Navigating the "Dependency Trap": A Systematic Literature Review on Generative AI in EFL Learning and Its Contextual Barriers

Abstract

The rapid integration of Generative Artificial Intelligence (GenAI) is changing English as a Foreign Language (EFL) education. Students often use GenAI as a backup resource or private tutor because of challenges such as large class sizes and delayed teacher feedback. However, this creates a major risk: the technology can easily shift from a helpful learning support into a tool that does the actual thinking and writing for the student. This systematic literature review synthesizes research from 2023 to 2026 to examine how classroom barriers and student habits combine to create a “dependency trap”. It is a self-reinforcing cycle where over-reliance on AI hinders independent language development. By following PRISMA guidelines, a systematic search was conducted across Scopus and Web of Science. The screening focused on peer-reviewed studies investigating the structural, cultural, and pedagogical challenges of using GenAI in global EFL contexts. The synthesis shows that a lack of clear university policies, unequal access to technology, and teacher stress all accelerate student dependency on AI. Students often trust fluent AI outputs blindly because the text sounds academic, even when they lack the skills to spot errors. This leads to two major risks: cognitive offloading, which creates fluency without comprehension, and a loss of unique student voice, where writing becomes repetitive and formulaic. The study concludes that simply allowing students to use AI without guidance is insufficient. Instead, educational institutions must establish clear usage policies and formal AI literacy training. This will help students critically evaluate AI tools and protect their own learning autonomy.