Analyses reporting the absence of some Nobel laureates from earlier versions, the inclusion of journalists and editors among researcher records, and criticism of the scoring method highlight the list’s limitations.
Two questions underpin the debate: Are citation data collected accurately and comprehensively? And can those data justify describing researchers as “the world’s best scientists”?
John P. A. Ioannidis, a Stanford University researcher and one of the database’s developers, draws an important distinction on the second question. In an interview published in the Indian Journal of Medical Research, he explains that the data were not intended to serve as awards or certificates of recognition. Inclusion does not necessarily establish that someone is a great scientist, nor does exclusion mean that someone conducts poor science.
Some Nobel laureates were reported missing from the list
One specific source behind this discussion is an analysis published by John Philip on Mendeley Data on November 17, 2023.
Examining the career-long list covering data through the end of 2022, Philip reported that 12 of the 71 physics laureates and two of the 66 chemistry laureates awarded Nobel Prizes between 1998 and 2022 were absent. All 62 medicine laureates examined were included.
The physicists identified as absent from the career-long list included Andrea Ghez, Rainer Weiss, Arthur B. McDonald, and Albert Fert.
An important distinction is necessary here: This work was published as an openly available analysis note, not as a peer-reviewed journal article. Its findings concern the historical version examined; they do not establish that the same individuals are absent from the 2026 list. The analysis alone also cannot conclusively determine why those names were missing.
Peer-reviewed research also shows that Nobel recognition and citation rankings do not coincide
A study by Marek Kosmulski, published in Scientometrics in 2020, assessed 97 Nobel laureates in physics, chemistry, medicine, and economics from 2010–2019 using several bibliometric indicators.
In the career-long data examined, 90 of these laureates ranked among the top 100,000 by composite citation score. Seven fell outside that threshold, while 45 ranked among the top 6,000.
This finding cannot be interpreted as showing that seven laureates are absent from today’s expanded “top 2%” database. The study used a different threshold and data period. It does, however, demonstrate that contributions recognized at Nobel level and citation rankings do not measure the same thing.
The Nobel Prize is not a perfect measure of scientific value either. Nevertheless, the comparison exposes the problems with treating a ranking position as a direct measure of scientific superiority.
How is the “top 2%” determined?
The resource commonly known as the Stanford–Elsevier list is a database of standardized citation indicators compiled from Scopus records.
Selection is based on placement among the overall top 100,000 by a composite indicator called the “c-score,” or within the top 2% of a researcher’s primary subfield. Career-long citation impact and citations received during a single year are presented in separate tables.
A researcher who places in the top 2% of their subfield may therefore rank much lower overall without this necessarily indicating a calculation error. The two rankings use different comparison groups.
The ninth version, released on October 7, 2026, includes career-long data through the end of 2025 and citations received during 2025. The label “2026 list” does not mean that all scientific work published in 2026 has been evaluated.
Six indicators do not necessarily provide six independent measurements
The c-score combines total citations, the h-index, the coauthorship-adjusted hm-index, and citations to publications in different authorship positions.
A 2025 study by Marta Kuc-Czarnecka and Andrea Saltelli in Scientometrics examined the relationships between these indicators. Combining variables that carry similar information can cause some characteristics to influence the score more strongly than others.
The study’s central criticism is that the weights assigned in a formula may not match the indicators’ actual influence on the result. A mathematical formula therefore does not produce a measure of scientific value independent of methodological choices.
A 2026 letter to the editor by Gangan Prathap in the same journal also raised concerns about the ability of such composite indicators to distinguish size-related effects from characteristics associated with intensity or quality.
Researcher records included 105 journalists and editors
Data quality concerns are not merely theoretical.
While examining citation and authorship patterns in the health sciences, a study by Thomas Perneger, published in Research Integrity and Peer Review in 2023, identified records for 105 journalists or editors. Their texts were reported to have been incorrectly classified as original research. These records were excluded from the study’s analysis.
This finding does not diminish the value of science journalism. The issue is the treatment of scientific research and science reporting as though they were the same activity.
Perneger’s analysis of 18,231 researchers also showed that the h-index and hm-index favored different authorship patterns. The h-index was more strongly associated with middle-author positions, while the hm-index displayed a more balanced pattern. The choice of indicator can therefore influence which forms of work are rewarded.
High citation impact can coexist with retracted publications
Research published by Ioannidis and colleagues in PLOS Biology in 2025 provided evidence that high citation impact cannot be treated as an assurance of research integrity.
Among researchers on the career-long list examined, 7,083—equivalent to 3.3%—had at least one retracted publication. The figure was 4% for the single-year list. Among the top 1,000 researchers in the career-long ranking, 13.8% had at least one retracted publication.
These figures concern earlier datasets used in the study, not the 2026 list.
Furthermore, a retraction is not proof of misconduct by every coauthor. The reason for the retraction, the individual’s responsibility, and their publication history must be examined together. The limitation demonstrated by the data is clearer: A high citation score does not certify that a researcher’s publications are free of problems.
Removing self-citations changes the list
Self-citation occurs when researchers refer to their own earlier work and can be a normal part of scientific practice. High rates, however, warrant separate examination when interpreting total citation scores.
In their 2020 PLOS Biology article, the database’s developers reported that 4.9% of researchers who placed in the top 2% of their subfield for career-long impact when self-citations were included fell outside that group when self-citations were removed.
This is why the database’s provision of indicators both with and without self-citations matters. Announcements that selectively use only the more favorable figure can give readers an incomplete picture. Nevertheless, a self-citation rate alone cannot establish improper intent.
Whom does reliance on a single database make less visible?
A study by Carl P. Lipo, Beau DiNapoli, and Benjamin Andrus, published in Scientometrics in August 2026, compared Scopus and OpenAlex using the records of 600 researchers selected from the 2024 list.
The analysis of 564 researchers for whom coverage could be calculated reported differences associated with publisher and publication type. It raised questions about the representation of book-oriented disciplines and the higher coverage of researchers publishing in Elsevier journals.
This suggests that the same body of scientific work can be measured differently depending on the database used. Elsevier’s position as both the owner of Scopus and a major publisher increases the importance of independent scrutiny of database coverage.
The study does not, however, prove deliberate favoritism. OpenAlex, the comparison database, also has matching, classification, and duplicate-record problems. The findings support caution against treating results from a single source as universally definitive.
Do these criticisms invalidate the entire list?
An analysis published by Massimo Del Fabbro and colleagues in June 2026 identified problems involving document types, historical records, and retracted publications in the 2025 version.
Yet when the proposed filters were applied together, 99.30% of the top 1,000 researchers and 97.47% of the top 100,000 were retained. The findings demonstrate that some problems require correction, but they do not support the claim that the entire list is meaningless.
The database’s developer issues a clear warning to universities
Ioannidis’s message in his Indian Journal of Medical Research interview is brief: “These are not medals!”
He notes that some websites republish the data under grandiose titles and even sell certificates of recognition. He explains that these websites are not affiliated with his team and that the database was not created for that purpose.
Taken together, the research and these warnings indicate that the list can help assess citation impact, but cannot independently establish scientific quality, research integrity, or professional competence.
Instead of announcing that a researcher has been “named one of the world’s best scientists,” universities should specify which year’s citation table includes that researcher. Academic assessment should look beyond ranking positions to examine research methods, original contributions, reproducibility, and real-world impact. The foundational methodological paper behind the database also calls for citation indicators to be complemented by detailed qualitative assessment.
Sources
- John P. A. Ioannidis interview in the Indian Journal of Medical Research: “Meta-research, metrics, gaming and top 2 per cent scientists”
- John Philip, analysis of Nobel laureates’ representation in the Stanford–Elsevier lists, Mendeley Data, November 17, 2023
- Marek Kosmulski, study comparing Nobel laureates using bibliometric indicators, Scientometrics, 2020
- Ioannidis and colleagues, methodological paper on the Stanford–Elsevier top 2% database, PLOS Biology, 2020
- Marta Kuc-Czarnecka and Andrea Saltelli, study of weighting in composite citation indicators, Scientometrics, 2025
- Gangan Prathap, letter to the editor on the limitations of composite indicators, Scientometrics, 2026
- Thomas Perneger, study addressing the classification of journalists and editors in researcher records, Research Integrity and Peer Review, 2023
- Ioannidis and colleagues, study of highly cited researchers and retracted publications, PLOS Biology, 2025
- Carl P. Lipo, Beau DiNapoli, and Benjamin Andrus, comparison of Scopus and OpenAlex, Scientometrics, 2026
- Massimo Del Fabbro and colleagues, analysis of data and classification problems in the Stanford–Elsevier list, 2026
