Abstract Despite the availability of evidence-based treatments like CBT, many individuals still lack adequate care for anxiety and depression due to workforce shortages, systemic barriers, and cultural differences. AI-delivered CBT (AI CBT) offers a scalable solution to this issue, yet existing evidence remains fragmented and lacks diversity. This systematic review addresses this gap regarding AI CBTs. We searched Web of Science, PubMed, and PsycINFO for RCTs 2016 onward that evaluated AI CBT for anxiety or depressive symptoms. Of 208 studies screened, 16 met the inclusion criteria. Data was extracted and thematically synthesized. Study quality was assessed using the Cochrane Risk of Bias Tool 2. AI CBT showed limited efficacy for anxiety, with only 1 of 14 studies reporting significant improvements. Third-wave CBT and interventions targeting younger populations may show particular potential in reducing anxiety. For depressive symptoms, only 4 of 16 studies showed significant improvements. Across both conditions, highly preliminary evidence suggest that AI CBTs show less promise for older adults. Additionally, personalization and human support appeared to not be reliably associated with improved outcomes. Study quality also raised concerns. Overall, current AI CBTs require further refinement through high-quality and user-centered research. No funding went into this manuscript (PROSPERO: CRD42024615340).
Yip et al. (Fri,) studied this question.