Background Infrared ear thermometry is frequently used in children, because this is a quick method of taking temperature and the ear is easily accessible. Our aim was to evaluate agreement between temperature measured at the rectum and ear in children. Methods We did a systematic review of studies comparing temperature measured at the rectum (the reference site) using mercury, electronic, or indwelling probe thermometers, with temperature measured at the ear (the test site) using infrared ear thermometers. Heterogeneity between studies was investigated by exploring subgroups according to the mode of the infrared ear thermometer. Findings 44 studies containing 58 comparisons (5935 children) were eligible for inclusion in this review. Outcome data were available in reports from 12 comparisons (2312 [39%] children), and data on individual patients were obtained for a further 19 comparisons (2129 [36%] children). 31 comparisons (4441 [75%] children) were therefore included in the meta-analysis. The pooled mean temperature difference (rectal minus ear) was 0·29°C (95% limits of agreement −0·74 to 1·32). We pooled data by ear device mode and the mean temperature differences were rectal mode 0·15°C (−0·95 to 1·25), actual 0·70°C (−0·20 to 1·60), core 0·25°C (−0·78 to 1·27), oral 0·34°C (−0·86 to 1·54), tympanic 0·62°C (−0·40 to 1·64) and mode not stated 0·32°C (−0·57 to 1·21). There was significant residual heterogeneity in both mean differences and sample SDs within the groups of ear device mode. Interpretation Although the mean differences between rectal and ear temperature measurements were small, the wide limits of agreement mean that ear temperature is not a good approximation of rectal temperature, even when the ear thermometer is used in rectal mode. Our finding suggests that infrared ear thermometry does not show sufficient agreement with an established method of temperature measurement to be used in situations where body temperature needs to be measured with precision.