Limitations of ChatGPT in Chinese-to-English News Translation: A Functional Equivalence Perspective

ChatGPT has emerged as an influential tool for machine translation since recent advancement of artificial intelligence. However, news translation presents difficulties including political sensitivity, formal register, and cultural variability, and these challenge the limits of AI‑generated output. This study uses functional equivalency to assess ChatGPT's ability to translate news from Chinese to English. ChatGPT (GPT-4o) outputs were compared with published official translations of government news reports, and the study conducted a detailed error analysis at four levels, namely lexical, syntactic, discourse, and stylistic. Though it struggles with term consistency, handling complex sentence structures, maintaining coherent flow, and replicating journalistic tone, ChatGPT frequently succeeds in maintaining essential information and generating sentences that are generally fluent. These flaws presented by AI may erode confidence in vital news materials and make it more difficult to replicate the desired reader effect. The results show the usefulness of functional equivalence theory as a model for assessment and extend it to the field of AI-assisted translation. The findings of the study point to areas where human oversight and supervision is crucial for higher standard quality in news and media translation. This study offers implications to educators, translators, and AI developers that aim to enhance translation quality for high-stakes situations.