Qur'anic Balaghah and the Challenges of Artificial Intelligence (AI) in Semantic Interpretation

Abstract

This study aims to identify Qur'anic rhetorical (balaghah) elements produced by artificial intelligence models and to measure the capacity of three models, Claude, Gemini, and GPT, in handling Qur'anic semantics. A comparative testing method was employed by providing identical Arabic prompts to all three models on Q.S. Al-Baqarah verses 17–18 and Q.S. Al-Fatiha verses 2–5, with results benchmarked against Al-Zamakhshari's Tafsir Al-Kasysyaf as the primary reference. Results indicate that Claude and GPT were more terminologically accurate in identifying the rhetorical features of Al-Baqarah verse 17, while Gemini proved more precise on verse 18. For Al-Fatiha verses 2–5, Gemini produced the analysis closest to Al-Kasysyaf, particularly in explicating the iltifat phenomenon. However, all three models consistently failed to classify the rhetorical phenomena within the specific branch of 'Ilm al-Bayan. This study concludes that AI holds considerable promise as an auxiliary tool in rhetorical analysis, yet has not attained the epistemological depth of classical tafsir scholarship.