Human Agency in GenAI-ssisted Translation: Lessons from Translating Chinese Policy Reports

The rapid development of Generative AI and large language models (LLMs) has significantly transformed the landscape of professional translation, especially for large-volume, high-stakes texts such as policy reports and research papers. While LLMs such as ChatGPT can produce fluent and contextually plausible drafts, their output often requires careful post-editing to ensure accuracy, appropriateness, and alignment with genre conventions. This paper explores the translator's agency in AI-assisted translation through a detailed examination of case studies drawn from a corpus of Chinese policy reports on sustainable agricultural supply chains. Each case demonstrates how the human translator identifies issues, evaluates alternatives, and makes contextually informed revisions, resulting in translations that are not only accurate but also rhetorically effective and suitable for policy discourse. The paper argues that human agency is not diminished but rather reconfigured in AI post-editing. Translators act as diagnosticians and evaluators, working with AI as co-creators to shape the text while ultimately serving as the final gatekeepers of meaning and quality. The study concludes by stressing the qualities translators must cultivate—critical reading, domain knowledge, genre awareness, stylistic sensibility, and editorial judgment—to fully harness AI as a productive partner while maintaining ultimate accountability for the final product.