The study systematically reviewed 17 studies with 9,949 patients and 37 predictive models for acute kidney injury (AKI) in acute pancreatitis (AP). Of the 32 models that passed internal validation, the majority had an area under the curve (AUC) above 0.7. The meta-analysis showed a pooled AUC for internal validation of 20 models of 0.790 (95% CI = 0.761–0.818) and for external validation of 5 models of 0.766 (95% CI = 0.684–0.845). In all 17 studies, the risk of bias was high, but the overall applicability of the models was satisfactory. Predictive models showed moderate predictive power. However, clinical applicability requires further investigation due to high risk of bias and lack of external validation. The study is registered under the identifier CRD420251008769.