Putting equity first: reorienting artificial intelligence for population-level postpartum depression prevention

Postpartum depression (PPD) affects 10–20% of new mothers globally, yet detection and treatment rates remain low, particularly in underserved populations. Artificial intelligence (AI)—here spanning supervised machine-learning prediction, natural-language and acoustic analysis, and generative conversational agents—is increasingly promoted as a solution, from risk prediction to chatbot-delivered therapy. Drawing on selected evidence from clinical informatics, digital mental health, public-health ethics, and implementation science, we argue that although AI may genuinely expand reach, its current trajectory, driven by technical feasibility rather than public-health value, could widen rather than narrow maternal mental-health inequities. This risk plays out across four domains: (1) algorithmic accuracy that does not translate into population outcomes; (2) training data that structurally exclude marginalized groups; (3) algorithmic care that may erode relational support; and (4) a regulatory vacuum that permits deployment without accountability. Distinguishing published evidence from our own recommendations, we conclude that AI for PPD prevention should be governed as a public good through equity-centered design, real-world evidence standards (including calibration and subgroup performance beyond AUC), tiered device oversight, layered data consent, and clearly specified safety pathways. We call for a global dialogue among policymakers, developers, health systems, researchers, and women with lived experience to operationalize these principles before large-scale deployment.