<![CDATA[The rapid adoption of generative artificial intelligence (AI) has expanded its application to Qur'anic translation and interpretation, raising important questions regarding the reliability of AI-generated outputs for theologically sensitive verses. Among the most challenging passages are anthropomorphic (mutashābihāt) verses, whose translation requires careful negotiation between literal meaning and established theological traditions. This study compares the translation strategies employed by ChatGPT and Gemini in rendering anthropomorphic Qur'anic expressions and evaluates the consistency of their outputs across repeated interactions. This research employs a qualitative comparative design using library research and content analysis. Four categories of anthropomorphic expressions—wajh (face), aʿyun (eyes), yad (hand), and istawā (rising over the Throne)—were translated repeatedly by both AI systems using identical prompts administered in different sessions. The resulting outputs were classified into three interpretive approaches: literal translation, ta'wīl (figurative interpretation), and tafwīḍ (consigning the ultimate meaning to God). Comparative analysis was then conducted to examine translation preferences and output consistency. The findings demonstrate substantial differences in theological translation strategies between the two AI models. ChatGPT predominantly adopts literal translation (39 of 72 outputs), followed by ta'wīl (19) and tafwīḍ (14), whereas Gemini produces more literal renderings (42 of 72 outputs) together with ta'wīl (30) but does not employ tafwīḍ. ChatGPT also exhibits strategy shifts across repeated sessions, particularly in translating istawā, where responses change from tafwīḍ to literal interpretation. By contrast, Gemini maintains comparatively stable translation patterns despite relying primarily on literal formulations. These findings indicate that generative AI systems differ not only linguistically but also in their implicit theological orientation when processing anthropomorphic Qur'anic expressions. This study contributes to the emerging field of digital Qur'anic studies by providing a systematic framework for evaluating AI translation strategies from a theological perspective. It demonstrates that AI-generated Qur'anic translations should not be regarded as doctrinally neutral, particularly when dealing with verses carrying significant theological implications. The study is limited to two large language models and four categories of anthropomorphic expressions. Future research should expand comparisons to additional AI systems, multilingual translation settings, and broader corpora of mutashābihāt verses to develop more comprehensive evaluation standards for AI-assisted Qur'anic interpretation.]]>
Copyrights © 2026
