\n<\/p>\n
Over a billion people in the world are living with some form of mental crisis, and the numbers aren’t seeing a downward trend. It is all hands on deck to find potential ways to address the rising public health concern. A recent study<\/a> revealed something quite encouraging. A conversational AI called Kai made a positive impact on the lives of university students dealing with mental health issues.<\/p>\n Students who interacted with Kai reported lower anxiety and an overall improvement in well-being compared to those who received group therapy. They also experienced greater relief from depression and higher life satisfaction than students who received no treatment.<\/p>\n The researchers are hopeful that conversational AI could become a scalable mental health resource, reaching more people while keeping support efficient and accessible. The findings are published in JAMA Network Open<\/i>.<\/p>\n Even though one in eight people suffers from mental health issues<\/a>, only about 25% of those in need actually receive care. A major reason is the lack of trained professionals needed to address the urgent issue. Traditional one-on-one therapy also faces significant constraints from uneven access to care and the persistent social stigma that discourages many from reaching out for help.<\/p>\n Mental health apps have attempted to improve access, but users often feel they are impersonal or not fluid enough to meet individual needs. This causes many people to drop from using such apps, as per studies, fewer than 4%<\/a> of users stay active after the first month.<\/p>\n The rise of generative AI has brought renewed focus to conversational chatbots, which scientists see as a promising way to expand access beyond traditional digital mental health tools that struggle to truly hold users’ attention. However, solid evidence remains limited regarding whether conversational AI<\/a> can meaningfully reduce psychiatric symptoms and whether people can form a therapeutic alliance with chatbots.<\/p>\n To study these aspects, the researchers recruited 995 university students who were experiencing emotional distress. To keep the students safe and ensure they received the right kind of support, those facing severe mental health challenges or active suicidal thoughts were not included and directed toward appropriate clinical care.<\/p>\n One-third of the participants interacted with a conversational AI platform called Kai, developed by a separate company where one of the study’s authors worked. Through natural, back-and-forth text conversations, Kai provided personalized psychological support, drawing on approaches like cognitive behavioral therapy to tailor its responses to each user.<\/p>\n The other two groups consisted of participants who attended traditional face-to-face group therapy sessions with professional psychologists, and a group that received no active treatment during the study.<\/p>\n The results showed that conversational AI outperformed both group therapy and no treatment in reducing anxiety. For depression, both face-to-face therapy and AI led to significant improvements.<\/p>\n Students using the AI also reported greater happiness and life satisfaction than those in the other groups. However, the AI failed to show improvement in PTSD symptoms, suggesting that trauma-related conditions may still require more specialized, human-led care.<\/p>\n About 35% of those recruited dropped out before the follow-up, which may have influenced the results. While the findings point to the potential of conversational AI being introduced as a supportive tool for expanding access to mental health care, more robust research is needed to truly understand its impact.<\/p>\n \n Written for you by our author Sanjukta Mondal<\/a>, edited by Sadie Harley<\/a>, and fact-checked and reviewed by Robert Egan<\/a>\u2014this article is the result of careful human work. We rely on readers like you to keep independent science journalism alive. Anat Shoshani et al, Efficacy of a Conversational AI Agent for Psychiatric Symptoms and Digital Therapeutic Alliance, JAMA Network Open<\/i> (2026). DOI: 10.1001\/jamanetworkopen.2026.6713<\/a><\/p>\n \n\t\t\t\t\t\t\t\t\t\t\t\t\tJournal information:<\/strong>Can AI help?<\/h2>\n
\n <\/figcaption><\/figure>\n<\/p><\/div>\n<\/div>\n
\n If this reporting matters to you, please consider a donation<\/a> (especially monthly). You’ll get an ad-free<\/b> account as a thank-you.\n <\/p>\n<\/p><\/div>\n\n\t\t\t\t\t\t\t\t\t\t\t\t\t\tPublication details\t\t\t\t\t\t\t\t\t\t\t\t\t<\/h6>\n
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\tJAMA Network Open<\/cite><\/a>
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t