The corpus in six numbers
- scholar → citing-paper citation edges
- 1,140,065scholar → citing-paper citation edges
- distinct citing papers
- 953,433distinct citing papers
- fully-mapped scholars (2,875 incl. partial)
- 2,334fully-mapped scholars (2,875 incl. partial)
- distinct citing institutions
- 173,287distinct citing institutions
- ISO country codes observed
- 231ISO country codes observed
- of 2,607,701 citing authorships have a resolved country
- 43.6%of 2,607,701 citing authorships have a resolved country
Citation counts are usually reported as a single number. This report asks a different question: where do citations come from? Every scholar mapped on CitationMap between July 2025 and June 2026 contributes their citing papers, and every citing author's institutional affiliation is geocoded and canonicalized against the Research Organization Registry (ROR). The result is one of the larger publicly described citation-geography datasets we're aware of: 2,607,701 citing authorship records, of which 1,136,133 (43.6%) resolve to one of 231 ISO country codes. Five findings follow. Methodology and known limitations are documented in full below.
1. China and the United States are effectively tied as the world's largest sources of citations
Of the 1,136,133 country-resolved citing authorships in the corpus, China accounts for 289,636 (25.5%) and the United States for 283,854 (25.0%) — a gap of less than six thousand records, or about half a percentage point. Together the two countries produce just over half of all geocoded citations. The United Kingdom, in third place, is roughly a fifth the size of either.
| Rank | Country | Citing authorships | Share of geocoded total |
|---|---|---|---|
| 1 | China | 289,636 | 25.5% |
| 2 | United Statesincl. 22,870 records labeled “USA” merged | 283,854 | 25.0% |
| 3 | United Kingdom | 54,847 | 4.8% |
| 4 | Germany | 49,738 | 4.4% |
| 5 | India | 30,807 | 2.7% |
| 6 | Italy | 29,172 | 2.6% |
| 7 | Canada | 27,968 | 2.5% |
| 8 | Australia | 26,224 | 2.3% |
| 9 | France | 25,795 | 2.3% |
| 10 | South Korea | 23,967 | 2.1% |
| 11 | Japan | 22,615 | 2.0% |
| 12 | Spain | 20,317 | 1.8% |
| 13 | Switzerland | 18,548 | 1.6% |
| 14 | Brazil | 15,202 | 1.3% |
| 15 | Netherlands | 14,776 | 1.3% |
| 16 | Singapore | 9,897 | 0.9% |
| 17 | Sweden | 9,853 | 0.9% |
| 18 | Iran | 8,914 | 0.8% |
| 19 | Belgium | 7,682 | 0.7% |
| 20 | Taiwan | 7,322 | 0.6% |
Top 20 citing countries by geocoded citing authorships. Shares computed against the 1,136,133 country-resolved authorships, not the full 2.6M corpus. One paper with three authors in the same country counts three times — see methodology.
2. Asia now produces nearly half of institution-resolved citations
For the 755,044 authorships that resolve to a ROR-canonicalized institution with continent data, the continental split is lopsided: Asia alone accounts for 49.1% — more than North America and Europe combined. Africa and South America together contribute 3.0%, a gap that says at least as much about the geography of research funding (and of this corpus's field skew) as about the geography of talent.
| Continent | Citing authorships | Share |
|---|---|---|
| Asia | 370,612 | 49.1% |
| North America | 194,014 | 25.7% |
| Europe | 145,409 | 19.3% |
| Oceania | 22,681 | 3.0% |
| Africa | 12,875 | 1.7% |
| South America | 9,453 | 1.3% |
Continental distribution of the 755,044 authorships resolvable to a ROR-canonicalized institution with continent data.
3. The median mapped scholar is cited from 15 countries
Across all 2,875 scholars with geographic data, the median number of citing countries is 15 and the mean is 22.3. Two-thirds of scholars (66.4%) are cited from 10 or more countries; fewer than one in eleven (8.7%) is cited from only one or two.
| Citing countries per scholar | Scholars | Share (n=2,875) |
|---|---|---|
| 1–2 countries | 251 | 8.7% |
| 3–5 countries | 333 | 11.6% |
| 6–10 countries | 461 | 16.0% |
| 11–20 countries | 706 | 24.6% |
| 21+ countries | 1,124 | 39.1% |
The practical implication: for an active, internationally visible researcher, international impact is the norm, not the exception. This matters for anyone building an EB-1A or O-1 "international recognition" argument: the question is not whether your work has crossed borders — it almost certainly has — but how far it ranks against this distribution. A scholar cited from 21+ countries sits in the top 39.1% of this corpus; one cited from 40+ is in genuinely rare territory.
4. The widest-reaching scholars are cited from over two-thirds of all countries on Earth
At the extreme right of the distribution, the ceiling is striking. Yann LeCun is cited from 139 countries and Yoshua Bengio from 127. For context, there are roughly 193–195 UN member states: the most-cited researchers in this corpus have measurable citation footprints in over two-thirds of the countries that exist. Both maps are publicly browsable, so these numbers can be verified paper by paper.
5. Eight of the top 15 citing institutions are Chinese
The institution-level table tells the same story as the country table, but sharper. The Chinese Academy of Sciences is the single largest citing institution in the corpus, and 8 of the top 15 institutions are Chinese. Stanford is the highest-ranked US institution at #3; the highest-ranked European entries are Cambridge (#9) and Oxford (#12).
| Rank | Institution | Location | Citing authorships |
|---|---|---|---|
| 1 | Chinese Academy of Sciences | China | 8,024 |
| 2 | Tsinghua University | China | 6,803 |
| 3 | Stanford University | United States | 6,609 |
| 4 | Zhejiang University | China | 5,606 |
| 5 | Shanghai Jiao Tong University | China | 5,299 |
| 6 | Peking University | China | 4,738 |
| 7 | Fudan University | China | 4,462 |
| 8 | University of Washington | United States | 4,437 |
| 9 | University of Cambridge | United Kingdom | 4,231 |
| 10 | University of Michigan | United States | 4,223 |
| 11 | Huazhong University of Science and Technology | China | 3,915 |
| 12 | University of Oxford | United Kingdom | 3,832 |
| 13 | MIT | United States | 3,634 |
| 14 | University of Science and Technology of China | China | 3,534 |
| 15 | National University of Singapore | Singapore | 3,480 |
| 16 | CERN | International | 3,472 |
| 17 | Sun Yat-sen University | China | 3,441 |
| 18 | Sichuan University | China | 3,323 |
| 19 | Harvard University | United States | 3,257 |
| 20 | University of Pennsylvania | United States | 3,203 |
Top 20 citing institutions by geocoded citing authorships, after ROR canonicalization. One known institution-dedup artifact (a US medical center that was absorbing unrelated records) was excluded — see methodology.
If your field's citation network looks nothing like this table — if it's clustered in three institutions on one coast — that asymmetry is itself a finding. Our guide to finding a research gap covers how to use geographic citation gaps as evidence of an under-synthesized literature.
Methodology and known limitations
Everything in this section is load-bearing. Read it before quoting any number above.
- Sample and time window. The corpus is every scholar whose citation map was generated on citationmap.com between July 2025 and June 2026 — 2,334 fully-mapped scholars, 2,875 including partially mapped scholars with geographic data. This is a self-selected sample: it skews toward computer science, AI, and biomedical fields, because those are the people who generate citation maps. It is not a census of global science.
- Country attribution. Each citing author's institution affiliation string (from Google Scholar citation records) is geocoded and canonicalized against the Research Organization Registry (ROR). Of 2,607,701 citing authorship records, 1,136,133 (43.6%) resolve to a country. The unresolved majority have missing or ambiguous affiliation strings; we leave them out rather than guess.
- Unit of analysis. Counts are citing authorships, not citing papers: one paper with three China-affiliated authors contributes three records to China's total. This weights research cultures with larger author lists upward. We consider it the right unit for measuring institutional engagement, but a per-paper recount would narrow the gaps at the top of the country table.
- Label merging. The United States total merges 22,870 records whose country label was the literal string "USA" into the canonical United States bucket. No other manual merges were applied to the country table.
- One exclusion. A single known institution-dedup artifact — a US medical center whose ROR record was absorbing unrelated affiliation strings — was excluded from the institution table. No other rows were removed or adjusted.
- Replication. The underlying per-scholar maps are publicly browsable on citationmap.com, so individual data points (LeCun's 139 countries, Bengio's 127) can be independently checked. We welcome corrections: if you find an error, we will fix the report and document the change on this page.
Cite this report
An, T. (2026). The Geography of Citations: Where the World's Research Influence Comes From. CitationMap. https://citationmap.com/guides/citation-geography-report
Journalists, librarians, and bloggers may reuse the tables and figures in this report with attribution and a link to this page. No further permission is required.
Frequently asked questions
Where does the data in this report come from?+
From the CitationMap production corpus: every scholar whose citation map was generated on citationmap.com between July 2025 and June 2026. For each scholar we record the papers that cite their work (from Google Scholar citation records), then resolve each citing author's institution affiliation by geocoding the affiliation string and canonicalizing the institution against the ROR (Research Organization Registry). The snapshot for this report was computed on June 10, 2026: 2,334 fully-mapped scholars (2,875 including partially mapped scholars with geographic data), 953,433 distinct citing papers, and 1,140,065 scholar-to-citing-paper edges.
Why do only 43.6% of citing authorships have a resolved country?+
Country attribution requires a usable affiliation string. Of 2,607,701 citing authorship records in the corpus, 1,136,133 (43.6%) resolve to a country; the rest have missing, truncated, or ambiguous affiliation strings that neither geocoding nor ROR matching can place with confidence. We deliberately leave those unresolved rather than guess. There is no obvious reason resolution failure would be strongly correlated with one region, but we cannot rule out residual bias — for example, affiliation strings in non-Latin scripts may fail more often.
Does counting authorships instead of papers inflate China's numbers?+
It weights them upward, yes — and the report says so explicitly. The unit of analysis is the citing authorship: one citing paper with three China-affiliated authors contributes three records to China's total. Fields and research cultures with larger average author lists (and Chinese institutions do tend to have larger author teams in many STEM fields) therefore weigh more than a per-paper count would give them. We chose authorships because they measure how many researchers at an institution are engaging with a body of work, which is the quantity the underlying maps visualize. A per-paper, fractional-credit recount would narrow but not eliminate the China–US gap at the top of the table.
Is this corpus representative of all academic research?+
No, and we don't claim it is. The corpus is selection-biased toward the people who generate citation maps: it skews toward computer science, AI, and biomedical fields, and toward scholars with an active reason to demonstrate international impact (job market, grants, EB-1A/O-1 visa petitions). It is best read as a large, internally consistent sample of citation geography for active, internationally visible researchers — not as a census of global science. Field-normalized databases like OpenAlex or Web of Science will give different absolute numbers; we'd expect the broad geographic pattern (China–US parity at the top, the long international tail per scholar) to hold.
Can journalists and bloggers reuse these tables and figures?+
Yes. The tables and figures in this report may be reused with attribution and a link to this page. Suggested citation: An, T. (2026). The Geography of Citations: Where the World's Research Influence Comes From. CitationMap. https://citationmap.com/guides/citation-geography-report. The underlying per-scholar maps are publicly browsable on citationmap.com, so any specific claim can be spot-checked. If you find an error, email us — we'll correct the report and note the change.
Explore the data yourself
Every number in this report aggregates maps you can open directly: start with Yann LeCun (139 citing countries) or Yoshua Bengio (127), or generate a map for any scholar and see where their citations fall against the distributions above.