Stanford study: sustained AI companionship predicts lower well-being by displacing human interaction

stanfordnlp · x · 2026-09-11

A Stanford team (Yutong Zhang, Dora Zhao, Diyi Yang, et al.) ran a two-wave longitudinal study of CharacterAI users: 1,182 surveyed at baseline and 439 re-surveyed after a mean 12-month follow-up.

Key findings:

The authors argue the results carry design implications: AI companion products should avoid boosting engagement at the cost of real-world social interaction.

Related event: Year-Long Study Finds Heavy AI Companion Use Lowers Well-Being(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →