A Bibliometric and Systematic Review of Retrieval- Augmented Generation Research: Evolution, Collaboration, and Emerging Themes

Retrieval-Augmented Generation, is a prominent AI technique in the field of artificial intelligence meanwhile, it integrates large-scale language models to boost the interpretability of output, contextual grounding, and factual accuracy. In this paper, we conduct a comprehensive bibliometric analysis to identify the scientific map of RAG research during the period 2018–2025.Findings reveal that there is an explosive growth in scientific production was fivefold between 2020 and 2024, coinciding with advancement of transformerbased models. A global overview of research evidences a profoundly interconnected, international collaborative context wherein top-ranked countries like U.S., Chinese, British, and Indian academic institutions also hold central places within the network of cooperation. Above all, the co-citation and thematic analyses successfully demarcate the core intellectual structure of the field, organized around three principal focuses: generative models enhanced by retrieval, knowledge-heavy applications, and approaches for fact grounding with language models. In particular, the prevailing tone of research has shifted from merely integrating basic components of retrieval to concentrate on trust and scalability benchmarks suitable for realworld applications within specific domains. The present study, from this perspective, not only clarifies the structural development of RAG but also acts as a crucial strategic guide for researchers and practitioners willing to push the frontier of Retrieval-Augmented Generation technology forward.