Background: Metacognition has been described as the knowledge of our own cognitive processes. Metacognitive deficits are common in schizophrenia, but little is known about metacognition before the onset of full-blown psychosis. Aims: This study aimed to longitudinally characterize metacognition in a sample of individuals at clinical high risk (CHR) for psychosis, and to determine if metacognition was related to later conversion to psychosis. Method: Participants (153 CHR individuals; 68 help seeking controls, HSC) were part of the large multi-site PREDICT study, which sought to determine predictors of conversion to psychosis. They were tested at baseline and 6 months using the Meta-Cognitions Questionnaire (MCQ) that has five sub-scales assessing different domains of metacognition. Results: Results of the mixed-effect models demonstrated significantly poorer scores at baseline for the CHR group compared to the HSC group in Negative beliefs about uncontrollability, Negative beliefs and the overall MCQ score. At the 6-month assessment, no difference was observed in metacognition between the two groups, but both groups showed improvement in metacognition over time. Those who later converted to psychosis had poorer performance on metacognitive beliefs at baseline. Conclusions: A poorer performance in metacognition can be seen as a marker of developing a full blown psychotic illness and confirms the potential value of assessing metacognitive beliefs in individuals vulnerable for psychosis.