Ask someone today whom they turn to when feeling anxious or struggling to sleep, and increasingly the answer is a chatbot, not a friend or a doctor. In a survey of 19,000 adults across 18 countries that AXA conducted with Ipsos earlier this year, 61% of respondents said that they use AI for mental-health matters, with the share rising to more than 80% in some Asian countries. This implies that, for a significant segment of the global population, AI is a viable psychological-support tool.
The increased use of chatbots for mental-health prob lems is not entirely surprising, given that 26% of respondents showed symptoms of depression, anxiety, or stress at severe or more extreme levels—a three-percentage-point rise since 2022. Topping the list of stated causes for this “silent epidemic” were fi nancial in security and rapid economic and social change.
Always-online technology has also contributed to the deterioration in mental health. A majority of young adults reported that time spent on screens negatively affects their daily lives, even as they spend an average of more than five hours per day in front of one outside of work. This is the paradox of modern digital life: people recognise that their habits are not serving them well, yet social and professional pressures lock their behaviour in place.
It is worrying that people are readily turning to digital tools to cope with mental-health problems that, in some cases, such tools have exacerbated. But even more alarming is that among the respondents using a chatbot for mental-health matters, 42% said that they almost always follow the advice it gives. And while 38% said that they trust these platforms more than human mental-health professionals, 28% admitted that following a chatbot’s recommendation has led them toward harmful behaviour.
This is not a fringe activity. The survey shows a relatively new tool being used, often without guardrails, by people who frequently are the least equipped to evaluate what they are being told—either because they are inexperienced and unwell, or simply because a professional is unavailable. Globally, only 40% of people who needed mental-health support over the past year saw a professional, with cost, time, and stigma the most cited reasons for not seeking care.
People are integrating AI into their emotional lives not as an add-on to a well-functioning health-care system, but rather as an accessible entry point into care that has remained out of reach for most. The question is not whether AI belongs in mental-health treatment (it is already there, providing unsupervised psychological support). It is whether health leaders, employers, insurers, and AI builders take responsibility for shaping its role.
I see three reasons to be op timistic about the technology being steered in a beneficial direction, based on what the data show and on my own clinical experience.
First, AI can be harnessed to enhance access to mental-health support, acting as a bridge between a troubled individual and a complex system. For someone who cannot afford an appointment or is afraid to ask for one, a well-designed AI tool could help them name what they are feeling and point them toward appropriate care. There is certainly a need for such a tool: only 60% of respondents trust their friends and family to provide mental-health support, and nearly half of respondents with a job said that they would not discuss psychological struggles with their employer. An initial conversation with a chatbot that is private and judgment-free could be the difference between seeking help and staying silent.
Second, AI could save clinicians and caregivers time—a scarce resource. One of the most cited causes of burnout among health professionals is the heavy administrative load. And burnout often has cascading effects, degrading the quality of every interaction with patients. The development of responsible AI tools that can absorb documentation tasks would free clinicians to focus more on human connection, which is one of the strongest protective factors with regard to mental health.
Lastly, AI could serve as a pattern-finder. As the survey shows, many people already sense that their mental health is shaped by many overlapping factors, including work, finances, loneliness, screen time, and sleep. AI systems could help users—and the health systems supporting them—better identify and address these pressures by connecting them with the right kind of help sooner.
Of course, using AI in mental-health interventions requires robust safety features. AI-powered tools should be built and governed to disclose clearly what they can and can not do, and to escalate to a qualified human whenever risk is detected. Moreover, clinicians and patients must be involved in these tools’ development from the beginning, rather than being asked for feedback on finished products. Anything less could compound the very crisis these tools seek to address.
AI will not wait for health-related institutions to catch up. As more people with psycho logical problems turn to chatbots instead of calling a doctor, it becomes even more important that the tools at their disposal do more good than harm.
The writer is Chief Medical Officer of AXA Health Business.
