<![CDATA[The rapid development of generative artificial intelligence has reshaped English language teaching, particularly in writing instruction. This study investigates the integration of ChatGPT in improving applied English writing skills among hospitality students in a vocational higher education context. Using a sequential explanatory mixed-methods design, this study involved 32 students in the quantitative phase and 12 purposively selected students in the qualitative phase. Quantitative data were collected through pre-test and post-test writing assessments, while qualitative data were gathered through semi-structured interviews, classroom observation, and document analysis. Students’ writing was assessed using a hospitality-specific rubric covering tone and politeness, lexical capability, grammatical accuracy, and contextual appropriateness. The findings show a significant improvement in students’ writing performance after ChatGPT-assisted instruction, with the overall mean score increasing from 21.47 to 29.34. The paired samples t-test indicated a statistically significant difference, t (31) = 7.42, p < .001, with a large effect size, Cohen’s d = 1.31. The largest improvement appeared in lexical capability, followed by grammatical accuracy, hospitality tone and politeness, and contextual appropriateness. Qualitative findings reveal that students used ChatGPT for vocabulary expansion, grammar checking, brainstorming, drafting support, and tone consultation. However, the findings also indicate potential risks, including over-reliance, loss of personal voice, and cultural mismatch in hospitality communication. This study concludes that ChatGPT can function as effective pedagogical scaffolding when integrated through structured guidance, reflective practice, and critical evaluation.]]>
