Can an observer with image information accurately predict visual performance of a na•ve self or peer? In two experiments, participants viewed degraded images of celebrity faces that slowly resolved to full clarity. Participants stopped the resolution process when they were able to identify the celebrity (phase 1). Following phase 1, Experiment-1 participants were given a surprise memory test in which they adjusted the clarity of each face until it looked as it did when recognized in phase 1. Experiment-2 participants viewed phase-1 faces plus a new set of faces in full clarity a priori, and then stopped the resolution process when they thought a naive peer would identify the face. In both experiments, participants showed visual hindsight bias by overestimating identification performance of na•ve observers. These results suggest that once the content of a visual image is known, observers are unable to discount this information. As a result, they overestimate their ability, and the ability of others, to identify the image at a more degraded state.