Global research trends in clinical studies of ginseng: a bibliometric and text-as-data analysis

Clinical research on ginseng has expanded considerably, yet most bibliometric syntheses still rely on publication counts and co-occurrence maps. We mapped research on Panax ginseng and demonstrated a text-as-data and network pipeline. English-language publications indexed in the Science Citation Index Expanded (2004-2026) were retrieved using ginseng- and clinical-study-related terms and screened by two reviewers. CiteSpace and R-based clustering and citation-burst analyses were supplemented with topic modelling, word embeddings, a cross-validated study-type classifier, social-network analysis, negative-binomial citation modelling, Gini concentration, and disease-intervention and pathway mapping. Of 1386 records, 270 met inclusion criteria. Annual publications increased from six in 2004 to 26 in 2021, accumulating 6990 citations (h-index 44). South Korea produced 53.7% of publications, whereas Australia had the highest network centrality; international co-authorship remained low (24.4%) and did not increase. Seventeen CiteSpace clusters and eight validated topics indicated a shift from herb-drug-interaction safety toward supplementation, type 2 diabetes, cognition, and sexual or hormonal health. The author-keyword network showed a small-world structure centered on the intervention rather than a disease. Citations were concentrated (Gini 0.562), and only team size predicted citation rate (IRR 1.06). The classifier achieved 89.6% accuracy, while ginsenosides, NF-κB/inflammation, and insulin/AMPK were the molecular terms most frequently co-mentioned in mechanism-focused abstracts. Randomized designs comprised 190 studies (70.4%), with 10 observational (3.7%) and 70 (25.9%) other designs (uncontrolled, single-arm, pharmacokinetic, or mechanistic sub-studies). Clinical ginseng research is increasingly disease-specific but remains geographically concentrated, fragmented, and dependent on short-term trials. This pipeline quantifies uncertainty, reveals patterns beyond counts, and highlights the need for international collaboration, standardized ginsenoside reporting, and long-term cohorts.