Empathy is a cornerstone of human sociality. It has important consequences for our interpersonal relationships and for navigating our social world more generally. Although research has identified numerous psychological factors that can influence empathy, evidence suggests that empathy may also be rooted in our biology and, in particular, the gonadal steroid hormone testosterone. To date, much of the research linking testosterone and empathy has focused on the 2D:4D ratio (i.e., the ratio of the lengths of the index and ring fingers), and the results have been mixed. These mixed results, however, may be due to reliance on self-report measures to assess empathy, which can be vulnerable to self-presentation, as well as social-cultural norms about gender/sex differences in empathy. Moreover, although some have argued that digit ratio is an indicator of prenatal androgen exposure, the evidence for this to date is weak. Here, we aimed to follow up on this prior work, using a naturalistic empathic accuracy task in which participants dynamically track, in real-time, the emotional state of targets. We show that lower digit ratio (Study 1; N = 107) and higher circulating testosterone (Study 2; N = 76) are associated with poorer empathic accuracy performance; critically, these effects hold when controlling for sex/gender. In neither study, however, did we find effects on self-reported empathy. Our results highlight the limitations of self-report measures and support the notion that endogenous testosterone levels as well as 2D:4D ratio are related to key social-cognitive competencies like empathic accuracy.