<![CDATA[The rapid advancement of artificial intelligence in translation has created an urgent need to examine its effectiveness in producing high-quality academic texts, particularly in relation to linguistic accuracy, naturalness, and terminological consistency. This study investigates the relationship among these three dimensions and their contribution to the quality of academic translation, while comparing the performance of ChatGPT and Google Translate. A quantitative quasi-experimental design with a non-equivalent control group was employed, involving two classes (experimental = ChatGPT; control = Google Translate; n = 30 each). Data were collected using a Likert-scale rubric covering four indicators: accuracy, naturalness, terminology consistency, and overall quality. The instrument was validated through Rasch analysis, and data met the assumptions for parametric testing. The findings indicate that the experimental group demonstrated consistent improvement across all indicators, with notable gains in naturalness and terminology consistency. Statistical analysis revealed significant differences between the two groups in key aspects of translation quality, favoring the use of ChatGPT. In addition, effectiveness analysis showed that the improvement achieved by the experimental group was higher than that of the control group. Structural modeling further suggests that naturalness and terminology consistency play a substantial role in shaping overall translation quality. Overall, the results highlight that high-quality academic translation is achieved through the integration of accurate meaning, fluent expression, and consistent terminology. ChatGPT demonstrates a comparative advantage in supporting these dimensions, suggesting its potential as a complementary tool in academic translation practices and higher education contexts.]]>
Copyrights © 2026
