Prior studies have found metacognitive impairments are linked to a transdiagnostic dimension of anxious-depression, manifesting as reduced confidence in performance (‘metacognitive bias’). However, previous work has been cross-sectional and so it is unclear if under-confidence is a trait-like marker of anxious-depression vulnerability, or if it resolves when anxious-depression improves. Data were collected as part of the ‘Precision in Psychiatry’ study, a large-scale transdiagnostic, four-week observational study of individuals initiating internet-based cognitive behavioural therapy (iCBT) or antidepressant medication. Self-reported clinical questionnaires and perceptual task performance were gathered to assess anxious-depression and metacognitive bias at baseline and four-week follow-up. Primary analyses were conducted for individuals who received iCBT (n=649), with comparisons between smaller samples that received antidepressant medication (n=88) and a control group receiving no intervention (n=82). Prior to receiving treatment, anxious-depression severity was associated with under-confidence in performance in the iCBT arm, replicating previous work. From baseline to follow-up, levels of anxious-depression were significantly reduced, and this was accompanied by a significant increase in metacognitive confidence (B=0.17, SE=0.02, p<0.001). These changes were correlated (r(647)=-0.12, p=0.002); those with the greatest reductions in anxious-depression levels had the largest increase in confidence. In the antidepressant arm, anxious-depression reduced (B=-0.61, SE=0.09, p<0.001) and confidence increased (B=0.31, SE=0.08, p<0.001). Among controls, confidence remained stable from baseline to follow-up (B=0.11, SE=0.07, p=0.103). Metacognitive biases in anxious-depression are state-dependent; when symptoms improve with treatment, so does confidence in performance. Our results suggest this is not specific to the type of intervention.