Interpret and write up the bibliometric review

A bibliometric paper is not a collection of charts. Reviewers expect you to turn the numbers and maps into a story about the field: how it developed, what it is built on, where it is now and where it should go next. This lesson shows how to get from analysis to manuscript.

From results to findings

1. Name every cluster by reading its most cited and most central documents. “Cluster 3” is not a finding; “Cluster 3: explainable and trustworthy AI in clinical decision support” is.
2. Connect performance and maps: are the most productive countries also central in the collaboration network? Are the most cited authors in the intellectual base also active in the current fronts?
3. Triangulate trends: emerging themes on the thematic map, recent terms in the overlay view, keyword bursts and recent coupling clusters should tell a consistent story. In the demo data, they all point to large language models, generative AI and deep learning as the newest fronts, with telemedicine and patient satisfaction rising since 2019.
4. Explain, do not just describe: why did output accelerate in a given year? Why is a theme niche rather than motor?

Building a research agenda

The research agenda is often the most cited part of a bibliometric review. Derive it systematically:

• Emerging themes: topics that are rising but not yet developed need foundational work.
• Basic themes: central but undeveloped topics need deeper, more specific research.
• Niche themes: well-developed but isolated topics could be connected to the mainstream.
• Gaps between clusters: weak links between themes (for example, between AI methods and patient satisfaction) suggest interdisciplinary questions.
• Geographic gaps: regions with little output or collaboration.

Formulate each agenda item as a concrete research question, not a vague suggestion.

The structure of the paper

1. Introduction: why the field matters, why a bibliometric review is needed now, and the research questions.
2. Methods: database, search string and date, inclusion criteria, PRISMA diagram, cleaning steps, software, and the settings of each analysis.
3. Results: performance analysis: main information, production, sources, authors, countries and citations.
4. Results: science mapping: conceptual, intellectual and social structure.
5. Discussion: interpretation, the research agenda, implications.
6. Limitations: database coverage, language, document types, keyword dependence, citation lag, name ambiguity.
7. Conclusion.

The methods paragraph

Records were retrieved from Scopus on 15 October 2026 with the search string reported in Table 1, restricted to articles and reviews in English published between 2008 and 2025. After removing duplicates and screening, 418 documents were analysed in Bibliometrics Studio (Xavier, 2026). Keyword synonyms were merged and search terms removed before co-word analysis. Co-occurrence and co-citation networks used association-strength normalisation and Louvain clustering; thematic maps followed Cobo et al. (2011), with centrality and density computed from the co-word network. The selection process is summarised in a PRISMA 2020 flow diagram (Figure 1).

Adapt the example to your study; the Studio’s export page writes a version that describes your actual settings.

Figures and tables

A typical bibliometric paper has 6 to 10 figures: PRISMA diagram, annual production, a source or country chart, the co-occurrence network, the thematic map, thematic evolution, the co-citation network and the collaboration map. Save networks as SVG for vector quality, or as 300 dpi PNG. Make sure labels are readable at journal column width; show fewer nodes rather than illegible ones. Tables usually cover main information, top sources, authors, countries and the most cited documents.

Limitations to acknowledge

• Coverage: one or two databases never include everything; non-English and regional journals are under-represented.
• Time lag: recent papers have had little time to be cited.
• Keywords: maps depend on author and database indexing practices.
• Name ambiguity for authors and institutions.
• Bibliometric indicators measure attention and structure, not the quality or validity of findings.

You have completed the course. The final assessment covers all eight lessons; pass it to download your certificate, and use the Studio for your own review.