Cultural nuance and pragmatic loss in AI-generated arabic–english translation: A comparative study of figurative language

Abstract

Recent advances in artificial intelligence, particularly neural machine translation (NMT), have significantly improved the fluency and efficiency of automated translation systems. However, concerns persist regarding their ability to convey culturally and pragmatically embedded meaning. This study investigates how AI-generated Arabic–English translations handle figurative language, metaphor, and culture-bound expressions in comparison with human translations. Adopting a qualitative, theory-driven approach, the study analyzes a purpose-built corpus of culturally marked Arabic texts, including religious, poetic, idiomatic, and political discourse. The analysis is grounded in Functional Equivalence Theory, Relevance Theory, and the domestication–foreignization framework. Findings reveal that AI systems consistently prioritize literal and surface-level rendering, resulting in pragmatic loss, weakened rhetorical force, and diminished cultural resonance. Human translators, by contrast, employ adaptive strategies such as explicitation, idiomatic substitution, and semantic amplification to achieve communicative adequacy. The study concludes that while AI excels in formal accuracy, it remains limited as a cultural mediator, reaffirming the indispensable role of human translators in figurative and culturally sensitive translation.