During Israel's period of heightened security tensions, when anxiety and depression surged across the population, approximately 1,000 Israeli students found themselves facing an unexpected therapist: an AI-powered conversational platform called Kai operating quietly in their messaging apps. A new study from Reichman University, published in JAMA Network Open, reveals that this digital companion didn't just offer support—it matched and, in some measures, outperformed human-led group therapy.
The research, led by Prof. Anat Shoshani of the Baruch Ivcher School of Psychology, tackled a question growing more urgent as mental health crises deepen worldwide: Can artificial intelligence bridge the gap between those who need help and a system that rarely has enough clinicians to provide it? The answer emerging from this rigorous randomized clinical trial suggests something remarkable may be possible.
Researchers divided participants into three groups. Some received traditional in-person group therapy with psychologists. Others were placed on a waitlist. A third group gained access to Kai, which operates within a familiar messaging platform and is available around the clock. What Kai could do mattered: the system was grounded in established therapeutic frameworks including cognitive behavioral therapy, acceptance and commitment therapy, dialectical behavior therapy, mindfulness practices, and positive psychology. It could sustain conversations over weeks, recall personal details from previous interactions, detect signs of emotional distress, and deliver real-time tools like breathing exercises and guided reflective writing.
The results startled even seasoned researchers. Among participants presenting with clinical anxiety levels, 58 percent moved into the healthy range following the intervention. Kai users experienced anxiety reduction that actually exceeded what the in-person group therapy cohort achieved. Nearly half of those experiencing depression reported substantial improvement. Yet perhaps most striking was what researchers call "therapeutic alliance"—the intangible sense of trust and connection that typically takes months to build with a human therapist. Participants rated the AI system as empathetic, professional, and supportive at levels comparable to human clinicians. Many revealed it felt easier to confide in a digital system, freed from fears of judgment or embarrassment.
Persistence also mattered. Mental health apps typically suffer steep dropout rates, yet 61 percent of Kai users remained engaged throughout the entire 12-week study, using the platform an average of three days per week. The system's ability to remember users, their histories, and their contexts created a sense of continuity that kept people coming back—especially during vulnerable moments: on the bus, in the library, during sleepless nights.
"AI sits in the user's pocket," Prof. Shoshani reflected, "providing continuous support precisely when it is needed. The goal is not to create machines that sound more human, but to build systems that help make our society more human."
The researchers are careful to acknowledge limits. Kai is not a substitute for human treatment of complex conditions like PTSD. The platform operates within a hybrid model where human clinicians are available 24/7, and when the system detects acute distress or risk, it alerts a human team for immediate intervention. This blended approach—AI as continuous companion, humans as crisis responders—may point toward the mental health infrastructure the world actually needs: not replacement, but intelligent support that bridges the distance between crisis and care.