A Japanese study followed 6,852 people aged 16 and older who were not receiving routine care. Regular emotional-support AI use was reported by 20.7% of the high mental-health-risk group and 4.6% of the low-risk group. The analysis linking mental health in 2024 with use in 2025 produced an adjusted association of 2.71, which should be distinguished from the roughly 4.5-fold raw proportion. This suggests people in distress may rely on AI more, not that AI caused or treated their symptoms. Persistent distress warrants connection to real clinical or counseling support rather than relying solely on chatbot conversations.
Back to Health
Emotional AI use and mental-health risk: distinguish association from cause
Chatbot use neither establishes treatment benefit nor replaces professional support.