of GDP), labour force with advanced education (%) |
Required citation: Yaqtin AI Readiness Index (2010-2023 series), built from World Bank Open Data. https://yaqtin.net/ai-readiness
Licence & attribution: Our own composite, published under CC BY 4.0 — reuse it with attribution. Built entirely from World Bank Open Data (CC BY 4.0 for World Bank-produced datasets). The underlying indicators are compiled by the World Bank from national statistical offices and agencies including the ITU, UNESCO Institute for Statistics and the Worldwide Governance Indicators programme, which carry their own terms; the indicator selection, scaling and weighting are Yaqtin's work and Yaqtin's responsibility.
AI Preparedness Index — International Monetary Fund
Covers 174 economies for 2023 on a 0-1 scale. The overall index is the sum of four dimensions, each published as a weighted contribution on a 0-0.25 scale rather than an independent 0-1 sub-score - so pillar values here are read against their own scale (verified: the four dimensions sum to the overall index across all 174 economies). Values were retrieved from the IMF DataMapper API and match the IMF's published aipidata.xlsx exactly. Note the IMF's own indicator metadata mislabels the "RE" series with the Human Capital label; the published spreadsheet confirms it is Regulation and Ethics, which is how it is mapped here. The AIPI itself aggregates indicators from eight third-party institutions, which carry their own terms.
Native scale: 0–1 (higher is better) · pillars published on their own 0–0.25 scale · https://www.imf.org/external/datamapper/datasets/AIPI
| Government | Regulation and Ethics dimension |
|---|---|
| Technology | Innovation and Economic Integration dimension |
| Data & Infrastructure | Digital Infrastructure dimension |
| Human Capital | Human Capital and Labor Market Policies dimension |
Required citation: Cazzaniga and others. 2024. "Gen-AI: Artificial Intelligence and the Future of Work." IMF Staff Discussion Note SDN2024/001, International Monetary Fund, Washington, DC.
Licence & attribution: Reuse terms not independently verified: the IMF's "Copyright and Usage" page (imf.org/en/about/copyright-and-terms) could not be retrieved when checked on 2026-07-26. World Bank Data360 records this dataset as "Public Use", and the IMF publishes it openly via its DataMapper API and a public spreadsheet. Reproduced here with the citation the IMF requires. © International Monetary Fund; not offered under an open licence, and explicitly not CC BY.
Known limitations
- Two indices, measuring the same idea differently. The Yaqtin AI Readiness Index is our own composite, rebuilt from World Bank open data as an annual series from 2010 to 2023. The IMF’s AI Preparedness Index is an authoritative external benchmark, but a single 2023 snapshot. They are never blended; switching source switches the whole dataset. Across the 148 countries both cover they agree closely — Spearman rank correlation 0.94 — which is a useful cross-check, not proof either is right.
- Our index is our responsibility. We chose the indicators, the scaling and the equal weighting. Those are defensible choices, not discovered facts, and a different set of reasonable choices would produce a somewhat different ranking. Treat it as indicative rather than authoritative.
- Neither index updates monthly. The underlying statistics — internet penetration, tertiary enrolment, R&D spending — are collected annually by national statistical offices, so there is no monthly ground truth to report. We check the upstream sources on a schedule and republish only when something actually changes.
- A country must have all four pillars to be scored. Averaging over whichever pillars a country happens to have flatters those with sparse statistics, because the pillar they lack is usually their weakest. Countries missing a pillar therefore show “no data” overall while still displaying the pillars they do have.
- Years are comparable because they share one ruler. Each indicator’s scaling bounds are computed once across every country-year and reused for all years. Scaling each year against its own distribution would measure only relative position — if every country improves, every score stays flat and real global progress disappears from the chart. On the shared ruler the world mean rises from 0.393 in 2010 to 0.494 in 2023.
- Each year uses only figures reported for that year. No carry-forward and no interpolation, so a country that skipped a reporting year shows a gap rather than a repeated value. Between 109 and 137 countries are scored in any given year.
- The series stops at 2023, not because progress did. National statistics arrive with a lag: 2024 currently has about 58 reporting countries and 2025 almost none. Publishing them would show a cliff that is an artefact of reporting, not a fall in readiness.
- The IMF’s reuse terms could not be verified. Their “Copyright and Usage” page was unreachable when checked; that source’s licence note says exactly what is and is not established. Our own index is published under CC BY 4.0.
- IMF pillars are weighted contributions, not sub-scores. The IMF’s four dimensions are published on a 0–0.25 scale and sum to its 0–1 overall index; ours are independent 0–1 sub-scores. Both are normalised for display, but they are not the same construct.
- Different providers measure different things; a country’s score is only comparable within one source and one year. That is why sources are never blended or averaged here.
- Normalising to 0–100 preserves each source’s ordering but not its distribution; small differences between countries may not be meaningful. The AIPI is itself built from indicators published by eight other institutions.
- Missing values are rendered as “no data” and excluded from ranks and charts — never zero-filled, never interpolated. Economies the provider does not cover simply have no record. Regional and income aggregates published alongside country rows are excluded.
- Country regions and income groups come from the World Bank; where the World Bank has not classified an economy, the income group shows as unclassified rather than being guessed.
- Pillar mappings across providers are editorial judgements documented per source above; the underlying methodologies differ.
Reproducing or updating the data
The data pipeline is documented in the repository README: a provider CSV is converted with scripts/ingest.ts, validated against JSON Schemas, and dropped into data/scores/<sourceId>/<year>.json. Adding a year or provider requires no code changes.