AI chatbot alone ‘may be insufficient’ for maternal mental health: 6 study notes

Advertisement

Pregnant and postpartum individuals who used a generative AI maternal health chatbot the most reported higher anxiety scores. Overall, the tool did not improve anxiety or depression, according to a University of Maryland-led randomized clinical trial published Nov. 4 in JAMA Network Open.

Elizabeth Norell, PhD, of the University of Maryland School of Public Health in College Park, and Quynh Nguyen, PhD, of the National Institute of Nursing Research in Bethesda, Md., led the trial. It enrolled 400 participants from 49 states and Washington, D.C., between October 2023 and December 2024, with follow-up completed in December 2025. Participants identified as members of racial or ethnic minority groups and were pregnant or parenting an infant younger than 6 months. Half got access to Rosie, an AI chatbot offering maternal and infant health guidance. The other half joined a monthly children’s book club as the active control.

The researchers said a generative AI maternal health chatbot alone “may be insufficient to improve maternal mental health” and that such chatbots may work better as complementary tools than as standalone interventions.

Here are six things to know:

1. At 12 months, the chatbot group’s anxiety and depression scores showed no significant difference from the control group’s.

2. The chatbot also had no significant effect on emergency department visits, postpartum checkups, contraception use or sleep.

3. Engagement was high: 98.4% of chatbot users asked at least one question, and the median user asked 119. Users earned 50 cents per question, up to $45.

4. Heavier users reported higher anxiety but about half an hour more sleep per night. The researchers said anxious participants may have used the tool more, or repeated symptom-focused questions may have heightened worry.

5. Among participants pregnant at enrollment, chatbot users reported more pregnancy-related complications in one adjusted analysis. The finding did not hold in longitudinal analyses, and the researchers cautioned against drawing conclusions from it.

6. Automated safety alerts triggered 262 times among 97 users. Each alert prompted review and outreach from study staff.

At the Becker's Fall Behavioral Health Summit, taking place November 4–5 in Chicago, behavioral health leaders and executives will explore strategies for expanding access to care, integrating services, addressing workforce challenges and leveraging innovation to improve outcomes across the behavioral health continuum. Apply for complimentary registration now.

Advertisement

Next Up in Research & Analysis

Advertisement