Bridging Code and Perception: An Integrative Review of AI-Generated Virtual Humans Informed by the TCCM–SOR Framework

As an emerging form of digital interaction, AI-Generated Virtual Humans (AI-VHs) are progressively assuming socially functional roles across entertainment, e-commerce, education, and other domains, attracting widespread attention from interdisciplinary scholars. However, existing research remains fragmented, lacking a systematic analytical framework that integrates technology, platforms, and user behavior, thereby hindering the revelation of their developmental trajectories and impact mechanisms. To address this gap, this paper employs the SPAR-4-SLR methodology to conduct a quantitative analysis and thematic review of 889 core publications from Web of Science and Scopus databases. It constructs a four-tier analytical framework: “technology, virtual human, platform, user” and integrates TCCM (Theory-Context-Characteristics-Method) and SOR (Stimulus-Organism-Response) perspectives to systematically identify the field's research actors, hot topics, and evolutionary trajectory. Findings reveal that this field has formed an interdisciplinary knowledge network centered on computer science and business/human-computer interaction. Research pathways exhibit an evolutionary pattern of “technological foundation, platform embedding, user transformation.” Early studies focused on 3D modeling and generative AI, while recent research has shifted toward virtual influencers and trust-building in live-streaming e-commerce, social presence, and consumer behavior mechanisms. These findings theoretically elucidate the evolutionary logic and impact pathways of AI-VHs within a multi-layered technology-society architecture. They provide practical foundations for technology developers to optimize virtual human generation and interaction systems, for platform operators to design scenario-based application strategies, and for relevant policymakers to advance technological governance.