Artificial Intelligence in Banking: A Bibliometric Analysis of Research Trends, Intellectual Structure, and Performance-Related Themes
Purpose: This study provides a comprehensive bibliometric analysis of AI integration in banking, mapping research trends, intellectual structures, and performance-related themes. While AI’s growing importance in financial services has attracted scholarly attention, bibliometric studies specifically focusing on the AI-banking-performance triad remain limited. This study addresses this gap by systematically mapping the research dynamics, intellectual structure, and thematic evolution of this domain. Materials and Methods: Publications were retrieved from the Scopus database using predefined search criteria, resulting in 891 articles published between 2014 and 2024. Bibliometric indicators were employed to examine publication trends, authorship patterns, and institutional contributions. VOSviewer (version 1.6.20) and the R-based Bibliometrix (version 4.3.3) package were used to construct co-authorship networks, keyword co-occurrence maps, co-citation structures, and thematic maps. Results: Findings reveal exponential growth, particularly after 2018, with a peak of 277 articles in 2024. IEEE Access and Expert Systems with Applications are the leading sources, while Baesens B. emerges as a highly influential author. China and India dominate output, though European countries achieve higher per-article impact. Highly cited works focus on credit scoring, fraud detection, fintech, and financial inclusion. Conceptual mapping identifies five thematic clusters, with “AI in banking” as a motor theme, NLP as a niche, and credit detection as emerging. Conclusions: AI research in banking is rapidly expanding, interdisciplinary, and globally distributed. Theoretically, the findings are framed by TAM, TPB, DOI, and Dynamic Capabilities Theory, revealing that AI adoption represents a multi-level phenomenon spanning individual acceptance, institutional diffusion, and strategic reconfiguration. These theoretical lenses explain why fraud detection and credit scoring dominate as early adoptions while NLP and governance remain underdeveloped. The study highlights key contributors, emerging themes, and future research directions. However, findings are constrained by reliance on a single database (Scopus), exclusion of non-English and non-peer-reviewed sources, and inherent limitations of bibliometric methods.