<p>Non-invasive sampling using faeces can provide valuable ecological information on rare and elusive species, for example histological diet analyses. However, inaccurate field identifications can prejudice our understanding of carnivore ecology. We assessed the accuracy of field identifications for scats collected (2014–2016) across the Greater Limpopo Transfrontier Conservation Area, in southern Africa, that were putatively identified as African lion (<i>Panthera leo</i>) and leopard (<i>P. pardus</i>) in the field. 314 and 413 scats were collected and identified in the field as from lion and leopard. These scats were subjected to genetic analyses, producing genotypes from 134, 213 and 58 scats from lions, leopards and cheetah (<i>Acinonyx jubatus</i>), respectively. 21% (<i>n</i> = 57) and 3.7% (<i>n</i> = 10) of scats identified as leopards in the field were genetically confirmed to be cheetah and lion respectively. 6.7% (<i>n</i> = 9) and 0.8% (<i>n</i> = 1) of scats identified as lions in the field were genetically confirmed to be leopards and cheetah scats, respectively. We then compared diet compositions for lions and leopards derived from scats identified in the field, relative to those identified genetically, using a subsample of the scats collected. Lion diets did not differ significantly between the methods of scat identification. Conversely, leopard diet compositions differed slightly between the methods of scat identification, based on the contribution of different prey size classes to leopard diets and leopard prey preferences. We also provide weak evidence that field identifications of scats can underestimate the degree of diet overlap between lions and leopards. We urge those using non-invasive scat sampling to genetically confirm the identity of the scat donor, to ensure unbiased descriptions of carnivore ecology.</p>