Despite the importance of observers to collect data for effective fisheries management worldwide, their species-identification abilities are rarely assessed. Misidentifications could compromise observer data particularly in diverse, multi-species fisheries such as those in the tropics where visual identification is challenging. Here, we provide the first estimates of the ability of scientific observers to identify five species of morphologically similar carcharhinid sharks (Carcharhinus leucas, C. amboinensis, C. tilstoni, C. sorrah and C. brevipinna) in a fishery in northern Australia. We compared observer field identifications of sharks with genetic validation (814. bp mtDNA NADH dehydrogenase subunit 4) to quantify species identification errors. We used binomial generalised linear models to determine the influences of species, gender, total length, and the observer's experience on identification error. We found that identification error (?20%) depended predominately on the species in question (highest error for C. tilstoni). Male sharks were misidentified less frequently than females, and error decreased marginally with increasing total length. Surprisingly, we found no statistical evidence that observer experience influenced identification error. Our results provide the first benchmark of identification accuracy of observers for carcharhinid sharks in northern Australia and show that estimates of error in species identifications need to be incorporated into management strategies to ensure successful recovery of the many recently over-fished shark populations.