Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and...

Area of Science:

• Digital Health
• Artificial Intelligence
• Mental Healthcare

Background:

• Effectiveness, scalability, and safety of direct-to-consumer AI mental-health chatbots are not well-established.
• Uncertainty exists regarding their real-world impact and monitoring capabilities.

Purpose of the Study:

• To systematically review and meta-analyze the evidence on commercial AI mental-health chatbots.
• To assess their effectiveness for depression, anxiety, and loneliness.
• To evaluate completion rates, attrition, and safety reporting.

Main Methods:

• Systematic review and meta-analysis of peer-reviewed studies from 2015-2025.
• Random-effects models for randomized controlled trials (RCTs) and binomial-logit models for completion.
• Safety analysis focused on harms-reporting interpretability.

Main Results:

• Fifty-two studies (22 RCTs) involving 13 commercial chatbots were analyzed.
• Modest depression improvement (g = -0.35) was observed, below clinical significance.
• Anxiety effects were uncertain (g = -0.37), loneliness showed partial reduction (g = -0.21).
• Completion rates (79.5%) may overestimate retention; attrition was similar to controls.
• Harms reporting was uninterpretable, with low certainty across findings.

Conclusions:

• Commercial AI mental-health chatbots may offer modest benefits for depression.
• Evidence for loneliness is limited, and anxiety outcomes are heterogeneous, with potential for harm.
• Completion rates might inflate perceived efficacy; safety cannot be assumed without robust monitoring.
• These tools are best used as low-intensity adjuncts with built-in safety features.