Test anxiety is common among university students and impairs examination performance. Existing interventions for test anxiety are not particularly effective. Prior to developing an effective intervention, the key psychological beliefs that predict test anxiety need to be identified. Two transdiagnostic models, the intolerance of uncertainty (IU) and Self-Regulatory Executive Function (S-REF) models, propose different beliefs that result in emotional disorder, with beliefs about uncertainty emphasized in the IU model, and metacognitive beliefs emphasized in S-REF model. This study examines if IU and metacognitive beliefs predict test anxiety, and, if the relationship between these beliefs and examination performance is mediated by test anxiety. Undergraduates (n = 134) completed self-report questionnaires at two time points, approximately 3 months apart. At Time 1, during term time, participants completed questionnaires measuring their IU and metacognitive beliefs. At Time 2, participants completed a measure of “state” test anxiety immediately before their examination. IU and metacognitive beliefs were significantly positively correlated with test anxiety, but regression analyses found only the metacognitive belief domain “negative beliefs about the uncontrollability and danger of worry” predicted test anxiety. The relationship between “negative beliefs about the uncontrollability and danger of worry” and examination performance was mediated by the worry dimension of test anxiety. Overall, “negative beliefs about the uncontrollability and danger of worry” appear key to test anxiety. Modification of these metacognitive beliefs in the context of a well-being or study skills program for students could reduce test anxiety and ultimately improve academic performance.