The Algorithm That Is Making PTSD Treatment Work for Veterans

An AI-assisted PTSD treatment programme developed in the Netherlands is producing better outcomes than conventional therapy for military veterans. Here is how it works.

The Dutch Ministry of Defence has been running a controlled trial of an AI-assisted post-traumatic stress disorder treatment programme since 2024, and the results published in the first quarter of 2026 are remarkable enough to have triggered immediate interest from defence ministries across Europe and from civilian mental health systems facing similar challenges.

The programme, developed by a consortium of Radboud University Medical Centre, the Netherlands Defence Academy, and several technology partners, uses a combination of real-time physiological monitoring, personalized adaptive cognitive behavioural therapy delivered partly through a conversational AI interface, and structured human therapist oversight to address the specific features of combat-related PTSD that standard therapeutic protocols handle poorly.

The specific innovation is in the adaptive personalisation. Conventional CBT for PTSD follows a protocol designed for the average patient — the average symptom profile, the average trauma exposure, the average complicating factors. Veterans with combat PTSD frequently have non-average symptom profiles: hypervigilance that prevents engagement with standard exposure therapy, moral injury dimensions that standard CBT frameworks don't adequately address, and specific triggers related to combat scenarios that civilian therapists who haven't served are poorly equipped to engage with.

The AI component of the programme continuously adjusts the therapeutic protocol based on physiological data (heart rate variability, galvanic skin response) that tracks the patient's stress level in real-time during sessions, patient-reported outcomes collected between sessions, and pattern recognition from the thousands of previous patients whose de-identified data trained the system. The human therapist remains central — they make all clinical decisions — but the AI provides them with real-time analysis that dramatically improves their ability to read where the patient actually is rather than where a protocol assumes they should be.

The trial results show a 34 percent improvement in validated PTSD symptom scores after 16 sessions, compared to 21 percent improvement for the standard protocol. The dropout rate — a critical issue in PTSD treatment where patients frequently disengage before completing sufficient sessions to achieve durable benefit — was 18 percent, compared to 31 percent in the standard protocol.