Summary Capnography (ETCO2) is routinely used as a non‐invasive estimate of arterial carbon dioxide (PaCO2) levels in order to modify ventilatory settings, whereby it is assumed that there is a positive gap between PaCO2 and ETCO2 of approximately 0.5 kPa. However, negative values (ETCO2 > PaCO2) can be observed. We retrospectively analysed arterial to end‐tidal carbon dioxide differences in 799 children undergoing general anaesthesia with mechanical ventilation of the lungs in order to elucidate predictors for a negative gap. A total of 2452 blood gas analysis readings with complete vital sign monitoring, anaesthesia gas analysis and spirometry data were analysed. Mean arterial to end‐tidal carbon dioxide difference was −0.18 kPa (limits of 95% agreement −1.10 to 0.74) and 71.2% of samples demonstrated negative values. The intercept model revealed PaCO2 to be the strongest predictor for a negative PaCO2‐ETCO2 difference. A decrease in PaCO2 by 1 kPa resulted in a decrease in the PaCO2‐ETCO2 difference by 0.23 kPa. This study demonstrates that ETCO2 monitoring in children whose lungs are mechanically ventilated may paradoxically lead to overestimation of ETCO2 (ETCO2 > PaCO2) with a subsequent risk of unrecognised hypocarbia.