<![CDATA[Poetry translation presents a unique challenge, as it requires maintaining a delicate balance between semantic fidelity and the preservation of the aesthetic qualities of the target text. This study aims to evaluate the quality of poetry translations produced by university students both with and without the assistance of artificial intelligence (AI), using four assessment criteria based on Nababan's (2012) framework: accuracy, acceptability, readability, and poeticness. The research employed a descriptive quantitative approach with an evaluation instrument using a three-point scale, where a score of 3 indicates high quality, 2 denotes moderate quality, and 1 reflects low quality. Data were obtained from poetry translations evaluated by experienced raters using a standardized rubric, with results analyzed in terms of frequency distribution and percentages. Findings reveal that in AI-assisted translations, the highest score attainment was observed in readability (52.76%), followed by acceptability (51.14%), accuracy (50.39%), and poeticness (24.34%) as the lowest. In contrast, in translations without AI, the highest score attainment was also found in readability (34.76%), followed by accuracy (33.06%), acceptability (32.43%), and poeticness (12.33%). This comparison indicates that AI use consistently enhances accuracy, acceptability, and readability scores, yet shows limited capacity to improve poeticness significantly. This aspect requires creativity, stylistic sensitivity, and nuanced linguistic judgment beyond the current capabilities of AI systems.]]>
