Effectiveness of a Fully Automated Mobile Therapeutic Versus a General Chatbot in Reducing Depression and Anxiety and Improving Well-Being: Feasibility Randomized Controlled Trial
Bibliographic record
Abstract
Background: Given the increasing prevalence of depression and anxiety disorders and enduring barriers to care, there is a critical need for alternative treatment options. Generative artificial intelligence (AI) chatbots show promise for increasing access to mental health care, though more direct research is needed to establish their efficacy. Objective: This pilot study aimed to test the efficacy of a generative mental health chatbot rooted in solution-focused therapy compared to the general-purpose ChatGPT and an assessment-only control (AOC) group on depression, anxiety, and well-being. Methods: A total of 185 English-speaking adults were recruited online and randomly assigned to one of three groups: AI therapy, ChatGPT, or AOC. Of these, 147 eligible participants filled out a pretreatment assessment. Over a 3-week period, the AI therapy group (n=44) was instructed to complete 3 structured, fully automated app-based sessions per week (9 total), while the ChatGPT group (n=60) was instructed to engage in 9 unstructured conversations with ChatGPT (GPT-4o-based models). The control group (n=43) received no intervention. In the AI therapy group, 39% (n=17) completed all sessions, as did 62% (n=38) of those in the ChatGPT group. Primary outcome measures, self-assessed online at baseline and postintervention, included the Patient Health Questionnaire-9 (PHQ-9), Overall Depression Severity and Impairment Scale (ODSIS) (depression), 7-item Generalized Anxiety Disorder Scale (anxiety), and World Health Organization Well-Being Index (5-item version) (well-being). Linear mixed effects models were used for data analysis. Results: Compared to AOC, both the AI therapy group (d=-0.47; P=.01) and the ChatGPT group (d=-0.44; P=.02) demonstrated significant reductions in depression scores measured by PHQ-9. The AI therapy group showed nonsignificant reductions in anxiety (d=-0.37; P=.11) and ODSIS depression scores (d=-0.25; P=.22) and an increase in well-being (d=0.12; P=.53) compared to AOC. Similarly, a nonsignificant reduction in anxiety (d=-0.27; P=.22) and ODSIS depression scores (d=-0.12; P=.53) and an increase in well-being (d=0.20; P=.29) were observed in the ChatGPT group compared to AOC. The AI therapy group did not significantly outperform the ChatGPT group on any outcomes (PHQ-9: b=-0.19; d=0.03; P=.87; 7-item Generalized Anxiety Disorder Scale: b=-0.57; d=-0.11; P=.62; ODSIS: b=-0.59; d=-0.13; P=.50; and WHO: b=-0.38; d=-0.07; P=.69). Conclusions: Both the structured generative AI chatbot and ChatGPT showed a significant reduction in depression scores compared to the control group. No significant effects were observed across other outcomes, although descriptive trends indicated improvements in anxiety. While the AI therapy group showed descriptively better outcomes for depression and anxiety, differences between groups were not significant. A larger sample and longer intervention may be needed for the emerging trends to yield clinically meaningful effect sizes.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.007 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.004 | 0.003 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.014 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".