Investigating the Translation Dynamics of Arabic-English Code-Switching by AI and EFL University Students in Saudi Arabia

In Saudi Arabia’s increasingly bilingual academic environment, Arabic-English code-switching has become a common phenomenon, especially among university students. This study investigates the translation dynamics of such code-switched texts, comparing the performance of Artificial Intelligence (AI), represented by ChatGPT 4, and English as a Foreign Language (EFL) university students at Imam Mohammed Ibn Saud Islamic University. Despite technological advancements, there remains a gap in understanding how both AI and EFL students handle the cultural and contextual complexity of bilingual communication. The study aims to evaluate the accuracy, fluency and contextual appropriateness of translations produced by both AI and EFL students, identifying translation challenges and strategies employed. Grounded in Koehn’s (2010) Statistical Machine Translation (SMT) model, this qualitative study engaged ten purposively selected EFL students. Participants collaboratively translated ten validated Arabic-English code-switched texts and were later interviewed to reflect on their translation experiences. These translations were compared to AI-generated outputs through qualitative textual analysis and thematic coding. The findings revealed that while AI produced fluent and grammatically accurate translations, it often failed to interpret cultural references, emotional tone and idiomatic expressions. In contrast, EFL students demonstrated greater flexibility, cultural mediation and pragmatic sensitivity. The study recommends enhancing AI systems with contextual-awareness capabilities and integrating AI tools into EFL translation pedagogy for guided post-editing. It also calls for training programs to prioritize cultural competence alongside linguistic skills. These findings have implications for the development of more adaptive translation technologies and enriched translator education in bilingual settings.