Stanford Study Links Intense AI Companion Use to Worsened Well-being for Socially Isolated Individuals
A new Stanford study has found that for individuals with smaller real-world social networks, intense use of AI companions was linked to worse well-being, rather than improvement. The study surveyed over 1,100 users of Character.AI and analyzed hundreds of donated chat transcripts. A key finding was that opening up to a chatbot about personal matters left these users feeling worse, not relieved. This outcome is counterintuitive compared to human relationships, where sharing vulnerabilities typically fosters closeness and improves well-being. Researchers suspect this negative effect occurs because chatbots cannot genuinely reciprocate sharing and are engineered primarily to maintain user engagement, offering what they term a 'social snack'—a temporary fix for connection that lacks sustained nourishment. 
