Chatbot Tutor vs. Human Tutor: Assessing Learning Effectivenes in AI- and Human-Led EFL Learning

Abstract

This study investigates the comparative effectiveness of chatbot-led and human-led instruction in teaching English grammar to undergraduate EFL learners in Indonesia. Using a quasi-experimental design, the study involved 30 students from the English Department of Universitas Serambi Mekkah who were divided equally into a chatbot group (CG) (n=15) and a human tutor group (HTG) (n=15). Both groups received instruction on the simple past tense with a focus on irregular verbs during a single structured learning session. Grammar development was measured through pre-tests and post-tests, supported by semi-structured interviews to explore their perceptions. The findings show that both instructional modes resulted in measurable improvement in grammatical accuracy. The human tutor group achieved slightly higher gains in sentence construction, indicating deeper conceptual understanding, while the chatbot group reported higher engagement, reduced anxiety, and greater learner autonomy. These results suggest that chatbot tutors are effective for form-focused practice and affective support, whereas human tutors remain essential for nuanced explanation and adaptive teaching. Regarding the limitations, the chatbot group noted chatbot’s ineffectiveness in providing contextual explanations while the human tutor group considered time as the major constraint in accessing continuous and individualized support. The study concludes that a blended approach integrating chatbot tools with human instruction offers the most pedagogically effective model for EFL grammar learning.