Physician empathy is considered essential for good clinical care. Empirical evidence shows that it correlates with better patient satisfaction, compliance, and clinical outcomes. These data have nevertheless been criticized because of a lack of consistency and reliability. In this paper, we claim that these issues partly stem from the widespread idealization of empathy: we mistakenly assume that physician empathy always contributes to good care. This has prevented us from agreeing on a definition of empathy, from understanding the effects of its different components and from exploring its limits. This is problematic because physicians’ ignorance of the risks of empathy and of strategies to manage them can impact their work and wellbeing negatively. To address this problem, we explore the effects of the potential components of empathy and argue that it should be conceived as a purely descriptive and wide term. We end by discussing implications for medical education.