Abstract
BACKGROUND: This randomized controlled trial compared AI-driven and human peer counselor-delivered internet-based Cognitive Behavioral Therapy (iCBT) for depressive symptoms (primary outcome) in young adults. Addressing a gap in literature, we explored the comparative effectiveness and acceptability/perception of these two modalities of online intervention. METHODS: Ninety young adults were randomized to AI-driven iCBT, human peer counselor iCBT (participant-blinded), or a waitlist control. Interventions consisted of eight hours over four weeks. Depressive symptoms (primary outcome), suicidal ideation, and self-efficacy were assessed at baseline, week two, and week four. Qualitative analysis explored participant perceptions. RESULTS: Both iCBT interventions significantly reduced depressive symptoms (p < 0.05). No significant difference was observed between intervention groups at week two. The AI group improved significantly from baseline to week two (p = 0.004) but showed no further significant reduction by week four (a plateau effect), while the human group demonstrated continued improvement. Qualitative analysis indicated that participants valued AI’s convenience and accessibility but expressed concerns regarding its emotional understanding and personalization. CONCLUSION: AI-iCBT shows promising short-term efficacy, comparable to human counseling up to week two, but its limitations in emotional perception and sustaining therapeutic momentum resulted in a plateau effect. Future AI development must focus on improving emotional interaction and personalized support. TRIAL REGISTRATION: ChiCTR2400088423. Registered on 19 August 2024. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12888-026-07925-1.