Embodied models of social cognition argue that others’ emotional states are processed by re-enacting a representation of the same state in the observer, along with associated somatic and physiological responses. In this framework, previous studies tested whether a strong sensitivity to interoceptive signals (i.e., inputs arising from within one’s body) facilitates the understanding of others’ affect, leading to mixed results. Such heterogeneity in the literature could reflect methodological differences in paradigms employed, with some probing classification of a precise condition, and others requiring the assessment of supra-ordinal dimensions orthogonal to many states. Here, we engaged fifty young women in a study where they evaluated others’ naturalistic facial reactions to painful and disgusting stimuli of comparable unpleasantness. Separately, we measured their interoceptive abilities through a well-known heartbeat counting task. We found that individuals that were more accurate in tracking their heartbeats across time were also more prone to judge facial expressions as more unpleasant (supra-ordinal assessment). However, when specifically asked to discriminate between comparably-unpleasant pain and disgust (state-specific assessment), participants’ performance was not influenced by their interoceptive abilities. Although confined to a female sample, this study extends our knowledge on the role of interoception in the understanding of others, which influences only the evaluation of general features such as unpleasantness (common between pain and disgust), without extending to the appraisal of a precise state. This finding supports multi-componential models of social cognition, suggesting that only part of our ability to assess others’ affect is mediated by a representation of one’s affective/somatic responses.