From benchmarking to human–AI collaboration: a systematic review of ChatGPT in translation studies (2022–2025)
TL;DR
The literature demonstrates a transition from early benchmarking of ChatGPT against human translators and conventional translation technologies toward a more interaction-oriented understanding of human-AI collaboration, suggesting that current translation research is increasingly concerned with whether ChatGPT can translate effectively, but with how it can be integrated responsibly and productively into translation practice, translator education, and human decision-making.
Abstract
The rapid development of large language models, particularly ChatGPT, has generated growing interest in their potential applications in translation studies. Despite the increasing number of publications on this topic, a comprehensive overview of how ChatGPT is being investigated and applied in translation research remains limited. This study presents a systematic review of research on the use of ChatGPT in translation studies. Using the Scopus database, 1,316 publications published between 2022 and 2025 were initially identified. After the application of explicit inclusion and exclusion criteria, 140 studies were retained for the final synthesis. The selected studies were analyzed to identify publication trends, language pairs, domains of application, and changes in research focus over time. The findings indicate a marked increase in research output across the examined period, peaking in 2025 with 83 publications. English-Arabic emerged as the most frequently investigated language pair in studies involving source-to-target translation tasks. Thematic analysis identified several major areas of ChatGPT use, including translation pedagogy and training, medical and healthcare translation, literary and creative translation, legal and technical translation, audiovisual translation and subtitling, post-editing practices, and translation quality evaluation. The analysis also revealed a chronological shift in research priorities: studies published in 2023 largely examined user acceptance and preliminary comparisons, research in 2024 increasingly focused on performance testing in specialized domains, and studies published in 2025 placed greater emphasis on effective human-AI interaction and the integration of ChatGPT into translator training. Overall, the literature demonstrates a transition from early benchmarking of ChatGPT against human translators and conventional translation technologies toward a more interaction-oriented understanding of human-AI collaboration. This shift suggests that current translation research is increasingly concerned not simply with whether ChatGPT can translate effectively, but with how it can be integrated responsibly and productively into translation practice, translator education, and human decision-making.