Rule-Based Chatbots Improve Depressive Symptoms Modestly
Researchers conducted a systematic review and meta-analysis of 15 studies (2020–2025) comparing LLM-based and rule-based chatbots for depression and anxiety using robust variance estimation and random-effects REML pooling. They found rule-based chatbots produced a small but significant improvement in depressive symptoms (Hedges g=0.266, P=.04), while LLM-based results were non-significant with wide confidence intervals; anxiety outcomes were non-significant for both approaches. The study identifies a 4–8 week window as most effective and recommends larger trials to assess LLM interventions.
Key Points
- 1Found rule-based chatbots achieved small significant effect on depression (Hedges g=0.266, P=.04).
- 2Showed medium-term (4-8 weeks) interventions yielded greatest depression improvement versus blank controls.
- 3Recommend rule-based chatbots as scalable low-resource option; LLM evidence remains underpowered, needs larger trials.
Scoring Rationale
Provides actionable evidence favoring rule-based chatbots, but limited LLM sample size and wide CIs reduce overall certainty.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems
