Bots become companions through listening, which can secretly worsen loneliness.

Over-reliance on AI may reduce genuine human connection and social skill development.

AI comfort can damage autonomy; we must choose genuine human engagement.

Something is happening in the intimate space between us and our screens. A subtle shift that no one expected is underway. Two studies, Anthropic's analysis of 4.5 million Claude conversations in 2025, and an OpenAI/MIT Media Lab evaluation of 40 million ChatGPT interactions in 2026—tell a story that their headline numbers seem to contradict. Both teams lead with reassurance: emotional engagement is rare, companionship accounts for a small fraction of use. That framing, while accurate, points away from the finding that matters most.

In the Anthropic data, long conversations point to a tendency that is anything but uplifting. In exchanges of 50 or more messages, users move into processing psychological trauma, navigating workplace conflicts, and exploring questions of personal meaning. When users face existential dread or loneliness AI-mediated coaching sessions morph into companionship. The bot is not invited as a friend. It is becoming one—through the accumulated weight of being listened to, never judged, always available. Following the same pattern the OpenAI study brings to light that users tend to follow the advice that they receive from “their” bot; but instead of improved well-being the evolving dynamic leads to more loneliness and a decrease in perceived quality of life.

These findings confirm a cost that we tend to not factor in when we delight in our expanding artificial treasure chest.........

© Psychology Today