Large language models are increasingly used by the public for mental health-related questions, including moments of acute suicidal crisis. Yet general-purpose AI systems are not specifically developed, clinically validated, or systematically evaluated for this highly sensitive context. This creates both a promising opportunity and a major responsibility: AI may offer immediate, scalable support in moments of crisis, but inappropriate responses may also carry substantial risks.AI for Suicide Prevention examines how artificial intelligence can be responsibly developed and evaluated for patient-facing suicidal crisis support. The project starts by evaluating how currently available general-purpose large language models respond to suicidal crisis situations, with a particular focus on clinical appropriateness, safety, and potentially harmful response patterns.Building on these findings, the project will develop and optimize an AI-supported crisis-support system specifically designed for acute suicidal crisis support and responsible use in mental health contexts. AI-generated crisis-support responses will first be assessed for clinical appropriateness and safety using standardized suicidal crisis scenarios evaluated by clinicians and individuals with lived experience. In a subsequent experimental study, the optimized crisis-support system will be compared with general-purpose large language models in simulated crisis conversations with LLM-based patient agents. In a later phase, the system will be examined in a pilot feasibility study with patients directly within psychiatric care. The project is conducted in cooperation with the TUM Klinikum, Department of Psychiatry and Psychotherapy, and the Chair of Health Informatics. The project aims to inform the responsible development and clinical evaluation of patient-facing AI tools for acute suicidal crisis support

Study protocol: Digital psychological interventions are effective in improving mental-well being, reduce psychological distress and prevent the onset of mental disorders, although standardised, one-size-fits-all approaches might limit uptake, engagement and effectiveness. Recent technological advances in AI allow for a greater personalisation of digital interventions. The ADAPT study aims to evaluate the efficacy, safety and user experience of two personalized, AI-supported digital mental health applications – an ML-based modular online program (Mindfit), and an LLM-powered chatbot, Ello - each independently compared to a no-intervention control group. ADAPT is a three-arm, open-label randomized controlled trial conducted online. Target participants are adults experiencing everyday psychological distress. Participants are randomized (1:1:1) to MindFIT, Ello or the no-intervention control group for 6 weeks, with an uncontrolled follow-up at three months post randomization. The primary outcome is perceived stress (PSS-10) at 6 weeks post randomization. Confirmatory secondary endpoints include well-being (WHO-5, SWEMWBS), self-efficacy (SWE), and depressive symptoms symptoms of depression (PHQ-9) and generalised anxiety (GAD-7). Additional exploratory outcomes include transdiagnostic risk factors, such as emotion regulation, social skills and avoidance. N = 420 individuals will be randomized to the three conditions. Statistical analyses of primary and secondary endpoints following an intention-to-treat approach using linear mixed models, as well as responder analyses have been prespecified. Additionally, a completer analysis will be performed and information about safety and usability will be descriptively reported. The ADAPT trial is among the few first rigorous RCTs to evaluate commercially available AI-based personalized solutions for mental health, including both a modular online program and an LLM-powered chatbot.

The University years are a developmentally crucial period and a peak age for the onset of mental disorders. The World Mental Health International College Student (WMH-ICS) initiative is aimed at developing and implementing a system for improving prevention and early interventions for mental health problems among college students. The initiative consists of three core elements. The first element is a web-based survey to assess the magnitude and nature of emotional problems, the effects of these problems on students' functioning, and barriers to seeking treatment. All first-year students in participating colleges (n= 71; 19 countries) were invited to participate. The second element is an infrastructure to test digital health interventions aimed at the prevention and early intervention in mental health problems. Participating colleges can develop and test digital interventions in randomized trials.The third element is the dissemination and continuous quality improvement monitoring of the evidence-based interventions developed in WMH-ICS. By addressing these three core elements, the WMH-ICS aims to integrate epidemiological and clinical research to offer scalable and effective evidence-based interventions for mental health problems at a critical life course stage.

StressLess is developing and evaluating a low-threshold, internet-based blended-care program for stress management aimed at TUM students. The pilot study examines acceptance, adherence, changes in stress experience, and the promotion of health-related habits. Digital self-study modules are combined with two guided group sessions.

In a pilot study, StudiCare Procrastination is investigating whether an integrated intervention combining online modules and group sessions improves acceptance and adherence among students who procrastinate. In addition, qualitative interviews are used to identify barriers, facilitating factors, and prerequisites for sustainable implementation.

The BESTFORCAN research consortium focuses on the dissemination of effective psychotherapeutic treatments for children, adolescents, and young adults after exposure to traumatic events.

The project investigates the effectiveness of psychological interventions for preventing major depressive disorder in adults who already experience clinically relevant subthreshold depressive symptoms but do not meet criteria for major depressive disorder at baseline. To this end, individual participant data from randomized controlled trials conducted internationally are systematically identified, pooled, and harmonized.Using an individual participant data meta-analysis, the project examines short- and longer-term intervention effects on the first or recurrent onset of major depressive disorder, depressive symptom severity, and clinically meaningful symptom change. A further focus is the analysis of potential effect modifiers at the participant and study levels, including sociodemographic and clinical characteristics, previous experience with psychotherapy, and characteristics of the interventions and their delivery. We will also examine, as a foundation for potential future precision mental health, machine-learning based prediction models, to predict intervention success on an individual level.The statistical analyses account for the hierarchical structure of the data, between-study differences, and missing data, and are complemented by sensitivity analyses. The overarching aim is to strengthen the evidence base for indicated prevention of depressive disorders and to identify approaches for targeted, personalized, and routinely implementable prevention services.

PSYCHOnlineTHERAPY investigates the integration of online interventions into routine outpatient psychotherapeutic care for patients with depressive disorders and anxiety disorders. The project implements and evaluates blended psychotherapy, a form of care that has so far received limited research attention and combines online treatment components with psychotherapy sessions delivered by psychotherapists as part of routine care. This new model of care is being implemented and evaluated within the framework of a pilot project aimed at improving psychotherapeutic care.The blended psychotherapy approach under investigation has the potential to improve psychotherapeutic care as it combines personal contact between patients and therapists with the possibility of adapting treatment to patients’ individual needs. At the same time, scientifically evaluated online components allow parts of therapy to be transferred into patients’ everyday lives. This may enable a more flexible delivery of therapy and contribute to an improved transfer of therapeutic content into patients’ daily routines. In addition, blended psychotherapy may increase therapists’ treatment capacities and thereby increase the number of available therapy places.As part of the PSYCHOnlineTHERAPY project, blended psychotherapy is being investigated in a three-arm cluster-randomized non-inferiority study in two variants: PSYCHOnlineTHERAPYfix, which follows a predefined sequence of online sessions and joint therapy sessions with a therapist, and PSYCHOnlineTHERAPYflex, in which the sequence of online sessions and joint therapy sessions is determined flexibly at the therapist’s discretion. Both variants are compared with standard psychotherapy as delivered in routine care, consisting of therapy sessions with a psychotherapist either on site or via video.A maximum of 16 therapy sessions are delivered in each condition, composed of joint therapy sessions and online sessions depending on the study arm. The joint therapy sessions consist of cognitive behavioral therapy delivered by licensed psychotherapists in Baden-Württemberg who participate in the PNP selective contract of AOK Baden-Württemberg and Bosch BKK. Access to the evidence-based online lessons, which patients complete independently, is provided via a secure online platform. Treating therapists can assign different lessons individually according to the needs of participating patients.The scientific evaluation of the project includes both quantitative and qualitative assessment methods at several measurement points. The two variants of blended psychotherapy are evaluated with regard to their effectiveness, cost-effectiveness, acceptance, and feasibility.Funding & CooperationInnovation Fund of the Federal Joint Committee for New Forms of CareUlm University, Department of Clinical Psychology and PsychotherapyAOK Baden-WürttembergBosch BKKMEDIVERBUND AGFreie Liste der PsychotherapeutenGerman Psychotherapists AssociationMEDI Baden-Württemberg e.V.