AI systems increasingly interact with users on an emotional level. The simulation of emotional intelligence has become a fundamental aspect of human–AI design. This meta-analysis synthesizes findings from 32 empirical studies (N = 4720) investigating the relationship between elements of emotionally intelligent AI and three major user outcomes: empathy, trust, and engagement. Across various platforms, from chatbots to social robots to voice assistants, emotionally expressive features such as empathic tone, backchanneling, and affective design were strongly related to increased user responses. The analysis revealed a moderate, consistent effect size (r = 0.31), suggesting that emotional intelligence consistently impacts socio-affective experiences of human–AI interaction. Implicit emotional cues were strongest, often outperforming explicit self-report measures. While benefits are evident, ethical problems exist, such as emotional manipulation, anthropomorphism, and cultural conflict. Implications of this work contribute to theory and practical advice toward creating ethically responsible, emotionally intelligent artificial intelligence systems.