Talk to Me: A Preliminary Review on the Evolution and Impact of Emotional AI

This paper presents an ongoing and comprehensive systematic literature review on AI bots endowed with emotional capabilities (“emotional AI bots”) across a wide range of application domains. While emotional AI bots are increasingly integrated into digital interactions—spanning mental health support, customer service, online communities, education, and social platforms—existing research is highly fragmented and lacks a unified analytical perspective. To address this, we employ a generalized four-stage framework to categorize and synthesize the literature on emotional AI bots. The four stages—Promotion (enhancing positive emotional experiences and engagement), Prevention (identifying and addressing potential negative emotional states), Treatment (responding to acute negative emotions or critical incidents), and Recovery (supporting the restoration and maintenance of positive emotions and ongoing relationships)—are adapted from the Continuum of Mental Health Promotion and Prevention originally proposed by Mrazek & Haggerty (1994). This framework enables a structured analysis that transcends traditional domain boundaries and allows for the comparison of developments in both supportive and commercial contexts. Our review systematically covers representative studies from fields such as human-computer interaction, information systems, marketing, and applied psychology. Key research themes include the design and effectiveness of empathic chatbots, AI-driven emotion detection and expression in digital communication, and the impact of emotionally intelligent bots on user outcomes and experiences. The analysis reveals that most advances in emotional AI bots are concentrated in the Treatment stage, where real-time emotion recognition, adaptive dialogue strategies, and supportive interventions have been most actively developed and evaluated. Prevention-oriented research, such as early risk detection and proactive mitigation of negative emotions, forms a smaller but significant subset. In contrast, both the Promotion and Recovery stages remain relatively underexplored, with limited work focusing on fostering long-term emotional engagement, relationship building, or sustained well-being. These findings highlight important trends and persistent research gaps in the current landscape of emotional AI bots. They also suggest significant opportunities for future innovation in designing bots that are capable of supporting the full continuum of user emotional experiences, rather than being limited to reactive or crisis-oriented functions. By offering a unified framework for analysis, this review provides new insights for both researchers and practitioners seeking to advance emotionally intelligent AI across diverse digital environments.