Deep learning in finance: a bibliometric review

This study presents a bibliometric analysis of deep learning (DL) in finance, based on 3,874 publications retrieved from the Web of Science (WoS) for 2016 to 2025 with a recall-oriented query whose precision is validated on a random sample. It explores publication trends, geographical distribution, thematic evolution, and keyword dynamics. The findings reveal a steady growth in DL-related finance research over the sample period, highlight the field’s inherently interdisciplinary nature, and identify a structural shift after 2022 driven by the rapid surge of large language models (LLMs) and generative artificial intelligence (AI). Extending earlier syntheses to November 2025, the study emphasizes transparency and validation: it discloses the full keyword-canonicalization dictionary, validates the relevance of the retrieved corpus on a random sample, re-estimates the headline results on a validated DL-core sub-corpus and benchmarks them against the DL literature at large, and subjects the post-2022 shift to statistical and robustness checks, including the exclusion of the provisionally indexed 2025 records. The study concludes by outlining prospective avenues for future research.