A systematic review and meta-analysis published in npj Digital Medicine on May 29, 2026, has confirmed the efficacy of AI- and rule-based conversational agents in providing mental health interventions. The study analyzed 48 randomized controlled trials involving 28,071 participants to determine the impact of these digital tools on depression, anxiety, and stress levels.

Researchers observed small-to-moderate but statistically significant reductions in symptoms across the evaluated conditions. The results showed a standardized mean difference (SMD) of −0.27 for depression, −0.20 for anxiety, and −0.26 for stress. To ensure the reliability of these findings, the team employed a Robust Variance Estimation (RVE) approach to account for dependencies within the studies, yielding results that remained consistent with the random-effects model.

Subgroup analyses revealed that conversational agents were particularly effective when used in clinical populations and during shorter-duration interventions. Meta-regression further indicated that the specific timing of outcome measurements had no appreciable influence on the results. While the overall risk of bias across the studies was found to be low, some data gaps—specifically related to missing information or selective reporting—were identified. However, the study noted that publication bias was negligible.

These findings suggest that conversational agents offer a scalable solution for addressing the growing global burden of psychological distress, especially where traditional care access is restricted. The authors conclude that while these digital interventions show promise, further research is necessary to investigate their long-term effectiveness and to determine the most effective methods for integrating these technologies into existing mental healthcare infrastructures.

A systematic review and meta-analysis published in npj Digital Medicine on May 29, 2026, has confirmed the efficacy of AI- and rule-based conversational agents in providing mental health interventions. The study analyzed 48 randomized controlled trials involving 28,071 participants to determine the impact of these digital tools on depression, anxiety, and stress levels.

Researchers observed small-to-moderate but statistically significant reductions in symptoms across the evaluated conditions. The results showed a standardized mean difference (SMD) of −0.27 for depression, −0.20 for anxiety, and −0.26 for stress. To ensure the reliability of these findings, the team employed a Robust Variance Estimation (RVE) approach to account for dependencies within the studies, yielding results that remained consistent with the random-effects model.

Subgroup analyses revealed that conversational agents were particularly effective when used in clinical populations and during shorter-duration interventions. Meta-regression further indicated that the specific timing of outcome measurements had no appreciable influence on the results. While the overall risk of bias across the studies was found to be low, some data gaps—specifically related to missing information or selective reporting—were identified. However, the study noted that publication bias was negligible.

These findings suggest that conversational agents offer a scalable solution for addressing the growing global burden of psychological distress, especially where traditional care access is restricted. The authors conclude that while these digital interventions show promise, further research is necessary to investigate their long-term effectiveness and to determine the most effective methods for integrating these technologies into existing mental healthcare infrastructures.