This study examines the use of generative artificial intelligence, i.e., ChatGPT, in English as a foreign language (EFL) learning, emphasizing the mediating role of entangled cognition and the effects of the learning outcomes of the tourism students. The research was designed to a quasi-experiment which included 96 participants (48 in an experimental group and 48 in a control group) who were sampled based on convenience to the Spring 2024 semester in one university in southern Taiwan. The “custom virtual language course” experimental group used ChatGPT for personalized language practice and culture learning, control group received traditional learning. A questionnaire package, including the cognitive technology use questionnaire (CTUQ), extended mind scale (EMS), distributed cognition questionnaire (DCQ), metacognitive awareness inventory (MAI), and TOEIC pre- and post-tests was administered to collect the data. The difference-in-differences design was adopted and observed a significant treatment effect such that the treatment group had an average increase in mean scores of 37.98 (standard deviation [SD] = 7.80) compared to 19.62 (SD = 7.80) for the control group and, therefore, an average treatment effect of 21.38 (95% confidence interval [18.74, 24. 01]). Findings suggest that ChatGPT promotes cognitive offloading, distributed cognition, and metacognitive awareness (CTUQ mean [M] = 3.701, EMS M = 3.421, DCQ M = 3.721, MAI M = 3.551), and the development of collaborative learning and cultural competence. These results reveal ChatGPT’s potential to reform EFL education, but they also indicate the necessity to mitigate the risks associated with ethical quandaries and over-dependence. Future studies need to create specific scales that can be used for entangled cognition and examine the long-term effects on cognition.