Physicians engage in risk stratification as a normative part of their professional duties. Risk stratification has the potential to be beneficial in many ways, and implicit recognition of this potential benefit underlies its acceptance as a cornerstone of the medical profession. However, risk stratification also has the potential to be harmful. We argue that ‘profiling’ is a term that corresponds to risk stratification strategies in which there is concern that ethical harms exceed likely or proven benefits. In the case of risk stratification for health goals, this would occur most frequently if benefits were obtained by threats to justice, autonomy or privacy. We discuss implications of the potential overlap between risk stratification and profiling for researchers and for clinicians, and we consider whether there are salient characteristics that make a particular risk stratification algorithm more or less likely to overlap with profiling, such as whether the risk stratification algorithm is based on voluntary versus non-voluntary characteristics, based on causal versus non-causal characteristics, or based on signifiers of historical disadvantage. We also discuss the ethical challenges created when a risk stratification scheme helps all subgroups but some more than others, or when risk stratification harms some subgroups but benefits the aggregate group.