The Over-Reliance Dilemma: Evaluating the Impact Of 100% Generative AI Utilization on Higher Education Student Assignment Quality and Critical Thinking

Abstract

The rapid integration of Large Language Models (LLMs) and Generative Artificial Intelligence (GenAI) in higher education has fundamentally altered student learning behaviors and academic workflow dynamics. While AI tools offer unprecedented advantages in information synthesis and writing efficiency, a growing subset of tertiary students exhibits near-total dependence—approaching 100% utilization—on AI tools to generate coursework submissions. This mixed-methods study investigates the motivations, output quality, and cognitive ramifications associated with near-total AI dependency in academic assignments among undergraduate students (N = 180). Data were gathered through structured questionnaires, qualitative student focus groups, and an expert-evaluated comparative analysis of fully AI-generated coursework versus human-guided assignments. The findings reveal that while total AI reliance yields structural fluency and high linguistic correctness, it significantly dampens higher-order critical thinking, original argumentative depth, and long-term skill acquisition. Furthermore, driving factors include severe time poverty, assessment anxiety, and unclear institutional guidelines regarding ethical AI integration. This paper offers practical recommendations for educators to redesign assessment models, foster authentic cognitive engagement, and establish balanced digital literacy frameworks.