<p>This study examines the behavioral and contextual determinants of fraud detection among external auditors in the Middle Eastern audit context. Drawing on the Theory of Planned Behavior and Social Cognitive Theory, it develops and tests a conditional mediation–moderation model linking transparency orientation, locus of control, and learning agility to fraud detection through audit skepticism, while accounting for the moderating role of audit environment complexity. Data were collected from 250 professional auditors and analyzed using Structural Equation Modeling in SmartPLS. The findings indicate that all three auditor attributes positively influence fraud detection, via enhanced audit skepticism, with learning agility emerging as the strongest predictor of skepticism. Audit environment complexity, however, exhibits a dual role: it weakens the relationships between auditor attributes and skepticism while moderating their direct effects on fraud detection. Moreover, higher complexity reduces the effectiveness of skepticism in detecting fraud, highlighting the constraints imposed by cognitively demanding audit environments. This study contributes to the literature by demonstrating how individual auditor characteristics interact with contextual complexity to shape fraud detection effectiveness, particularly in institutionally fragmented regulatory settings. The findings offer practical implications for auditor selection, training, and the design of organizational support mechanisms aimed at strengthening fraud detection capabilities.</p>