Mapping the Research Landscape of Artificial Intelligence Applications in English-Speaking Assessment: A Bibliometric Analysis from 2015 to 2025

This study carries out a bibliometric analysis of research on artificial intelligence (AI) applications in English speaking assessments published between 2015 and 2025. Drawing on 1,448 Scopus-indexed publications and using VOSviewer for analysis, the study examines publication trends, country/author contributions, keyword co-occurrence, citation networks, and thematic developments. Research output increased markedly after 2023, driven by rapid advances in deep learning-based speech processing, the wider adoption of automated speech evaluation technologies, and the growing demand for scalable digital assessment systems capable of supporting learning in increasingly technology-mediated educational environments. The USA‚ China‚ the UK, and Japan are the most productive countries․ They also exhibit the highest betweenness centrality in the international collaboration network․ Thematic analysis shows continued fragmentation in this field, with unsolved challenges including construct validity, ethical issues, and learner‑centered factors. The analysis reveals a rapid expansion of research after 2023, identifies the most productive countries, institutions, authors, and journals, and maps the evolution of research themes from automated speech recognition and pronunciation assessment to generative AI-supported speaking assessment. These findings establish the current knowledge structure of the field and identify emerging research directions that can inform the development, implementation, and evaluation of AI-supported English-speaking assessment. This study contributes by delivering a systematic overview of the field and identifying key directions for future research on valid, transparent, and pedagogically meaningful AI-assisted speaking assessment systems.