Digital Authority and Islamic Legal Reasoning: A Comparative Study of AI-Generated and Human-Issued Fatāwā

Abstract

In today’s digital world, artificial intelligence is becoming more common in every area of modern life, even religion. Increasingly, Muslims globally have been looking to AI chatbots for religious guidance about questions once reserved for qualified Islamic scholars. This paper compares the fatāwā produced by ChatGPT [model version GPT-4o], the most wide-ranging and deployed generative AI language model in human history to date with those of traditional Islamic institutions and contemporary Islamic scholars on a controversial contemporary topic yoga to identify whether or not the output of AI could ever be said in good faith to represent Islam. ChatGPT was chosen for its preeminence: with over 100 million active users, it’s the most likely A.I. tool Muslim laypeople turn to for answers to religious questions. However, AI is an evolving subject, and the results of this study pertain only to GPT-4o. Future models will not behave in the same way, and this should be regarded as a time-locked contribution. This study uses a comparative qualitative approach in examining fourteen human-authored fatāwā related to yoga obtained from publicly available online archives only, and therefore mirrors content that ChatGPT could have reasonably acquired. Each fatwā was presented to ChatGPT as a question, and then its response was evaluated according to factual correctness, citations of Islamic literature (if any), use of legal reasoning (Uṣūl al-Fiqh), and whether or not the original ruling remained intact. The results indicate that ChatGPT answered correctly six out of fourteen times, while it failed eight times. This failure reflects a systematic bias toward permissiveness: ChatGPT tends to rely on secondary sources rather than consulting authoritative collections of fatāwā, or at least in criticizing schools of thought where fatāwā have arisen. The rest of the paper seeks to argue that AI programs like ChatGPT can never achieve the status of an Islamic judge (fatwā). This assumes an important basis of digital literacies from Muslim users, integrating Islamic principles into replicable design systems for AIs with credentials such as scholars working alongside tech designers, and a broader rethinking of what ethical AI means and how to best be ethical about it in ways that speak to the life-worlds of the relevant Muslim religious contexts.