Abstract
Existing research on teachers’ ability to detect AI-generated texts has predominantly emphasized technical shortcomings, overlooking the behavioral and environmental factors that shape detection accuracy. As generative AI becomes embedded in education, understanding how institutional and personal contexts influence teachers’ detection performance is crucial for ensuring academic integrity. This study aims to identify and analyze the key internal (behavioral) and external (institutional and contextual) factors affecting teachers’ ability to distinguish AI-generated from human-written texts. It further seeks to examine how these factors interact across global regions to develop a more comprehensive framework for understanding detection challenges. An exploratory sequential mixed-method design was employed. The first phase involved 15 key informant interviews with educators from three continents to identify salient determinants of detection capability. Insights from this phase guided the development of a survey administered to 317 teachers across four continents. Data was analyzed using Structural Equation Modeling (SEM) to test interrelationships among identified factors. Findings revealed that rigid university policies significantly hinder teachers’ detection ability, especially in Europe, both directly and indirectly through time limitations and content indistinguishability. By integrating behavioral and contextual dimensions, the study advances beyond technically centered perspectives and proposes a global framework for understanding AI detectability. The results have theoretical and practical implications for policymakers and AI developers. Limitations include reliance on perception-based data and lack of African representation, warranting broader, experimental validation in future research.