This study systematically reviews research progress on micro-expression recognition in the context of artificial intelligence (AI)-assisted psychological counseling and its relevance to trust construction. Integrating theories of counseling relationships, affective computing, and human-AI trust models, it proposes a technology emotion-ethics integrative framework to organize existing evidence, reveal underlying connections, and guide future testable research directions. Current studies indicate that multimodal nonverbal cues—such as facial micro-expressions, vocal prosody, and physiological signals—can provide supplementary information for emotion and interaction understanding, yet their validity and reliability are significantly affected by contextual, cultural, and individual differences. Based on a technical review, this paper critically evaluates the ecological validity and cross-cultural generalization of major micro-expression databases, discusses algorithmic bias and potential misuse risks, and emphasizes the role of interpretability and human-AI collaboration mechanisms in trust maintenance, which require concrete verification in real counseling settings. Finally, the paper proposes future research agendas and practical implications to enhance the safety and trustworthiness of AI-assisted psychological interventions.
