The success of genome-wide association studies (GWAS) completed in the last 15 years has reinforced a key fact: polygenic architecture makes a substantial contribution to variation of susceptibility to complex disease, including Alzheimer’s disease. One straight-forward way to capture this architecture and predict which individuals in a population are most at risk is to calculate a polygenic risk score (PRS). This score aggregates the risk conferred across multiple genetic variants, ultimately representing an individual’s predicted genetic susceptibility for a disease. PRS have received increasing attention after having been successfully used in complex traits. This has brought with it renewed attention on new methods which improve the accuracy of risk prediction. While these applications are initially informative, their utility is far from equitable: the majority of PRS models use samples heavily if not entirely of individuals of European descent. This basic approach opens concerns of health equity if applied inaccurately to other population groups, or health disparity if we fail to use them at all. In this review we will examine the methods of calculating PRS and some of their previous uses in disease prediction. We also advocate for, with supporting scientific evidence, inclusion of data from diverse populations in these existing and future studies of population risk via PRS.