Sources and data lineage.
Every external source behind the AIR APAC AI Readiness Index, Q2 2026 results, with its family, vintage, coverage, normalisation, known bias and algorithmic role. External sources inform country priors and benchmark context; they do not validate the order any company holds.
How to read these scores
What the external sources do, and what they do not
- Company scores
- Scores reflect public-evidence visibility, through RPS (Raw Public Signal), the visibility measure built by AIR APAC, and a hiring-adjusted public signal. Not a private audit and not a statistical census.
- Market context
- CRP (Country Readiness Prior) is a market-level context benchmark used to interpret organisation scores. CERS (Country Enterprise Readiness Signal) is a market-level evidence and readiness context benchmark that shrinks raw company visibility by evidence coverage.
- Coverage
- The index covers all 23 markets. An additional enterprise-adoption benchmark is available for 12 of those markets; the remaining 11 show macro and state context only.
The division of labour matters. The published order is built from the AIR APAC public-signal scan over the screened universe. The external sources below feed the country priors and the benchmark context only. They do not validate an individual company's order, and no external source can move a company score directly.
The source register
Every source, its vintage and its role
Read a row left to right: the source and its publisher, the coverage it gives us, its vintage, how it enters the build, and whether it is live in Q2 2026.
| Source | Publisher | Coverage | Vintage | Normalisation and algorithmic role | Status |
|---|---|---|---|---|---|
| Oxford Insights Government AI Readiness Index 2024 composite total, 40 indicators across the government, technology and data pillars. Known bias: government and public-sector weighted; collinear with the International Monetary Fund (IMF) index at r of about 0.91; partly a GDP-per-capita proxy. | Oxford Insights | 21 APAC markets of 23. Hong Kong and Macau missing. | 2024 | APAC average-rank percentile, before median combination with IMF. Macro and state readiness block, collapsed with IMF in v0.2. | Live |
| IMF AI Preparedness Index (AIPI) Composite over digital infrastructure, human capital, innovation and regulation. Known bias: macro-structural; pre-2023 data; collinear with Oxford; a wealth proxy. | IMF | 21 APAC markets | 2023 | APAC average-rank percentile, before median combination with Oxford. Macro and state readiness block, collapsed with Oxford in v0.2. | Live |
| Asia Pacific AI Readiness Index 2023, Business AI Readiness score The Business AI Readiness 2023 score from Table 1, on a 0 to 100 scale. Known bias: a business-readiness proxy built from statistical indicators; not a direct company-level adoption survey; covers 12 of 23 markets only. | Salesforce, prepared by Access Partnership | 12 APAC economies: Australia, China, India, Indonesia, Japan, Malaysia, New Zealand, the Philippines, Singapore, South Korea, Thailand, Vietnam | 2023 | APAC percentile rank across the 12 covered economies; missing markets not imputed. Enterprise and business-readiness block in CRP v0.2A and in CERS where country coverage exists. Does not validate company scores. | Live, limited |
| AI innovation and talent output Research output, patents and developer density, drawn from Stanford HAI, WIPO, Scimago and GitHub. Known bias: research output is not broad deployment. | Various | Not yet sourced | Various | Planned APAC percentile rank. Intended to give innovation capacity independent of web visibility. Not yet computed. | Not implemented |
| Digital infrastructure Connectivity, compute and cloud-region presence, drawn from ITU, GSMA and cloud-region records. Known bias: necessary but not sufficient. | Various | Not yet sourced | Various | Planned APAC percentile rank. Intended as a feasibility constraint block. Not yet computed. | Not implemented |
| AIR APAC public-evidence company scoring (RPS) Five dimensions: discoverability, AI and digital evidence, platform and data footprint, freshness, and capability visibility. Known bias: English-language web visibility; job-board and BPO skew in offshore hubs; sampling-conditional. | AIR APAC. Instruments built and operated by Exeter Labs. | 23 APAC markets; more than 5,000 companies, identity-safe | 2026 | Web search results collected through the Serper search service, scored with fixed APAC weights. Produces each company's RPS visibility score and the published order. Input to adjusted RPS and to the Adjusted Enterprise Readiness measure that is not published this release. | Live |
Vintage notes
What the vintages cost, and what is still missing
Two macro sources, one signal
The Oxford and IMF indexes move almost together (correlation 0.91). We therefore treat them as one signal about the economy, not two. Both are structural and both are a partial wealth proxy, which is why no market-to-market comparison of CERS or mean RPS is published.
The enterprise block is partial
The 2023 enterprise and business-readiness block exists for 12 of the 23 markets, which is why the Adjusted Enterprise Readiness measure is not published this release. The remaining markets carry macro and state context only, and no value is imputed for them.
English-only evidence
All 5,820 Q2 records were sourced in English. The sources above inherit that limit, and the effect is not yet measured. It is disclosed as a limit rather than corrected around.
How each source is used
The method page states the weight of every signal and the exact step at which each external input enters the build.