(Para)Social by Design? A Systematic Review of parasocial relationship Forms with Generative AI

Relationships with generative AI are increasingly described using parasocial concepts. Parasocial interaction refers to situational experiences of perceived social engagement with a persona, whereas parasocial relationships denote more enduring cognitive and emotional bonds formed toward such personas over time. Classical parasocial theory conceptualizes both phenomena as one-sided and unidirectional. The dialogic characteristics of generative AI systems raise questions about the applicability and conceptual boundaries of parasocial theory in human–AI interaction. Recent empirical research shows that users can experience parasocial interactions and develop parasocial relationships with generative AI systems. Perceived responsiveness, personalization, and conversational continuity are associated with emotional engagement, trust, and feelings of companionship toward AI agents. The aim of this scoping review is to synthesize existing research on parasocial interaction and parasocial relationships. The review addresses the following research questions: (1) Which aspects and characteristics of parasocial interactions and relationships have been identified in empirical and conceptual research on generative AI? (2a) In which usage contexts do different aspects and characteristics of parasocial interactions and relationships with generative AI occur (e.g., work-related, therapeutic, educational, romantic, or friendship-oriented contexts)? (2b) Which psychological, relational, and behavioral outcomes are associated with parasocial interactions and parasocial relationships with generative AI, and how do these outcomes vary across usage contexts? (3) How can parasocial theory be conceptually extended to situate interactions with generative AI along a continuum between parasocial and social interaction while remaining distinct from interpersonal social relationships? To address these questions, a systematic literature search was conducted in scientific databases, including Web of Science, Scopus, and PsycINFO. The search strategy combined keywords related to generative AI (e.g., generative AI) with terms capturing parasocial and relationship concepts (e.g., parasocial interaction).