Peer-tutoring (PT) is a widely studied learning strategy. No studies, to our knowledge, have examined the effects of PT on tutors' metacognitive abilities, particularly in relation to implicit theories of intelligence. Study 1 aimed to determine the effects of PT on tutors' theory and beliefs of intelligence and self-confidence in their own intelligence. Study 2 aimed to explore the efficacy of a metacognitive training administered to tutors on some metacognitive dimensions for both tutors and tutees. In Study 1, we compared tutors and no-tutors before and after PT; in Study 2 we compared tutors, who completed a brief metacognitive training, with a group of standard tutors, as well as their relative tutees. Participants were assigned to experimental conditions and assessed using various standardised questionnaires. Study 1 revealed a surprising negative effect: after PT, tutors tended to abandon the idea of intelligence as a modifiable construct and adopt a more dysfunctional theory of entity intelligence ("Cohen's d" = 0.66). Study 2 also showed that the PT experience negatively affected tutors' theory of intelligence ("Cohen's d" = 0.94), but this held only for untrained tutors. We also found a positive indirect effect of the training on tutees' attitudes towards maths.