Large carnivores are depicted to shape entire ecosystems through top-down processes. Studies describing these processes are often used to support interventionist wildlife management practices, including carnivore reintroduction or lethal control programs. Unfortunately, there is an increasing tendency to ignore, disregard or devalue fundamental principles of the scientific method when communicating the reliability of current evidence for the ecological roles that large carnivores may play, eroding public confidence in large carnivore science and scientists. Here, we discuss six interrelated issues that currently undermine the reliability of the available literature on the ecological roles of large carnivores: (1) the overall paucity of available data, (2) reliability of carnivore population sampling techniques, (3) general disregard for alternative hypotheses to top-down forcing, (4) lack of applied science studies, (5) frequent use of logical fallacies, and (6) generalisation of results from relatively pristine systems to those substantially altered by humans. We first describe how widespread these issues are, and given this, show, for example, that evidence for the roles of wolves (Canis lupus) and dingoes (Canis lupus dingo) in initiating trophic cascades is not as strong as is often claimed. Managers and policy makers should exercise caution when relying on this literature to inform wildlife management decisions. We emphasise the value of manipulative experiments and discuss the role of scientific knowledge in the decision-making process. We hope that the issues we raise here prompt deeper consideration of actual evidence, leading towards an improvement in both the rigour and communication of large carnivore science.