AI research in libraries (2017–2025): a scoping review with bibliometric mapping across the COVID era

Abstract

Purpose This study examines how research on artificial intelligence (AI) in libraries evolved between 2017 and 2025, with specific attention to changes before, during and after the COVID-19 period. It seeks to identify shifts in publication trends, geographic participation, thematic emphases and methodological approaches and to assess what these shifts suggest about the changing focus and maturity of AI research in library scholarship and practice. Design/methodology/approach The study adopts a scoping review methodology following the Arksey and O'Malley framework, supported by PRISMA-ScR-based screening and selection procedures. A structured search of the Scopus database retrieved AI-related library research published between 2017 and 2025. The final dataset of 373 documents was analyzed using bibliometric profiling and author keyword co-occurrence mapping. Records were grouped into three periods – pre-COVID (2017–2019), during COVID (2020–2022) and post-COVID (2023–2025) – to enable comparative analysis of publication output, citation patterns, geographic distribution, thematic clusters and study designs across phases. Findings The results show a sharp increase in AI-related library research after 2020, with sustained high output after 2023 and an expected citation lag for the most recent year. Geographic participation broadens substantially over time, although publication output remains concentrated in a small number of countries and journals. Thematic analysis reveals a clear shift from early exploratory and infrastructure-linked research in the pre-COVID period, to service continuity, automation and operational concerns during COVID and toward post-COVID consolidation around governance, adoption, generative AI and AI-related literacies, alongside continued interest in service technologies. Methodologically, the literature becomes more diverse across periods, with growing use of quantitative, qualitative and mixed methods, although non-empirical studies remain the largest category overall. Originality/value By explicitly treating pre-COVID, COVID-period and post-COVID literature through 1 August 2025, this study extends earlier bibliometric work by examining publication trends, geographic participation, thematic clusters and methodological change within a single three-period framework. This makes visible changes that are less apparent in aggregate bibliometric analyses of AI-in-libraries research.