Artificial Reasoning and Islamic Law: Mapping LLM Capabilities Across Juristic Ranks and Reasoning Frameworks

The growing use of Artificial Intelligence (AI) in religious inquiry raises a question about how these systems can contribute to Islamic legal reasoning. The current study aimed to examine whether Large Language Models (LLMs) can approximate the forms of analysis that Islamic jurisprudence expects from its practitioners. Islamic legal tradition has identified ranks of competence, from the independent mujtahid to the mufti and to the muqallid, each with expectations in terms of language, evidence, and judgment. The study assessed LLM performance in relation to these expectations by examining their training data, reasoning patterns, and results of empirical evaluations. Recent models show improved results on structured and rule-based tasks, such as inheritance calculations, where fine-tuning and retrieval-augmented methods improve accuracy and reduce hallucination rates. They can classify concepts, organize texts, and reproduce deductive patterns found in juristic writings. These strengths support LLMs as juristic research tools. Important limitations remain. LLMs cannot judge the strength of evidence, reconcile competing juristic views, or distinguish between equivocal and unequivocal proofs. They treat textual inputs without considering evidentiary hierarchy and therefore lack calibration for the assessment of proofs. Most critically, they cannot access what this study calls the passive contextual element in juristic reasoning that relies on awareness of intention, social custom, social conditions, and consequences of a ruling, none of which LLMs can reliably infer or evaluate. The study concluded that LLMs hold an intermediate position. They assist juristic research through retrieval, organization, and structured reasoning, yet they cannot assume the epistemic or ethical responsibilities that shape Islamic legal judgment.