Does Long-Term AI Chatbot Use Cause Isolation? No One Knows

Researchers writing in STAT News are raising an uncomfortable question about conversational AI: whether people are slowly disconnecting from one another after months of leaning on chatbots for companionship, advice, and emotional support. Their core complaint is not a confirmed harm but a void. "What we don't know is whether people are disconnecting from one another, gradually, over months of interacting with chatbots," they write.

The problem is methodological. Most existing studies capture only short-term, snapshot effects, while the concern is cumulative and plays out over long stretches of real-world use. To study that pattern, researchers need longitudinal access to how people actually use products like ChatGPT, and that data sits inside private AI companies.

The authors argue the industry should open usage data to independent researchers so the mental health effects can be measured before the technology becomes further embedded in daily life. Without it, clinicians, regulators, and the public are left guessing about a product now used by hundreds of millions.