Generative AI as Emotional Support: A Scoping Review of Emotional Disclosure, AI Companionship, Psychological Dependence, and Human Help Seeking

Abstract

Generative artificial intelligence has moved from a mainly task based technology to a conversational system that people may use when they are lonely, worried, stressed, or simply in need of someone to listen. This scoping review looks at how people use generative artificial intelligence, or AI, as a source of emotional support. It brings together findings from experiments and surveys about four connected topics, namely emotional disclosure to AI, AI companionship, psychological dependence on AI, and changes in human help seeking behavior. Across many studies, people appear willing to share personal feelings with AI chatbots quite freely, at times more freely than with another person, because the AI feels less judging and less likely to gossip. Controlled experiments show that when a chatbot shares emotion back with the user, the user tends to open up more and feels more satisfied with the talk. Long running studies of companion apps such as Replika show that regular users can form real emotional bonds with their AI companion, and this bond can grow across repeated conversation in a pattern that looks like ordinary human relationship building. At the same time, a number of studies raise concern that leaning too heavily on an AI companion may lower a person's motivation to build human relationships or to seek a licensed professional during a mental health crisis. This review uses theories such as Computers Are Social Actors, Parasocial Relationship Theory, and Social Penetration Theory to explain why people respond to AI the way they do. The review closes with practical suggestions for safer design of emotional support AI and with directions for future research that should measure long term outcomes rather than single short conversations only.