How the rankings are made

What is counted

Research articles in OpenAlex, an open catalogue of scholarly work, published 2021–2025 in journals and proceedings on the CWTS core list, which is close to what Web of Science indexes. A paper belongs to a technology when its title or abstract matches the technology's keywords and its main research topic is on that technology's list. The topic check removes papers that only mention the technology. Retracted papers are excluded. One exception: large language models publish mainly at conferences and as preprints, so that technology counts papers and preprints from any source.

Every top paper is checked

Keywords and topics still catch papers that only mention a technology. So each paper in a technology's top tier is read by an AI model (Claude Haiku), which judges from its title and abstract whether the technology is the paper's subject. A second, stricter pass re-reads every paper the first one flagged, and a paper is left out only when both agree. Papers that use the technology as a tool stay in. Tested against 300 papers labelled by hand, every paper this check removed was one the hand labels also removed; it catches about three quarters of the clear-cut off-topic papers and leaves borderline ones in. The next most cited paper takes each removed paper's place. Each technology page lists the most cited papers it left out.

Records left out

Some catalogue records carry citations that belong to another paper. Two checks remove them before ranking. In every technology's top 1%, a record is left out when more than half its citations are dated two or more years before it was published. For preprints, each record's title is compared with arXiv's own; 21 records whose titles had been overwritten, while they kept a famous paper's citations, are left out. Each technology page lists any it lost.

How a country earns credit

Each paper is worth one point. The point is split equally between its authors, and each author's part is split between the countries of their institutions.

What "most cited" means

Papers are ranked by citations within their own year of publication, so a 2023 paper competes only with other 2023 papers. The top 10% and top 1% of each year are then pooled across the five years.

How it compares with a Web of Science ranking

Before this release, six of the technologies were built for 2019–2023, the years covered by ASPI's Critical Technology Tracker, which uses Web of Science. Agreement on the order of the top 20 countries, where 1.0 is identical:

Shares differ more than order does: China's share of the top 10% is 5 to 12 points lower here. Journal coverage, the citation window and review articles were each tested and do not explain it; differences in definitions explain part. This is a separate measurement, not a copy of theirs.

Who cites whom

Each technology page shows, for its leading countries, how many of the papers citing their top work come from the same country, next to the share you would expect from that country's size in the field.

For five technologies (quantum computing, quantum communication, post-quantum cryptography, atomic clocks, advanced aircraft engines) there is a third measure: home bias removed. For each of the most cited papers of each year, we count the citing papers with an author in any of the paper's own countries, remove the part of that count above what the countries' share of all papers in the field predicts, and rank the papers again. The same number of papers as the standard top 10% is credited as usual. For a paper, the expected home share is the chance that a random paper in the field has an author from any of its countries. Those pages also show what happens when every home citation is removed, for comparison; that version takes most from whoever writes most of a field, so it is not offered as a ranking.

A published study, Qiu, Steinwender and Azoulay's "Paper Tiger? Chinese Science and Home Bias in Citations" (Journal of International Economics, 2025), finds China's home bias the largest of major countries across science as a whole. It measures average citation impact across all fields; this site measures shares of the most cited papers in narrow technologies. The two answer different questions.

Known limits

• Citations measure attention from other researchers. They do not measure production, deployment or military capability.
• Some credit has no identifiable country. Each technology page states how much.
• The top 1% rests on few papers in small fields: 16 for atomic clocks. Any tier with fewer than 100 papers is marked as a small sample wherever its numbers appear.
• Where countries' shares are equal to one decimal place, their ranks are shown as a range, such as 11th to 14th. A rank that rests on fewer than five papers of credit is greyed.
• Countries cite their own work at different rates. The home-citation indicator on each technology page shows how much; the foreign-citations ranking exists for five technologies so far.
• The most recent years are the least settled: 2025 papers have had under two years to be cited. 2026 is shown only as a count of papers, never ranked.
• Keyword definitions catch some papers that do not belong. The most cited papers are listed on each technology page so you can judge.

Versions

Every definition is versioned; this build is version 0.5. When one changes, earlier numbers stay available under the earlier version.