Climate Resilience and Artificial Intelligence in Tourism: A Two-Stage Bibliometric Review and Post Hoc Evidence Audit (2019–2025)

Background: Tourism is among the most climate-sensitive service industries, yet it is simultaneously undergoing rapid digital transformation through artificial intelligence (AI). The overlap and thematic relationships between the retrieved climate resilience and AI-in-tourism literature require empirical description. Methods: We conduct a PRISMA 2020- and PRISMA-S-informed, two-stage bibliometric science mapping review of Scopus and the Web of Science (WoS) Core Collection (2019–2025), executed in 2026 and implemented in R (bibliometrix) under the SPAR-4-SLR protocol, complemented by a post hoc-focused full-text evidence audit developed during revision. Stage 1 maps the two parent literatures (N = 9513; Stream A = 2409, Stream B = 7177, shared = 73); Stage 2 is a separately searched exploratory corpus (N = 440) that is not the mathematical intersection of the parent streams. Key findings. The parent fields are large and expanding (24.59% annual growth) but show limited retrieved set overlap: cross-stream overlap is below 1% (0.64% unscreened; 0.77% screened; equivalently 3.03% of Stream A and 1.02% of Stream B), a measure of retrieved set overlap rather than of disciplinary isolation. The retained exploratory corpus has a 53.79% CAGR, and 214 of its 440 records fall outside the screened parent union. Its early relevance screening trail could not be reconstructed, so these findings describe the retained dataset and do not establish field-wide growth or convergence. Post hoc audit. Applying substantive tourism, AI, and climate hazard criteria to 100 assessed full texts identifies 14 studies for an exploratory audit (eight scenario-based, three prediction, and three association), spanning snow, thermal exposure, flooding and landslides, water, and recreation; none demonstrates an evaluated AI-led adaptation benefit. Implications: The findings motivate testing AI-supported climate risk decisions in tourism; they do not demonstrate improved resilience. Non-retrieval of 29 of 129 sought reports (22.5%) and post hoc selection limit the audit.