How do we learn about who is good at what? Others’ competence is unobservable and often must be inferred from observable evidence, such as failures and successes. However, even the same performance can indicate different levels of competence depending on the context, and objective evaluation metrics are not always available. Building on recent advances on children's use of emotion as information, here we ask whether expressions of surprise inform inferences about competence. Participants saw scenarios (sports, academics) where two students achieved identical outcomes but a teacher showed surprise to one student and no surprise to the other. In Exp.1, adults inferred that the successful student who elicited the teacher’s surprise was less competent than the other student, but this pattern reversed when both students failed. Exp.2 (4-9-year-olds) finds initial evidence for such inferences in school-aged children. These findings have implications for promoting healthy social comparisons and preventing acquisition of negative stereotypes from non-verbal cues.