Generative Artificial Intelligence as a Linguistic Crutch or Cognitive Scaffold: The Interplay of Self‐Beliefs and General English Proficiency in English Medium Instruction

ABSTRACT This study examines how self‐efficacy, self‐regulation, and generative artificial intelligence (GenAI) ‐related factors interact to shape academic success and general English proficiency (GEP) within an EMI context. It draws on data from 754 EMI students at a major public university in Turkey, focusing on the social sciences and engineering disciplines. The findings reveal that for social science students, besides GEP, a positive perception towards GenAI, self‐efficacy, and self‐regulation are key to academic achievement—self‐belief and the ability to regulate one's learning are essential for mastering both language and content. For engineering students, however, success is primarily driven by GEP, GenAI competence, and a positive perception toward GenAI. When it comes to GEP, social science students benefit from GenAI competence to improve their language skills, while engineering students rely more on self‐efficacy and self‐regulation. Additionally, excessive time spent on GenAI platforms is correlated to poorer academic outcomes, highlighting the importance of quality over quantity in GenAI engagement. The findings are discussed, and implications for enhancing EMI strategies are presented.