Normothermic machine perfusion (NMP) allows objective assessment of donor liver transplantability. Several viability evaluation protocols have been established, consisting of parameters such as perfusate lactate clearance, pH, transaminase levels, and the production and composition of bile. The aims of this study were to assess 3 such protocols, namely, those introduced by the teams from Birmingham (BP), Cambridge (CP), and Groningen (GP), using a cohort of high‐risk marginal livers that had initially been deemed unsuitable for transplantation and to introduce the concept of the viability assessment sensitivity and specificity. To demonstrate and quantify the diagnostic accuracy of these protocols, we used a composite outcome of organ use and 24‐month graft survival as a surrogate endpoint. The effects of assessment modifications, including the removal of the most stringent components of the protocols, were also assessed. Of the 31 organs, 22 were transplanted after a period of NMP, of which 18 achieved the outcome of 24‐month graft survival. The BP yielded 94% sensitivity and 50% specificity when predicting this outcome. The GP and CP both seemed overly conservative, with 1 and 0 organs, respectively, meeting these protocols. Modification of the GP and CP to exclude their most stringent components increased this to 11 and 8 organs, respectively, and resulted in moderate sensitivity (56% and 44%) but high specificity (92% and 100%, respectively) with respect to the composite outcome. This study shows that the normothermic assessment protocols can be useful in identifying potentially viable organs but that the balance of risk of underuse and overuse varies by protocol.