AI Is Becoming a Determinant of Health: Do We Need an “Epidemiology of AI”?: Nature Health | September 2026

What Does This Article Add?

As AI increasingly influences how people obtain health information and how healthcare systems make decisions, the authors argue that its effects should be studied at the population level—not only as individual AI interventions.

• The paper proposes treating AI exposure as a determinant of health.
• “Personal AI exposure” refers to deliberate use of tools such as AI assistants for health information or decision support.
• “Ambient AI exposure” occurs without active individual choice—for example, algorithmic curation or AI-mediated institutional decisions.
• The framework borrows concepts from environmental epidemiology to examine cumulative and chronic AI exposure.
• AI may act as an exposure, mediator, modifier or confounder within causal models of health.
• Conventional short-term experiments may miss long-term, population-scale effects of repeated AI exposure.
• Unequal access, performance and exposure could potentially influence health disparities and equity.

Why Does It Matter?

Healthcare research usually asks whether an individual AI tool works. This paper asks a broader question: what happens when entire populations increasingly live within AI-mediated information and healthcare environments?

GastroAGI Takeaway

The next phase of medical AI research may need to move from “Does this AI model perform well?” to “What does sustained AI exposure do to population health?” Measuring who is exposed, how often, through which systems and with what downstream consequences may become essential for evaluating safety, effectiveness and health equity.