Trophically transmitted parasites often alter their intermediate host's phenotype, thereby predisposing hosts to increased predation. This is generally considered to be a parasite strategy evolved to enhance transmission to the next host. However, the adaptive value of host manipulation is not clear, as it may be associated with costs, such as increased susceptibility to predator species that are unsuitable next hosts for the parasites. Thus, it has been proposed that, to be adaptive, manipulation should be specific by predisposing hosts more strongly to predation by target hosts (next host in the life cycle) than to non-hosts. Here we formally evaluate this prediction, and show that manipulation does not have to be specific to be adaptive. However, when manipulation is nonspecific, it needs to effectively increase the overall predation risk of infected hosts if it is to increase the parasite transmission probability. Thus, when initial predation risk is low, even highly nonspecific manipulation strategies can be adaptive. However, when initial predation risk is high, manipulation needs to be more specific to increase parasite transmission success. Therefore, nonspecific host manipulation may evolve in nature, but the adaptive value of a certain manipulation strategy can vary among different parasite populations depending on the variation in initial predation risk.