In “Can LLMs Write Faithfully? An Agent-Based Evaluation of LLM-generated Islamic Content” (Conference on Neural Information Processing Systems), Abdullah Mushtaq (Information Technology University) and others—including former PIL Fellow Ezieddin Elmahjub (Qatar University)—observe that “large language models are increasingly used for Islamic guidance, but risk misquoting texts, misapplying jurisprudence, or producing culturally inconsistent responses.” They “pilot an evaluation of GPT-4o, Ansari AI, and Fanar on prompts from authentic Islamic blogs. [Their] dual-agent framework uses a quantitative agent for citation verification and six-dimensional scoring (e.g., Structure, Islamic Consistency, Citations) and a qualitative agent for five-dimensional side-by-side comparison (e.g., Tone, Depth, Originality).
