Objective Perceptions of strangers’ self‐esteem can have wide‐ranging interpersonal consequences. Aiming to reconcile inconsistent results from previous research that had predominantly suggested that self‐esteem is a trait that can hardly be accurately judged at zero acquaintance, we examined unaquainted others’ accuracy in inferring individuals’ actual self‐esteem. Method Ninety‐nine target participants (77 female; Mage = 23.5 years) were videotaped in a self‐introductory situation, and self‐esteem self‐reports and reports by well‐known informants were obtained as separate accuracy criteria. Forty unacquainted observers judged targets' self‐esteem on the basis of these short video sequences (M = 23s, SD = 7.7). Results Results showed that both self‐reported (r = .31, p = .002) and informant‐reported self‐esteem (r = .21, p = .040) of targets could be inferred by strangers. The degree of accuracy in self‐esteem judgments could be explained with lens model analyses: Self‐ and informant‐reported self‐esteem predicted nonverbal and vocal friendliness, both of which predicted self‐esteem judgments by observers. In addition, observers’ accuracy in inferring informant‐reported self‐esteem was mediated by the utilization of targets’ physical attractiveness. Besides using valid behavioral information to infer strangers’ self‐esteem, observers inappropriately relied on invalid behavioral information reflecting nonverbal, vocal, and verbal self‐assuredness. Conclusions Our findings show that strangers can quite accurately detect individuals’ self‐reported and informant‐reported self‐esteem when targets are observed in a public self‐presentational situation.