AI Tools in Translation: Bridging Literalism and Cultural Fluency in English-Arabic Idioms – A Comparative Analysis of ChatGPT and DeepSeek

Abstract

Idioms are fixed phrases whose forms are usually unchangeable and whose meanings are always the same, metaphorical, indirect and inflexible. They are considered as cornerstone of linguistic and cultural nuance. Therefore, their translation poses formidable challenges for AI-driven machine translation. The aim of the current study is to evaluate the performance of two leading AI tools, ChatGPT and DeepSeek, in translating 20 English idioms into Arabic. The study employs two theoretical frameworks from translation studies—Nida’s dynamic equivalence as well as Vinay and Darbelnet’s translation strategies—alongside principles of neural machine translation (NMT). The analysis is based on three criteria. These are (1) accuracy, (2) cultural appropriateness, and (3) naturalness. Through this tripartite analysis, the findings of the study reveal distinct approaches of the AI tools under scrutiny. DeepSeek frequently employs modulation as well as culturally adapted alternatives This aligns closely with Nida’s emphasis on meaning preservation over literality. In contrast, ChatGPT often defaults to literal translations of meaning. Thus, it risks eroding idiomatic essence. The study concludes that translators must critically evaluate outputs to ensure cultural fidelity. For example, a translator must critically evaluate DeepSeek’s literal translation of the idiom “signing on the dotted line” as (“وقعوا على الخط المنقط”) and modify it to be “وقعوا على بياض”) to suit Arabic readers. Furthermore, the study advocates for a collaborative human-AI workflow (i.e., AI generates baseline translations and humans refine them through cultural vetting as well as stylistic adjustments). The study recommends that future advancements in NMT systems must prioritize idiom-specific training datasets and culturally annotated corpora, in addition to integrating dynamic equivalence principles with adaptive strategies with the aim of enhancing cultural fluency. Thus, by bridging algorithmic precision with human intuition, AI can evolve into a culturally intelligent partner that empowers translators.