AI mental health chatbots feel empathetic, but safety evidence still lags

A scoping review published in npj Digital Medicine assessed generative AI mental health chatbots. The review found that these chatbots are perceived as accessible and empathetic by users. However, evidence regarding their safety, sustained use, and clinical value is limited. The authors emphasize the gap between positive user experience and the lack of rigorous data on adverse effects.

The review examined multiple studies on AI chatbot interventions for mental health. Users reported feeling understood and supported, but systematic data on adverse effects or long-term outcomes were sparse. The authors highlight that while the chatbots appear engaging, there is insufficient evidence to confirm their therapeutic efficacy or safety. The review calls for more rigorous testing before these tools are deployed at scale.

The findings underscore the need for cautious implementation of AI in mental health settings. As chatbots become more common for delivering psychological support, regulators and clinicians must balance accessibility with proven safety and effectiveness. The review suggests that without proper evidence, the risks of harm or dependency remain unknown.