Cognitive biases have been implicated in the etiology and maintenance of depression and anxiety, but their utility in predicting future symptoms remains debated. This meta-analysis aimed to estimate the overall effect size of their predictive effects and to identify moderators relevant to theory and methodology. The study protocol was pre-registered on PROSPERO (record number: CRD42021232236). Searches of PsycINFO, Web of Science, PubMed, PsyArXiv Preprints, and ProQuest Dissertations yielded 81 studies with 621 contrasts and 17,709 participants through December 2024. The methodological quality of the included studies was evaluated using the Quality In Prognosis Studies (QUIPS) tool. Results from a three-level meta-analysis revealed a small overall effect size (β = 0.04, 95 %-CI [0.02, 0.06], p <
.002) and significant between- and within-study variance after removal of outliers. Equivalent effect sizes were found for the predictive utility of cognitive biases in children/adolescents and adults, for increased negative bias and decreased positive bias, and for anxiety and depression outcomes. The magnitude of the overall effect was moderated by the cognitive process, with significant effect sizes for interpretation bias and memory bias but not for attention bias. These findings support the predictive role of cognitive biases in anxiety and depression, with interpretation and memory biases emerging as key markers. These findings have implications for cognitive theories of depression and anxiety and for clinical interventions.