Practicing love with algorithms: How online communities narrate and make sense of human–AI intimacy

As artificial intelligence (AI) becomes embedded in everyday life, emotionally intimate interactions with AI companions are becoming increasingly visible. Yet prior research has focused mainly on individual experiences, leaving limited understanding of how human–AI intimacy is collectively interpreted and negotiated in online communities. This study examines community discourse on RedNote, a major Chinese social and lifestyle platform, using 13,059 comments from 58 relevant posts. Combining BERTopic modeling with a human–large language model (LLM) hybrid emotion classification procedure, the analysis identifies three thematic clusters: intimacy as co-construction, intimacy as infrastructure, and intimacy as boundary negotiation. Co-construction centers on persona design, roleplay, and narrative world-building and is marked mainly by joy and curiosity. Infrastructure-related discourse is associated with frustration and sadness, while boundary negotiation reflects ambivalence toward intimacy with AI.