ISSN : 2466-2542
This study categorised the information-searching difficulties experienced by library users, as identified in previous empirical research, and categorised the functions provided by AI search services implemented in libraries both domestically and internationally from the users’ perspective; it then examined the correspondence between these two categorisation systems and the level of empirical evidence supporting them. By combining a comprehensive literature review with a product environment scan, we identified nine types of difficulties encountered during the search process and five types of AI search functions based on task support criteria. We then assessed the level of correspondence by coding the supply of solutions for each functional type and the evidence—covering domain, user evaluation and outcome— using a scoring system. The analysis revealed that the combination for which the most substantial evidence of supply and validation had been accumulated was the function of searching for materials on similar topics and summarising search results. In contrast, functions addressing search failures caused by memory errors or mislabelling of materials emerged as the largest gap; despite academic research findings, neither validation within a library context nor commercial supply was confirmed within the scope of the study. Based on this, future tasks were proposed, including domestic empirical studies based on failed queries and the consideration of introducing these functions on a modular basis.