Development of an Equity-Centered Sociotechnical Architecture for Generative AI Integration in Public Health Promotion: Conceptual Framework.

Unlabelled: This article is a viewpoint: it presents the authors' perspective, informed by a critical synthesis of the current literature at the intersection of generative AI (GenAI) technologies, public health communication, and digital ethics, rather than original empirical data or analyses. The emergence of GenAI, including large language models (LLMs), represents a profound paradigm shift in digital health communication. By moving beyond traditional information retrieval to dynamic, human-like knowledge generation, GenAI offers unprecedented opportunities for public health promotion. However, the unguided integration of these powerful commercial models into health care systems poses profound sociotechnical risks. In this viewpoint, we aim to communicate three key messages to public health researchers, practitioners, policymakers, and AI developers: (1) GenAI offers transformative applications for public health promotion, spanning personalized health education, stigma mitigation, and accelerated epidemiological surveillance; (2) the unguided integration of commercial generative models simultaneously generates intersecting sociotechnical risks and ethical challenges, encompassing a widening "AI digital divide," algorithmic bias and epistemic opacity, and the erosion of data privacy and governance; and (3) an equity-centered sociotechnical architecture, built on 4 strategic pillars, is required to govern this transition safely. We conducted a critical synthesis of the current literature and theoretical frameworks at the intersection of GenAI technologies, public health communication, and digital ethics, systematically mapping both the translational capabilities and the sociotechnical vulnerabilities of generative models. GenAI demonstrates transformative potential across 3 primary domains: democratizing health education by translating complex medical jargon, mitigating societal stigma through nonjudgmental conversational interfaces, and accelerating epidemiological surveillance via rapid thematic synthesis. However, these benefits are counterbalanced by a matrix of sociotechnical risks. Specifically, unguided GenAI deployment threatens to exacerbate a novel "AI digital divide" driven by economic exclusion, prompt literacy demands, and linguistic biases; compromise clinical safety through deep-seated algorithmic biases and epistemic opacity; and erode patient privacy through profound vulnerabilities in cybersecurity and corporate data governance. The advent of GenAI marked an irreversible paradigm shift with the unprecedented capacity to democratize health literacy, dismantle stigma, and accelerate disease surveillance. However, treating GenAI as a technological panacea is a perilous oversight. Without intentional, equity-focused interventions, these technologies invariably scale and automate the structural inequalities they have the potential to solve. Ultimately, the future of digital health promotion depends not only on the computational power of these models but also on the ethical, regulatory, and inclusive sociotechnical architectures we design to govern them.