GenAI in Islamic Education Needs Human Mediation and Source Verification

A systematic review of generative artificial intelligence in Islamic religious education finds that the technology is being discussed mainly as a teaching aid rather than a substitute for teachers, scholars or recognised religious sources. The analysis, which combined 13 studies published between 2022 and 2026, also finds that ethical safeguards must extend beyond familiar AI concerns such as privacy, bias and plagiarism to include religious accuracy, source authenticity, teacher oversight and legitimate authority in interpreting knowledge. The authors caution that the evidence remains limited, fragmented and largely conceptual, exploratory or based on perceptions, so the findings describe an emerging research field rather than demonstrated classroom effects.

The review was conducted under PRISMA 2020 guidelines and searched peer-reviewed English-language journal articles in Scopus and Web of Science. The researchers initially identified 124 records, including 78 from Scopus and 46 from Web of Science. After removing 21 duplicates and three records that did not meet publication eligibility rules, 100 records underwent title-and-abstract screening. Sixty-one were excluded, 39 reports were sought for full text, and two could not be retrieved. Two reviewers then assessed 37 available reports independently against predefined inclusion and exclusion criteria, resolving disagreements by consensus; 13 studies were retained for the final synthesis. The search was conducted in April 2026 and used combinations of terms covering generative AI, Islamic or religious education, and ethical concerns such as academic integrity, privacy, bias, transparency and governance.

Only five of the 13 included studies were classified as direct evidence from Islamic religious education, Qur’anic learning, Islamic teachers or Muslim learners. Two studies concerned religious or theological education without an Islamic-specific focus, while six provided contextual evidence from broader educational debates about generative AI, academic integrity, learner attitudes, reliability and AI ethics. This separation was central to the synthesis: indirect and contextual studies were used to identify questions and safeguards that might matter in Islamic education, but they were not treated as direct measurements of how generative AI affects Islamic classrooms. The distinction matters because general AI findings cannot automatically be assumed to describe the experiences of teachers and learners in religious settings.

The evidence indicates that generative AI is most often proposed as a bounded assistant for practical tasks. Applications include preparing lessons and teaching materials, designing assessments, creating examples, supporting personalised learning, providing formative feedback, assisting with administrative work and encouraging teacher-student interaction. In Qur’anic learning, researchers have explored pronunciation assessment, tajweed recognition and adaptive feedback, but these uses remain supplementary to teacher-mediated instruction. Surveys and qualitative studies indicate that some Islamic teachers see tools such as ChatGPT as potentially useful resources, while Muslim learners have expressed acceptance of some AI-supported approaches. Such findings show perceived usefulness and willingness to adopt a technology, not that it improves religious understanding, moral development or learning outcomes.

The review identified risks that extend beyond ordinary academic errors. Hallucinated passages, fabricated citations, inaccurate or decontextualised explanations, bias, privacy violations, plagiarism and unclear authorship can occur in any educational use of generative AI. Their significance increases when the system is involved in Qur’anic knowledge, religious beliefs, ethical guidance or the interpretation of sacred texts. The literature therefore raises possible concerns about doctrinal distortion, confusion over who is qualified to interpret religious claims and the weakening of learners’ ability to evaluate authentic sources. These remain potential risks identified in research and theory, not established outcomes of classroom implementation, because direct empirical evidence is too limited to demonstrate that such effects routinely occur.

A particular problem is that fluent language can make a system appear authoritative even when it lacks the textual, scholarly, institutional or religious legitimacy required to interpret religious knowledge. The review distinguishes religious authority, which concerns recognised guidance from religious persons, institutions, traditions and sources, from epistemic authority, which concerns who or what can legitimately validate, interpret or transmit a knowledge claim. Teachers, scholars, texts and institutions may possess both forms of authority, but a generative model does not acquire them merely by producing coherent answers. This distinction addresses the risk that learners may mistake generated material for established religious knowledge, especially if institutions do not explain how its outputs should be checked.

To address that risk, the authors propose a preliminary Islamic ethical framework built around three complementary concepts: adab, amanah and epistemic authority. Adab concerns appropriate conduct towards knowledge, teachers, sources and the learning process, including whether AI supports diligent study rather than encouraging passive acceptance of generated answers. Amanah concerns trust, responsibility and accountable use of knowledge, encompassing disclosure of AI assistance, verification of outputs, protection of learner data, proper attribution and responsibility for final work. Epistemic authority concerns the legitimate validation and interpretation of religious claims. The framework is explicitly provisional and literature-derived; the three concepts are not presented as an exhaustive account of Islamic ethics or as validated classroom rules.

These ideas translate into several layers of recommended safeguards. Teachers should review AI-generated materials for accuracy and provide appropriate religious context, while students should learn to treat generative AI as an assistive tool rather than an authority on Islamic doctrine, jurisprudence, Qur’anic interpretation or theology. Recognised sources should be used to verify religious claims, and qualified educators or scholars should mediate content when necessary. Institutions were encouraged to adopt context-specific policies covering academic integrity, disclosure, privacy, assessment design, appropriate and prohibited uses, teacher responsibilities and procedures for checking generated information. Teacher preparation, AI literacy and student guidance were repeatedly identified as necessary components, while the review noted that broader religious and general AI literature provides supporting context rather than direct evidence of effectiveness in Islamic education.

The analysis also exposes major gaps that should temper expectations for rapid implementation. The directly relevant literature is geographically narrow and contains few classroom-based, longitudinal, comparative or intervention studies. Most included work was conceptual, exploratory, literature-based or perception-based, and the quality appraisal used a common six-part framework suited to heterogeneous designs rather than a design-specific risk-of-bias tool. All 13 studies met the review’s descriptive threshold, but their scores cannot be read as proof that the different study designs are methodologically equivalent. The search also excluded databases such as ERIC, ATLA Religion Database and ProQuest, so relevant work indexed only there may have been missed. Future studies should test the proposed framework in madrasahs, Islamic schools, universities, Qur’anic learning centres and teacher-education programmes.

The central result is therefore not that generative AI has been proven beneficial or harmful in Islamic religious education. It is that the available literature points toward a conditional, supervised and human-mediated role: systems may assist with bounded educational tasks, but teachers and recognised religious sources retain responsibility for validating and interpreting religious knowledge. Future research should move beyond measuring whether users accept the technology and instead examine effects on knowledge acquisition, source evaluation, critical thinking, academic integrity, disclosure practices, teacher-student relationships and moral formation. Those questions will require empirical evidence before the proposed ethical framework can be treated as an established model or used to claim concrete benefits in Islamic classrooms.

Subject of Research: Ethical and epistemic safeguards for generative AI in Islamic religious education

Article Title: Ethical integration of generative artificial intelligence in Islamic religious education involves human mediation and source verification

Article References: Mat Yusoff, S., Mohamad Marzaini, A. F., Zynuddin, S. N., Mat Jusoh, A. M., Basal, M. H., Alias, N. R., & Saad, N. (2026). Ethical integration of generative artificial intelligence in Islamic religious education involves human mediation and source verification. Discover Artificial Intelligence, 6(1), Article 1162. https://doi.org/10.1007/s44163-026-02251-7

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02251-7

Keywords: generative AI, Islamic education, religious education, AI ethics, teacher mediation, source verification, epistemic authority, academic integrity, Qur’anic learning, systematic review, Ethical, integration