Study Finds No Significant Mental Health Benefit From Maternal AI Chatbot

A recent study published in JAMA Network Open has evaluated the effectiveness of Rosie, an AI-powered chatbot designed to support maternal health. The research, which focused on the concept of maternal care and support, examined whether access to digital health advice could improve mental health outcomes for pregnant and postpartum individuals. The findings suggest that the chatbot offered no statistically significant advantage in managing anxiety or depression when compared to a control group receiving a monthly children's book program.

The AI chatbot Rosie showed no significant mental health advantage over a children's book program at 12 months.

The randomized trial enrolled 400 participants across the United States between October 2023 and December 2024. The study group consisted of English-speaking individuals aged 14 or older who were either pregnant or parenting an infant under six months old and identified as members of racial or ethnic minoritized groups. Participants were divided equally into two groups, with one group receiving access to the Rosie chatbot and the other receiving a monthly children's board book for one year. Both groups maintained access to study staff, and the Rosie intervention included human support from a clinical social worker who reviewed distress alerts.

Researchers measured outcomes using the GAD-7 questionnaire for anxiety and the PHQ-9 questionnaire for depression. At the 12-month follow-up, which concluded in December 2025, the data showed no clear benefit or equivalence between the two approaches. The adjusted difference for anxiety was 0.10 points, while the difference for depression was -0.23 points. Because the confidence intervals for both measures included zero, researchers noted uncertainty regarding the direction or size of any potential effect.

Understanding the limitations of maternal health tools

Several factors may influence these results. Participants generally began the study with low symptom scores, with median anxiety and depression scores of 2 and 3, respectively. Additionally, the inclusion of incentives and human intervention limits the ability to generalize these findings to routine, standalone chatbot usage. The authors of the study emphasized that these results should be viewed with caution, noting that whether different clinical populations or alternative AI interventions might produce different outcomes remains an open question.

Why it matters

The study highlights the challenges in determining the efficacy of digital health interventions for maternal populations. For those seeking to integrate technology into clinical care, these findings suggest that AI tools cannot currently be assumed to provide superior mental health benefits compared to traditional, low-cost educational support programs. The trial underscores the need for continued rigorous evaluation of AI in healthcare to distinguish between potential benefits and null results.