AI vs. Therapist: Mental Health Support in a War Zone

Source: Resume Genius / Unsplash

In war, help is not always a person. Sometimes, it is a chatbot.

A recent study published in BMC Psychology offers one of the most compelling real-world tests of AI psychotherapy yet.

Researchers conducted a randomized controlled trial with 104 women living in active combat zones in Ukraine—all of whom had been diagnosed with anxiety disorders. Half received traditional therapy with licensed psychologists three times a week. The other half used an AI-powered chatbot named Friend, designed to provide real-time, emotionally responsive psychological support.

The findings provide valuable insight into where AI excels—and where it still falls short—in the mental health landscape.

More Improvement With Human Therapists

The research design compared AI-powered chatbot support to human psychotherapy during ongoing war conditions. Participants were assessed using standardized clinical tools—the Hamilton Anxiety Rating Scale and the Beck Anxiety Inventory—at the start and end of an eight-week intervention.

The Friend chatbot offered daily, on-demand support via natural language processing and machine learning, drawing on techniques such as cognitive behavioral therapy (CBT) and motivational interviewing. These types of therapy are more readily offered by AI chatbots since both are based less on the relationship compared to other types of therapy such as relational work or psychodynamic psychotherapy. The control group had access to live therapists through regular video or in-person sessions.

Both groups experienced statistically significant reductions in anxiety. But the therapist group saw a 45% drop on the Hamilton scale and a 50% drop on the Beck scale—substantially greater than the 30% and 35% reductions observed in the chatbot group.

Why This Matters: AI as a Mental Health Adjunct

These results confirm what many clinicians and digital health experts have suspected: AI cannot yet match the emotional nuance and therapeutic depth of human connection. And yet, that is not the most important takeaway.

The true value of AI lies not in replacing therapists, but in extending care where therapists cannot reach. In this study, the Friend chatbot proved immensely valuable for women living in a war zone, where human support was irregular or inaccessible. The chatbot was available 24/7, and it was scalable, cost-effective, and adaptive. The AI chatbot offered a sense of continuity, structure, and basic emotional support and it did reduce anxiety symptoms, even though it was not as healing as human therapists.

AI can help bridge the gap—especially in crisis settings and underserved regions—by serving as an adjunct tool in mental health care.

The Empathy Gap: What AI Still Lacks

While the chatbot could adjust its responses based on sentiment analysis and learning algorithms, it lacked what we might call "emotional bandwidth." It could mimic empathy—but not feel it. It could guide, but not witness. For individuals in crisis, that absence of a real relational presence makes a difference.

The greater improvement in the therapist group speaks to the power of the therapeutic alliance: the nonverbal cues, the adaptive feedback, the human holding environment. These are elements that current-generation AI—however intelligent—cannot replicate.

This is why the authors, and many mental health experts, suggest a hybrid approach. AI can provide triage, psychoeducation, and emotional regulation coping strategies. Human therapists, meanwhile, can do the deeper relational work, process trauma, including experiential work, and provide more nuanced interventions.

The Friend chatbot’s model also points to a potential future in which culturally and linguistically adapted AI tools serve local populations more effectively than generic global apps. In this case, Friend was designed specifically for Ukrainian speakers and tailored for wartime stressors.

The Ethical Edge: Privacy, Trust, and Regulation

Of course, such tools raise critical questions about privacy, data protection, and informed consent. AI systems need access to sensitive psychological data to function well. This study addressed such concerns by adhering to the EU's General Data Protection Regulation (GDPR) standards, but widespread adoption will require broader regulatory frameworks and transparency about how data are used, stored, and protected.

Artificial Intelligence Essential Reads

Do You Have Bixonimania?

The Skill AI Can’t Replace: Agency

Moreover, trust in AI—especially in the domain of mental health—remains a cultural and psychological hurdle. People must feel that they are not just being processed by an algorithm, but actually heard and understood. Building trust will depend not only on performance metrics, but also on transparency, user experience, and integration with ethical clinical frameworks.

The Future: Augmentation, Not Replacement

This study affirms that AI-powered tools can provide real psychological relief, though not as powerful as human interventions—but that AI tools should not be viewed as stand-alone solutions. They work best as part of a hybrid system, augmenting rather than replacing the therapist’s role.

The future of mental health care will not be a binary choice between human and machine. It will be collaborative, layered, and dynamic. In moments of crisis, AI can be a source of support, but it still has limitations, especially if operating autonomously without professional oversight. In the long journey of healing, the human presence remains irreplaceable for now.

Copyright © 2025 Marlynn Wei, MD, PLLC. All rights reserved.