Trends in Multimodal Artificial Intelligence for Autism Research in Children and Adolescents: A Scientometric Study

Using scientometric and content-based approaches, this study explored the thematic structure and knowledge evolution in the application of artificial intelligence (AI) to autism spectrum disorder (ASD) in children and adolescents. A comprehensive search was conducted in the Scopus database up to June 30, 2025, using combined keyword sets related to ASD, AI, and child development, and 1401 English-language articles were retained. A bibliometric analysis was performed using VOSviewer to identify keyword co-occurrence patterns and citation linkages. The resulting thematic clusters were further analyzed through qualitative content examination. The analysis revealed a rapid growth in publications since 2015, with a surge after 2020, reflecting a growing interest in AI applications for ASD. Four important thematic clusters emerged: 1) AI-driven diagnostic models, 2) neuroimaging-based analysis, 3) early risk detection and behavioral assessment, and 4) emerging research on interventions and comorbidities, which remains comparatively limited. The United States, China, and the United Kingdom led in contributions, supported by active international collaborations. Highly cited and high-link-strength studies focused mainly on the use of deep learning in neuroimaging and early detection. Overall, the field demonstrates a strong diagnostic focus, increasing methodological sophistication, and interdisciplinary integration across computer science, neuroscience, and developmental psychology, leaving such areas as intervention and longitudinal tracking underexplored. These findings offer a comprehensive overview of the current landscape of the field and indicate critical opportunities for future development.