Emotional AI as Psychological Support: Functional Limitations and the Reshaping of Relational ExpectationsScoping Review

Abstract

Artificial intelligence chatbots and conversational agents are being used more and more often to provide emotional support in mental health. A scoping review of interdisciplinary studies published from 2018 to 2025 was conducted to identify structural and functional deficiencies in emotional AI for mental health support. Five connected thematic deficiencies have been identified: a predictive rather than an interpretive understanding, instability of the long-term dialogic context, absence of clinical authority and accountability, affirmation-oriented interaction design, and the perceived sufficiency of AI support. These topics are then related to user experience and, considering emotion, relationship expectations, and professional help-seeking, interpreted. Although emotional AI is more readily available for people, it also has some drawbacks; thus, people's expectations of emotional support may be reshaped, and they may fail to receive timely professional help. Implications for research, clinical practice and system design are proposed.