By Zubair Abbasi
This essay evaluates the potential of Generative Artificial Intelligence (GenAI) models—ChatGPT-4, Gemini, Co-Pilot, and DeepSeek—to apply the principles of Islamic inheritance law. Using a GenAI Evaluation Scale (GAIES), I assess the accuracy, comprehensiveness, and authenticity of AI-generated legal information on key topics such as gender-based inheritance differences, orphaned grandchildren's rights to inheritance, and state-reformed inheritance laws across multiple jurisdictions. My research finds that while ChatGPT-4 outperforms its counterparts, it still demonstrates significant shortcomings, such as fabricated references and legal misinterpretations.[1] While NotebookLM partially addresses this issue by citing exact page numbers from uploaded documents, it fails to provide comprehensive information. Based on my findings, I propose a GenAI Application Scale (GAIAS) to promote the ethical use of GenAI in Islamic legal education and research.[2]
Introduction
My research evaluates the performance of Large Language Models (LLMs)[3] such as ChatGPT-4, Gemini, Co-Pilot, and DeepSeek in providing accurate and authentic responses to inheritance-related queries under Sunnī and Shīʿī jurisprudence in diverse jurisdictions such as Pakistan, Egypt, and Iran. The objective of this study is to evaluate the performance of GenAI models in applying Islamic inheritance law, focusing on both classical legal principles and modern state-specific legal reforms.
To assess the performance of LLMs, I designed a GenAI Evaluation Scale (GAIES) to measure the accuracy of AI-generated legal information, and the authenticity of the references provided. Through various inheritance scenarios, the research tests the GenAI models' ability to handle both simple and complicated legal issues. It compares the AI-generated responses to actual jurisprudential rulings as prescribed by Sunnī and Shīʿī schools in authoritative classical and modern treatises. Based on the findings of this research, I propose GenAI Application Scale (GAIAS) to provide guidance on the practical utility of GenAI models for law students, scholars, and professionals in the field of Islamic law.
This research highlights both the potential and the limitations of GenAI models in the domain of Islamic inheritance law. GenAI models like ChatGPT-4 and DeepSeek hold promise as powerful tools for assisting legal scholars and researchers; however, their insufficient transparency and susceptibility to errors underscore the essential need for human oversight. NotebookLM may partially address this problem by citing exact page numbers from uploaded documents, but it lacks the capacity to provide comprehensive information. The study suggests that GenAI can play a supportive role in legal research by providing efficient, structured, and clear overviews that streamline the research process and contribute to a deeper understanding of complex legal rulings.
Methodology
This study evaluates the capabilities of ChatGPT-4, Gemini, Co-Pilot, and DeepSeek in applying the rules of Islamic inheritance law through a structured assessment. The study posed five questions ranging from basic to highly complex legal scenarios.[4] These included:
- Differences between Sunnī and Shīʿī inheritance laws.
- Simple propositions without Sunnī-Shīʿī divergences.
- Slightly complex cases with differing Sunnī and Shīʿī rulings.
- Moderately complex scenarios requiring adjustments among heirs.
- Jurisdiction-specific cases involving modern legal reforms.
The GAIES was developed to assess AI responses across five parameters:
- Accuracy: Alignment with Islamic legal principles.
- Comprehensiveness: Thoroughness of responses.
- Explainability: Clarity and reasoning behind responses.
- Authenticity of References: Reliability and authenticity of the sources cited.
- Structure of Information: Logical and coherent presentation.
Each GenAI model's responses were scored on a scale of 100, with each parameter contributing 20 points. This framework ensured a balanced evaluation of their capabilities and limitations.
Key Findings
After analyzing the responses generated by each GenAI model, three main conclusions emerged.
Firstly, ChatGPT-4 achieved an impressive overall score of 80.5% on the scale. DeepSeek followed by 77.5% score. These scores are significantly higher than Co-Pilot's medium score of 61.5% and Gemini's poor score of just 35.5%.
Secondly, a cascading effect was observed across the spectrum from simple to complex tasks. All four GenAI models performed better on simpler propositions and struggled with complex ones. Notably, the lowest scores for all models were on the final proposition, which required the application of Islamic inheritance law specific to different jurisdictions such as orphaned grandchildren's inheritance under Pakistani, Egyptian, and Iranian laws. This proposition required the models to apply modern reformed rules of Islamic inheritance law. For example, ChatGPT-4 cited the wrong statute in Egypt, Law No. 77 of 1943, instead of the correct statute, Law No. 71 of 1946,[5] also known as the compulsory bequest (wasīyyat wājibah), which grants orphaned grandchildren a share in the inheritance, up to one-third of the estate.[6] Similarly, while it correctly cited the relevant statute in Pakistan, section 4 of the Muslim Family Laws Ordinance 1961, it incorrectly stated that orphaned grandchildren do not inherit in the presence of a son of the deceased (their paternal uncle).[7]
Lastly, when evaluated on the five parameters of the GAIES, all four AI models scored highest in the areas of Structure of Information, Comprehensiveness, and Authenticity of References. Their scores were moderate in terms of explainability. However, accuracy was a weak point for all models, with Gemini scoring only 32% (6.5/20), DeepSeek and Co-Pilot 60% (12/20), and ChatGPT-4 67.5% (13.5/20).

The moderate accuracy of GenAI models raises concerns about the validity of AI-generated content. Google's NotebookLM partially addresses this problem by allowing users to upload documents and generating answers with exact page references. I tested it by uploading Coulson's Succession in the Muslim Family and posing five questions.[8] The accuracy rate was 100%, as all answers were based on the book and verifiable by checking the cited pages. However, NotebookLM fell short on other GAIES parameters. For instance, it failed to specify that a childless widow under Shīʿī school does not inherit land.[9] Similarly, on the inheritance rights of uterine brothers with full brothers, it did not clarify that, according to the Mālikī and Shāfiʿī schools, they share one-third of the estate—despite Coulson discussing this in detail.[10] This example highlights the limitations of even the latest GenAI models.
The GenAI Application Scale (GAIAS)
Based on the results of the GenAI Evaluation Scale (GAIES), I propose the GenAI Application Scale (GAIAS) to guide the effective use of AI models in Islamic inheritance law for students, educators, researchers, and scholars. This framework incorporates insights from current literature on GenAI integration in legal education.[11]
GAIAS serves as a standard for integrating GenAI into Islamic legal scholarship, with a primary focus on inheritance laws and potential applicability to other areas of Islamic law. It outlines practical guidelines for utilizing GenAI in learning, teaching, research and practice, emphasizing its benefits while addressing potential risks.
GenAI Application Scale (GAIAS)
1 | Full Use of GenAI as Research Assistant | GenAI models can generate structured and comprehensive information, including references to primary and secondary sources. However, all references must be verified for accuracy and authenticity. |
2 | Partial Use of GenAI Subject to Human Evaluation | GenAI models can assist in understanding the subject matter of Islamic inheritance law. However, users should exercise caution, as some of the AI-generated information may be unreliable or incorrect, even when accompanied by references to specific page numbers in books or articles. |
3 | Minimal Use of GenAI | GenAI models have limited utility in generating information regarding issues on which there are differences amongst various schools of Sunnī and Shīʿī inheritance laws. |
4 | No Use of GenAI | GenAI models should not be used to choose among multiple juristic opinions or interpret statutes across diverse jurisdictions, as these tasks require human expertise and nuanced judgment. |
5 | Ban on GenAI Use | GenAI must not be used to produce unethical or illegal content, such as material resembling primary Islamic sources (e.g., the Qur'ān and Sunna) or information intended to deprive rightful heirs of their inheritance shares. |
This Scale aims to integrate GenAI into Islamic legal education, research, and practice. Literature on GenAI in higher education highlights its benefits, including ChatGPT-4's role in enhancing learning. Empirical research shows that personalized chatbot interactions foster critical thinking, active learning, student engagement, and pedagogical innovation.[12] However, concerns remain about overreliance on GenAI, misinformation, trustworthiness, and declining academic rigor.[13]
Conclusion
Generative AI can assist in learning and applying Islamic inheritance law, one of the most important and complex areas of Islamic jurisprudence.[14] Tools like ChatGPT-4, DeepSeek, and NotebookLM can enhance learning experiences, streamline research, and foster comparative analyses. However, their limitations—such as fabricated references and incomplete answers—underscore the need for human oversight.
The GenAI Evaluation Scale (GAIES) used in this study reveals that significant improvements are still required, particularly in terms of accuracy, source validation, and adaptation to local legal developments and reforms. Therefore, Generative AI should be viewed as a supplementary tool that enhances and augments, rather than replaces, the expertise of human scholars and jurists in Islamic law.[15]
This study introduces the GenAI Application Scale (GAIAS) to guide the ethical and effective use of GenAI models in Islamic legal education and research. The objective of this scale is to serve as a guide to students, teachers, scholars, and practitioners of Islamic inheritance law to make the best use of GenAI models without falling into the trap of falsified information, which is presented in a confident and authoritative tone.[16]
Appendix I: Prompts
- You are an expert in Islamic law, please provide an answer to this question:[17] What are the differences between Sunni and Shia schools regarding Islamic inheritance law? Please provide references to your answer.
- You are an expert in Islamic law, please provide an answer to this question: Please calculate shares in inheritance if a deceased person leaves behind a husband and a son under Sunni and Shia inheritance laws. Please provide references to your answer.
- You are an expert in Islamic law, please provide an answer to this question: Please calculate shares in inheritance if a deceased person leaves behind a wife, a daughter, and a brother under Sunni and Shia inheritance laws. Please provide references to your answer.
- You are an expert in Islamic law, please provide an answer to this question: Please calculate shares in inheritance if a deceased leaves behind a husband, a mother, two uterine brothers, and two full brothers. Please provide references to your answer.
- You are an expert in Islamic law, please provide an answer to this question: Please calculate shares in inheritance if a deceased is survived by a son, and a deceased son's daughter under Islamic inheritance law as applied in Iran, Pakistan, and Egypt. Please provide references to your answer.
Appendix II: Glossary of Terms
- Generative AI (GenAI): A subset of artificial intelligence that generates human-like text, images, and other content by learning patterns from large datasets. GenAI is used in applications like chatbots such as ChatGPT-4, DeepSeek, Gemini, or Co-Pilot.
- GenAI Model: A machine learning system, such as ChatGPT-4, DeepSeek, Gemini, or Co-Pilot, that processes input data and generates responses based on training from vast amounts of text and structured data.
- GenAI Application Scale (GAIAS): A structured framework proposed in this study to guide the use of GenAI in Islamic legal education. The scale outlines the varying levels of GenAI assistance, ranging from full reliance for structured research to complete prohibition for ethical or legal concerns.
- GenAI Evaluation Scale (GAIES): A methodology developed to assess the accuracy, comprehensiveness, explainability, authenticity of references, and structural organization of AI-generated responses. This scale is designed to measure the effectiveness of GenAI tools in legal education and research.
- Large Language Model (LLM): A type of artificial intelligence system designed to process and generate human-like text by analyzing vast amounts of linguistic data. LLMs, such as ChatGPT-4, DeepSeek, Co-Pilot, and Gemini, use deep learning techniques, including neural networks and transformers, to generate coherent responses.
Notes:
[1] There have been several similar projects examining the (in)accuracy of GenAI models in correctly applying principles of Islamic law. See, for example, "Islamic Law and ChatGPT: Student Essays from the Islamic Law Lab," Islamic Law Blog, February 6, 2025, https://islamiclaw.blog/2025/02/06/islamic-law-and-chatgpt-student-essays-from-the-islamic-law-lab-2/.
[2] This essay is based on my current research project that explores the integration of GenAI into legal education, research, and adjudication. See Zubair Abbasi, "Can generative AI master Islamic inheritance law?," LSE Blog, October 11, 2024, https://blogs.lse.ac.uk/religionglobalsociety/2024/10/can-generative-ai-master-islamic-inheritance-law/.
[3] See Appendix II below for a glossary of terms used throughout the essay.
[4] See Appendix I below for detailed prompts.
[5] Ahmed Fouad, "Legalizing Bequests to Heirs in Egypt as a Legislative Application of Talfīq: towards a Purposive Interpretation of Article 37 of Law No. 71/1946," Yearbook of Islamic and Middle Eastern Law Online 23, no. 1 (2024): 204–32.
[6] Haseeb Fatima, "A Critical Appraisal of Obligatory Bequest as Prevalent in Muslim Countries," Islamic Studies 63, no. 2 (2024): 213–40.
[7] Shahbaz Ahmad Cheema, "Reappraisal of Lucy Carroll's Tripartite Thesis on Section 4 of Pakistan's Muslim Family Laws Ordinance 1961," Manchester Journal of Transnational Islamic Law & Practice 20, no. 2 (2024): 1–16.
[8] Noel James Coulson, Succession in the Muslim Family (Cambridge University Press, 1971).
[9] Muhammad Zubair Abbasi, "Fluidity of Sharī'a and the Modern State: Inheritance Rights of Childless Widows under Shīʿa Personal Law in British India and Pakistan," Yearbook of Islamic and Middle Eastern Law 23, no.1 (2024): 233–63.
[10] Coulson, Succession in the Muslim Family, 73–76.
[11] For details, see M. Perkins et al., "The Artificial Intelligence Assessment Scale (AIAS): A Framework for Ethical Integration of Generative AI in Educational Assessment," Journal of University Teaching and Learning Practice 21, no.6 (2024): 1–18; L. Furze et al., "The AI Assessment Scale (AIAS) in Action: A Pilot Implementation of GenAI Supported Assessment," arXiv preprint, arXiv:2403.14692 (2024): 1–18; Selçuk Kılınç, "Comprehensive AI Assessment Framework: Enhancing Educational Evaluation with Ethical AI Integration," arXiv preprint arXiv:2407.16887 (2024): 1–13; M. Perkins et al., "Navigating the Generative AI Era: Introducing the AI Assessment Scale for Ethical GenAI Assessment," arXiv preprint, arXiv:2312.07086 (2023): 1–8.
[12] John Bliss, "Teaching Law in the Age of Generative AI," Jurimetrics 64, no. 2 (2024): 111–61; Z. H. İpek, A. I. C. Gözüm, S. Papadakis, and M. Kallogiannakis, "Educational Applications of the ChatGPT AI System: A Systematic Review Research," Educational Process: International Journal 12, no. 3 (2023): 26–55; A. Tlili, B. Shehata, M. A. Adarkwah, A. Bozkurt, D. T. Hickey, R. Huang, and B. Agyemang, "What If the Devil Is My Guardian Angel: ChatGPT as a Case Study of Using Chatbots in Education," Smart Learning Environments 10, no. 1 (2023): 15; Y. Dai, A. Liu, and C. P. Lim, "Reconceptualizing ChatGPT and Generative AI as a Student-Driven Innovation in Higher Education," Procedia CIRP 119 (2023): 84–90.
[13] D. R. Cotton, P. A. Cotton, and J. R. Shipway, "Chatting and Cheating: Ensuring Academic Integrity in the Era of ChatGPT," Innovations in Education and Teaching International 61, no. 2 (2024): 228–39; D. O. Eke, "ChatGPT and the Rise of Generative AI: Threat to Academic Integrity?" Journal of Responsible Technology 13 (2023): 100060; R. H. Mogavi, C. Deng, J. J. Kim, P. Zhou, Y. D. Kwon, A. H. S. Metwally, and P. Hui, "ChatGPT in Education: A Blessing or a Curse? A Qualitative Study Exploring Early Adopters' Utilization and Perceptions," Computers in Human Behavior: Artificial Humans 2, no. 1 (2024): 100027.
[14] The Prophet Muhammad is reported to have said, "Learn laws of inheritance and teach them to people for they are one half of useful knowledge." Sunan Ibn Mājah, Chapter: 26, The Chapters on Shares of Inheritance. Syed Khalid Rashid, Muslim Law (4th edition, Lucknow: Eastern Book Company, 1963), 313–14.
[15] Mohammed Ghaly, "What Makes Work "Good" in the Age of Artificial Intelligence (AI)? Islamic Perspectives on AI-Mediated Work Ethics," The Journal of Ethics (2023): 1–25.
[16] Ali-Reza Bhojani and Marcus Schwarting, "Truth and Regret: Large Language Models, the Quran, and Misinformation," Theology and Science 21, no. 4 (2023): 557–63; C. Kidd and A. Birhane, "How AI can distort Human Beliefs." Science 380, no. 6651 (2023): 1222–23. L. Munn, L. Magee, V. Arora, "Truth Machines: Synthesizing Veracity in AI Language Models," AI & Society (2023): 1–15.
[17] I used "persona pattern" to give GenAI models a role to play when generating output. Jules White et al.,. "A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT," arXiv preprint arXiv:2302.11382 (2023).