A new paper posted on arXiv examines how sustained use of AI chatbots for companionship relates to well-being. The authors note that chatbots are increasingly used for emotional support and personal self-disclosure, but that little is known about how these social engagements unfold over time.
The study's central finding, as stated in the abstract, is that sustained AI companionship predicts lower well-being, and that this association is mediated by lower human interaction. In other words, the more people rely on AI companions, the less they engage with other people, and that displacement appears to drive the decline in well-being.
The paper does not suggest that all AI use is harmful, but it highlights a specific pathway of risk: when chatbots substitute for human contact, the long-term effects may be negative. The authors emphasize the need for further research into how these relationships develop over time.