An investigation into human vs AI English translations of Qur'anic euphemisms

This study examines the reliability of ChatGPT-5, an advanced AI language model, in translating Qur'anic euphemisms, expressions in the Holy Qur'an that convey sensitive meanings indirectly. It aims to evaluate whether AI can substitute or complement human translations of the Holy Qur'an. It compares ChatGPT-5 translation with that conducted by Abdullah Yusuf Ali, to highlight strengths and weaknesses of both translations. This study adopts the typology of euphemisms proposed by Samoškaitė, particularly those related to sex, to assess each translation using a combined framework drawn from Translation Quality Assessment, Multidimensional Quality Metrics and Scopos theory. This framework clearly defines the criteria of accuracy, fluency, contextual relevance, clarity, readability and cultural sensitivity, linking linguistic performance to the broader communicative goals of Qur'anic translation. The data of the study consist of eight Qur'anic verses containing euphemistic expressions, selected for their thematic and cultural sensitivity and are qualitatively analysed based on the adopted framework. The findings showed glaring disparities and inconsistencies in the accuracy of ChatGPT-5 translations, especially in the translations that involved subtlety of meaning inscribed in the Qur'anic euphemisms. It is noticed that while ChatGPT-5 demonstrated strengths in producing fast, grammatically correct and generally accessible translations, it struggled in rendering the deeper nuances of certain Qur'anic euphemisms. In particular, it tended to rely on literal, word-for-word translation without adequately capturing contextual subtleties. Although ChatGPT-5 outputs were sometimes clear and workable, they often lagged in clarity, readability and faithfulness to the original Arabic text. In contrast, Ali's translation is more accurate, faithful and effective, whereby it could provide clearer and more precise words without losing any meaning within the context. This study postulates that unlimited reliance on ChatGPT translations of Qur'anic texts is not warranted, as human involvement remains indispensable in building an understanding that is more accurate and sensitive to culture.