Metacognition about face recognition has been much discussed in the psychological literature. In particular, the use of self-report to identify people with prosopagnosia (“face blindness”) has contentiously been debated. However, no study to date has specifically assessed metacognition at the top end of the spectrum. If people with exceptionally proficient face recognition skills (“super-recognizers,” SRs) have greater insight into their abilities, self-report instruments may offer an efficient means of reducing candidate lists in SR screening programs. Here, we developed a “super-recognizer questionnaire” (SRQ), calibrated using a top-end civilian sample (Experiment 1). We examined its effectiveness in identifying SRs in pools of police (Experiment 2) and civilian (Experiment 3) participants, using objective face memory and matching tests. Moderate effect sizes in both samples suggest limited insight into face memory and target-present face matching ability, whereas the only predictor of target-absent matching performance across all samples was the number of years that an officer had been in the police force. Because the SRQ and single-item ratings showed little sensitivity in discriminating SRs from typical perceivers in police officers and civilians, we recommend against the use of self-report instruments in SR screening programs.