Forecasting tourism demand with search query: comparison of social media and search engine

Purpose The proliferation of multisource digital information has intensified tourists’ cross-platform search dependency. This study aims to systematically compare the predictive capacity of search engine-derived queries (Baidu) with social media-generated search indices (TikTok) in tourism demand forecasting. Design/methodology/approach This study applies a case analysis of Jiuzhaigou, a major tourist destination in western China, with robustness verified through replication at Mount Siguniang, a smaller-scale destination in the same region. The Boruta algorithm was used to identify predictor variables across different types of search queries. Daily tourist arrivals were forecasted, and the prediction accuracy was subsequently verified across different models and multiple data sources. Findings The Baidu Mobile Index demonstrated significantly superior performance compared to other search engine indices, while the TikTok single search index exhibited a moderate edge over the composite indices. Overall, social media search data rivaled or occasionally surpassed search engine data in performance when used with artificial intelligence (AI) models. However, integrating multisource search data yielded only marginal improvements in prediction accuracy. Practical implications The empirical findings provide actionable guidance for governmental agencies and tourism operators, particularly regarding optimal search data source selection and predictive model implementation for demand forecasting applications. Originality/value To the best of the authors’ knowledge, this study represents the first application of social media search data in tourism demand forecasting, advancing research on the predictive efficacy of various terminal search data from search engine.