Citation network analysis is a family of bibliometric techniques that map the citation relationships between papers, authors, or journals as a network graph, then use that graph’s structure — clusters, hubs, bridges — to reveal how a research field is organized, where it is heading, and which works or authors sit at its intellectual center. Where a single citation count answers how many times has this paper been cited, citation network analysis answers a structural question: which papers, people, or topics cluster together, and why.
It is distinct from citation indexing, the underlying infrastructure (Web of Science, Scopus, Crossref, OpenAlex) that captures and stores citation links in the first place. Citation network analysis is what you do with that indexed data once it exists — a research assessment and literature-review technique, not a database.
The two foundational relationship types
Almost every citation network technique reduces to one of two underlying relationships, both formalized in the information-science literature decades before today’s visualization software existed:
Co-citation
Two papers are co-cited when a third, later paper cites both of them together. The more often two papers are cited together across the literature, the stronger their co-citation link, and the more likely they address a related problem — even if the two papers never cite each other directly and were published independently. Co-citation analysis was introduced by Henry Small in 1973 as a way to map the intellectual structure of a scientific specialty from the citing behavior of the field itself, rather than from author-assigned keywords or subject categories.
Bibliographic coupling
Two papers are bibliographically coupled when they both cite the same third, earlier paper in their own reference lists. Unlike co-citation, this relationship is fixed at the moment of publication (a paper’s reference list doesn’t change), so bibliographic coupling is often used to find papers addressing similar problems at roughly the same point in time, while co-citation analysis is better suited to tracking how a field’s structure evolves as new papers accumulate. The technique was introduced by M. M. Kessler in 1963.
A third, simpler relationship — direct citation (paper A cites paper B) — is also mapped, typically to trace a research lineage forward or backward from a single seed paper, rather than to find topically similar clusters.
What the resulting network shows
Once citation links are assembled into a graph, standard network-analysis measures translate into research-assessment terms:
- Clusters in the graph typically correspond to sub-specialties or distinct research fronts within a broader field.
- Node size is usually mapped to citation count or publication count, so influential papers or prolific authors are visually larger.
- Central or bridging nodes — papers or authors connecting otherwise separate clusters — often indicate interdisciplinary or foundational work.
- Temporal layers (color-coded by publication year in most tools) show how a field’s research fronts have shifted, and can surface emerging topics before they show up in citation counts, since a new cluster of mutually citing recent papers is visible before any of them accumulates many citations individually.
Tools used for citation network analysis
Four tools account for most current citation-network work in academic settings, split roughly between dedicated bibliometric-mapping software (used for systematic, field-level analysis, often across hundreds or thousands of records) and lighter interactive literature-discovery tools (used to explore the neighborhood of a smaller set of papers, often for a single literature review).
VOSviewer
Developed by Nees Jan van Eck and Ludo Waltman at the Centre for Science and Technology Studies (CWTS), Leiden University, VOSviewer is free software for constructing and visualizing bibliometric maps from co-citation, bibliographic coupling, co-authorship, and keyword co-occurrence data. It is typically used with a bulk export (a set of records, often thousands) from Web of Science, Scopus, or a similar database, and produces the density- and cluster-map visualizations most common in published bibliometric-analysis papers.
CiteSpace
Developed by Chaomei Chen at Drexel University’s College of Computing and Informatics, CiteSpace is free Java-based software focused specifically on detecting and visualizing trends in scientific literature — research fronts, citation bursts (sudden spikes in a paper’s or term’s citation activity, often an early signal of an emerging topic), and structural turning points in a field over time. It is frequently used alongside VOSviewer in published bibliometric-review papers, with each tool covering complementary analysis angles rather than one replacing the other.
Litmaps
A citation-mapping and literature-monitoring tool aimed more at individual researchers and literature reviews than field-level bibliometric studies: it builds an interactive citation map from a seed set of papers and can monitor citation activity going forward, alerting the user when new papers cite into their map. See CASRAI’s dedicated guide: Litmaps: Citation Mapping and Literature Monitoring Tool.
ResearchRabbit
A free, citation-mapping literature-exploration tool that expands outward from a seed paper via authors, related works, and citation links, building growing collections rather than a single static graph. See CASRAI’s guide: Research Rabbit: Citation-Mapping Tool Guide.
Two related tools worth knowing in the same space: Connected Papers builds a one-shot visual graph of papers similar to a single seed paper, based on shared-reference (co-citation/bibliographic-coupling-style) similarity; OpenAlex and Semantic Scholar both expose citation-graph data via free, open APIs that underpin many of these visualization tools and are increasingly used directly for programmatic citation-network work. For a side-by-side comparison of the interactive discovery tools specifically, see Connected Papers Alternatives: Litmaps, ResearchRabbit, and Other Citation-Mapping Tools.
Applications
Literature review
Citation network analysis is increasingly used as a systematic complement to (not a replacement for) traditional narrative and systematic literature review. Two common uses:
- Bibliometric review: analyzing co-citation or bibliographic-coupling clusters across a large corpus of records to map the intellectual structure of an entire field — identifying its major sub-topics, most influential works, and how the field has evolved — typically published as a standalone bibliometric-analysis paper using VOSviewer and/or CiteSpace.
- Snowballing and gap-finding: using a tool like Litmaps, ResearchRabbit, or Connected Papers on a smaller seed set to surface papers a keyword search alone would miss (because they use different terminology for the same concept), and to spot under-cited connections between subfields that a manual search would not surface.
Research assessment
Citation network position is also used, alongside conventional bibliometric indicators such as the Field-Weighted Citation Impact (FWCI) or Relative Citation Ratio (RCR), as a qualitative input to research assessment: identifying whether a researcher’s or institution’s output sits at the center of a field, bridges multiple fields (a common signal of interdisciplinary impact that a raw citation count alone won’t distinguish from within-field impact), or occupies an emerging cluster before that cluster’s citation counts have caught up with its actual influence. It should be read as one input among several — CASRAI’s citation cartel and coercive citation entries cover ways that raw citation-network data can be gamed, which is exactly why the San Francisco Declaration on Research Assessment (DORA) and the Coalition for Advancing Research Assessment (CoARA) agreement both caution against relying on any single citation-based metric, network position included, as a standalone judgment of quality.
Limitations to keep in mind
- Coverage differs by source database. A network built from Web of Science data, Scopus data, or an open source such as OpenAlex will not be identical, since each database indexes a different (overlapping but non-identical) set of journals, conferences, and citation links — particularly for non-English-language or regional publications, which open bibliographic databases now generally cover more broadly than the legacy subscription indexes.
- Recency bias. Very recent papers have had little time to accumulate citations, so a citation network can under-represent genuinely important recent work — citation-burst detection (as in CiteSpace) is specifically designed to partly offset this by flagging early acceleration rather than absolute counts.
- A citation link is not an endorsement. Papers are sometimes cited critically, in passing, or as required background rather than as genuine intellectual influence; co-citation and bibliographic coupling infer topical relatedness from citation patterns, not from the sentiment or weight of any individual citation.
