Randomized controlled trials evaluating generative artificial intelligence-enabled mental health interventions among samples with clinically significant baseline symptoms: a systematic review and narrative synthesis
Clinical Psychology Review · 2026-11 · 102814
DOI: 10.1016/j.cpr.2026.102814 ↗Kurz gefasst
Systematische Übersicht in der Zeitschrift Clinical Psychology Review, 2026-11. Verfasst von Adam C. Jaroszewski und 5 weiteren Autor:innen. Die Arbeit nennt 74 Literaturangaben.
Abstract (Originalsprache)
Generative artificial intelligence (GenAI)-enabled interventions for mental health have expanded rapidly in recent years. To clarify what is known about their efficacy, we conducted a systematic review and narrative synthesis of randomized controlled trials (RCTs) evaluating mental health interventions that incorporated GenAI into treatment delivery. We excluded RCTs whose samples did not report clinically significant symptom levels on average at baseline. N = 10 RCTs met inclusion criteria. Most (90%) were published in 2025 or 2026. Trial design and quality varied. Samples were predominantly younger adults and skewed female. All trials used GenAI systems that were clinically customized to help power participant-facing chatbots. Most interventions (70%) were based on cognitive behavioral therapy (CBT). Most trials (60%) did not report whether any adverse events had occurred. Preliminary efficacy evidence indicated that clinically customized GenAI-enabled mental health interventions may improve symptoms relative to inactive comparators, such as a waitlist. However, evidence of efficacy relative to active comparators was limited, and efficacy evidence regarding general-purpose GenAI chatbots was mixed. There were recurring methodological limitations across trials: all trials relied on self-report outcomes; most trials did not require all participants to report clinically significant baseline symptoms, mask participants to allocation, or systematically assess treatment acceptability. Overall, GenAI-enabled trials show clinical promise. However, more rigorous trials are needed to determine if and for whom GenAI provides meaningful benefit and to establish its safety, acceptability, and effectiveness in clinical populations.
Zitieren
APA
Jaroszewski, A. C., Jonathan, G. K., Jones, G. K., Patel, A., Jacoby, R. J., & Wilhelm, S. (2026). Randomized controlled trials evaluating generative artificial intelligence-enabled mental health interventions among samples with clinically significant baseline symptoms: a systematic review and narrative synthesis. Clinical Psychology Review, 102814. https://doi.org/10.1016/j.cpr.2026.102814
Vancouver
Jaroszewski AC, Jonathan GK, Jones GK, Patel A, Jacoby RJ, Wilhelm S. Randomized controlled trials evaluating generative artificial intelligence-enabled mental health interventions among samples with clinically significant baseline symptoms: a systematic review and narrative synthesis. Clinical Psychology Review. 2026:102814. doi:10.1016/j.cpr.2026.102814
BibTeX
@article{Jaroszewski2026,
author = {Jaroszewski, Adam C. and Jonathan, Geneva K. and Jones, Grant K. and Patel, Anushka and Jacoby, Ryan J. and Wilhelm, Sabine},
title = {Randomized controlled trials evaluating generative artificial intelligence-enabled mental health interventions among samples with clinically significant baseline symptoms: a systematic review and narrative synthesis},
journal = {Clinical Psychology Review},
year = {2026},
pages = {102814},
doi = {10.1016/j.cpr.2026.102814}
}
EUVIMED