Abstract: Based on 623 publications on structural fire resistance indexed in the Web of Science Core Collection (2005–2025), this study applies CiteSpace to quantitatively map publication trends, journal/disciplinary distributions, and collaboration networks (authors, countries, and institutions), and to conduct keyword co-occurrence, clustering, and burst detection. A four-dimensional engineering lens (materials–members–fire scenarios–methods/design) is further used to semantically validate and consolidate thematic clusters. Results show three stages of development: slow exploration (2005–2007), rapid growth (2008–2021, 89.1% of the total; peak of 70 papers in 2021), and steady development (2022–2025). The United States, the United Kingdom, and China lead the output, with 156 papers from the United States. Gernay is the most productive author (19 papers), and the author collaboration network density is 0.0042. Keyword clusters are well-structured (Q=0.7846; S=0.9186), highlighting themes on fire performance and design, elevated-temperature material properties (concrete, A992 steel, and FRP, etc), key member responses (beams and columns), and the complementary use of OpenSees-based simulations and thermo-structural tests. Natural/traveling fire scenarios and AI-enabled prediction have intensified in recent years. Based on cluster evolution and burst patterns, this paper further distills testable research propositions and an engineering-oriented agenda for performance-based fire design and intelligent assessment.