VietTESOL International Convention 2026

In technical universities, general English (GE) curricula often suffer from low student motivation because non-English majors perceive standard topics as disconnected from their professional identities. While literature has extensively explored Artificial Intelligence (AI) in English for Specific Purposes (ESP), limited research addresses how Generative AI (GenAI) can transform basic GE tasks into career-oriented practices for low-proficiency learners, particularly after a prolonged instructional time lag. This quasi-experimental study investigated the integration of GenAI as a linguistic scaffold within a Task-Based Language Teaching (TBLT) framework at Hanoi University of Civil Engineering.
The participants comprised 46 pre-intermediate engineering majors divided into an Experimental Group (EG, n=21) and a Control Group (CG, n=25). To evaluate delayed language retention, the pedagogical intervention was purposefully conducted seven weeks after the formal instruction of a "City Description" writing task (Unit 5, English File). The CG utilized traditional methods, while the EG employed a structured three-step prompting framework to leverage GenAI (ChatGPT/Gemini) as a dynamic scaffold. Data from pre- and post-tests were evaluated using a double-blind grading procedure with a standardized analytic rubric. Independent samples t-tests revealed that while both groups shared homogenous baseline proficiency, the EG exhibited statistically significant improvements in final text quality ($p < 0.05$). More importantly, the AI-assisted drafts demonstrated a marked shift toward professional urban infrastructure terminology, suggesting that GenAI effectively bridges the gap between general language structures and professional identity during delayed production tasks.