Questions of how we know our own and other minds, and whether metacognition and mindreading rely on the same processes, are longstanding in psychology and philosophy. In Experiment 1, children/adolescents with autism (who tend to show attenuated mindreading) showed significantly lower accuracy on an explicit metacognition task than neurotypical children/adolescents, but not on an allegedly metacognitive implicit one. In Experiment 2, neurotypical adults completed these tasks in a single-task condition or a dual-task condition that required concurrent completion of a secondary task that tapped mindreading. Metacognitive accuracy was significantly diminished by the dual-mindreading-task on the explicit task but not the implicit task. In Experiment 3, we included additional dual-tasks to rule out the possibility that any secondary task (regardless of whether it required mindreading) would diminish metacognitive accuracy. Finally, in both Experiments 1 and 2, metacognitive accuracy on the explicit task, but not the implicit task, was associated significantly with performance on a measure of mindreading ability. These results suggest that explicit metacognitive tasks (used frequently to measure metacognition in humans) share metarepresentational processing resources with mindreading, whereas implicit tasks (which are claimed by some comparative psychologists to measure metacognition in nonhuman animals) do not.