Randomised trial of a maternal health chatbot finds no improvement in anxiety or depression at 12 months
400 pregnant and postpartum people from racially and ethnically minoritised groups across 49 US states were randomised to Rosie, a generative AI maternal health chatbot, or to monthly children's books, and after 12 months the chatbot arm showed no improvement in anxiety or depression scores, nor in emergency department visits, postpartum care, contraception use or sleep.
- Model
- Rosie, a generative AI maternal health chatbot answering from an evidence-based knowledge base (underlying model not named)
- Field
- Medicine
- Date
- 2026-11-02
- Human collaborators
- Elizabeth M. Norell, Jordan Boyd-Graber, Quynh C. Nguyen
What was found
Chatbots are widely proposed to fill the gap in support after delivery, when contact with the health system drops away while medical and psychological needs continue, and this trial is one of the few to measure whether one changes anything a year out. Participants were recruited online, enrolled between October 2023 and December 2024 with follow-up completed in December 2025, and randomised 1:1 against an active control receiving monthly children's books. The chatbot was built with community-engaged methods and answered questions from a shared knowledge base rather than tailoring responses by race or ethnicity, and it carried an automated safety alert for messages suggesting distress or urgent symptoms, reviewed by a clinical social worker who contacted the participant. Primary outcomes were the Generalized Anxiety Disorder-7 and Patient Health Questionnaire-9 at 12 months, and both were null: an adjusted mean difference of 0.10 (95% CI -0.77 to 0.96) for anxiety and -0.23 (95% CI -1.05 to 0.59) for depression. The authors conclude that such chatbots may be better positioned as adjunctive tools than as stand-alone interventions.
Novelty check
Europe PMC, 2026-11-09: TITLE:(chatbot OR "conversational agent" OR "generative artificial intelligence" OR "large language model") AND (pregnan* OR postpartum OR maternal) AND (randomized OR randomised) AND (trial) returned 81 records, overwhelmingly protocols, feasibility and acceptability studies, pilots and development papers; the same team's Rosie en Espanol pilot and its community-engaged design paper appear among them, and the only comparable completed randomised outcome trial found was a 2025 breastfeeding chatbot study on a narrower endpoint. The paper's own framing agrees, noting prior maternal chatbot studies emphasised feasibility, acceptability, knowledge or engagement rather than randomised clinical outcomes. ClinicalTrials.gov, 2026-11-09: NCT06053515, completed, 400 participants actual, University of Maryland as lead sponsor, matching the published report.
Caveats and known objections
Kept because this registry records negative results, and the null is the finding. Outcomes were entirely self-reported by participants who knew which arm they were in, with no clinical records consulted, and the authors note they did not measure concurrent mental health treatment or other support that could differ between arms. The developers ran and evaluated their own product, funded by the National Institute on Minority Health and Health Disparities, with the funders recorded as having no role; one author discloses personal fees and grants from industry sources outside this work. Engagement was paid at $0.50 per question up to $45 with a bonus for the most active fifth of users, so usage patterns are unlikely to match unpaid real-world use. 20 of 400 participants were lost to follow-up, unevenly at 13 from the chatbot arm against 7. Two signals run the wrong way and are not dismissed: participants pregnant at enrolment reported more pregnancy-related complications at 12 months in the secondary analysis though not in the longitudinal one, and in exploratory within-group analysis higher chatbot engagement went with higher anxiety scores, which the authors caution against overreading since people with more symptoms may use the tool more. No intervention-related adverse events were reported. The underlying language model is not named anywhere in the paper. Autonomy is ai-led rather than autonomous, the weaker reading: the chatbot answered participants directly with no clinician reviewing its responses, but the automated distress alerts fired 262 times across 97 chatbot participants and each was reviewed by a clinical social worker who then contacted the person, so a substantial share of the arm received human contact inside the intervention.
Trial registration
NCT06053515
That a study was registered, with the outcomes it declared and the date it declared them, so a published endpoint can be checked against the one planned. It does not certify the results, their analysis, or any clinical benefit; registry records are submitted by the sponsor and are not peer reviewed.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Peer reviewed and AI-led.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded peer reviewed for verification and ai-led for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Randomised trial of a maternal health chatbot finds no improvement in anxiety or depression at 12 months. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-11-02-rosie-maternal-chatbot-rct
BibTeX
@misc{whataifound-universityofmaryland-2026-rct,
title = {Randomised trial of a maternal health chatbot finds no improvement in anxiety or depression at 12 months},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by University of Maryland. Verification: Peer reviewed. Autonomy: AI-led.},
url = {https://whataifound.org/finding/2026-11-02-rosie-maternal-chatbot-rct}
}