Auditing Internal Consistency in Webometrics University Rankings: A Bidirectional, Historical, and Peer-Calibrated Approach

University rankings compress multidimensional institutional performance into a single ordered position, potentially obscuring inconsistencies between final ranks and their underlying indicators. This study develops a bidirectional audit framework to identify unusually favorable and unfavorable outcomes in the July 2026 Webometrics ranking. The analysis covers 5,500 institutions, including 5,491 linked historical profiles and 63,383 historical observations. Rank movements and current positions were evaluated using directional residuals from complementary robust and nonlinear models, followed by historical, institution-type, and predictor-defined peer calibration. Synthetic perturbation experiments assessed detector performance under known inconsistencies. The framework substantially reduced the initial candidate pool after comparability and temporal controls. Gwangju University emerged as the primary novel favorable case, whereas Middle East University Jordan was the primary novel unfavorable case; persistent or peer-sensitive signals were distinguished from new discrepancies. Synthetic validation yielded AUROC values of 0.955 and 0.954 for favorable and unfavorable perturbations, respectively, with false-positive rates below 1% in both directions. These findings demonstrate that ranking anomalies can be screened without interpreting residuals as evidence of institutional quality, manipulation, or provider error. The proposed framework offers a reproducible and conservative approach for prioritizing internally inconsistent ranking outcomes for further investigation.