People Are Asking AI for Fatwas… And the Answers Are Terrifying

Somewhere in Jakarta, a young woman types a question into ChatGPT about whether her husband’s divorce pronouncement over WhatsApp is legally binding under Islamic law. In Karachi, a first-generation university student asks an AI app whether his student loan is haram. In Dearborn, Michigan, a father queries an Islamic chatbot about whether the life insurance policy he is considering for his family is permissible. In Birmingham, a widow wants to know how her late husband’s estate should be divided among four children and a second wife.

None of them call a scholar. None of them walk into a mosque. They type. They wait three seconds. They receive an answer. And in the vast majority of cases, they believe it.

This is not a speculative scenario painted for dramatic effect. This is Thursday morning, somewhere in the world, among the 2.1 billion people who comprise the global Muslim community. And it is happening at a scale, and with a confidence, that should alarm every custodian of Islamic scholarship alive today.

The algorithm is issuing fatwas. It is getting them wrong. And almost nobody — not the platforms, not the governments, not the ulema — is doing anything serious enough to stop it.

Related:  How AI Created Fake Muslim Scholars and Spread Rulings They Never Made

The Machine That Never Studied Fiqh

To understand what is at stake, one must first understand what a fatwa actually is — and what it is emphatically not.

A fatwa is not a Google search result. It is not a frequently asked question from a religious website. It is a juristic opinion, rendered by a qualified scholar — a mufti — who has spent years, often decades, mastering the Quran, the Sunnah, the principles of usul al-fiqh (Islamic legal theory), the four schools of jurisprudence, and the vast inherited canon of classical Islamic law. Crucially, a fatwa is not issued in a vacuum. It is issued in context: the scholar listens, probes, considers the circumstances of the individual, weighs the doctrine against human reality, and — carrying the intense weight of accountability before Allah — delivers a ruling.

That accountability is not incidental. It is the very architecture of the institution. The mufti who errs does not simply give bad advice. In the theological framework of Islam, he bears a portion of the sin of those led astray. This is the fear that has made great scholars humble. It is the gravity that has made the institution of fatwa one of the most sophisticated moral technologies in human history.

Now feed it into a language model trained on the internet.

The results, as researchers are beginning to document, are exactly what one would expect: a system that sounds authoritative, moves with the confidence of certainty, and is — by any serious Islamic jurisprudential standard — profoundly unqualified. A 2024 academic study published in the Journal of Contemporary Islamic Law examined GPT-3.5’s performance in issuing rulings on Muslim family law, including matters of marriage, divorce, and inheritance. The verdict was damning: GPT-3.5 has issued incorrect fatwas in real family law cases. Not occasionally. Systemically.

The Islamic Information portal has documented cases where AI chatbots declared items halal that are haram, provided incorrect hadith reference numbers, and misattributed Quranic verses — fabricating citations with the serene self-assurance of a scholar who has never known doubt.

A peer-reviewed 2025 study in the Journal of Digital Islamicate Research assessed five dedicated Islamic chatbots and found “significant performance variability,” with every chatbot tested failing to meet the standards of Islamic legal reasoning. The failures were not edge cases. They were the norm.

The Manara in the Mosque

If you wanted a single image to capture the audacity of this moment, you could not do better than what Saudi Arabia unveiled in May 2025: a gleaming, 21-inch touchscreen robot named Manara, installed inside the Grand Mosque in Mecca itself — the holiest site in all of Islam — to answer pilgrims’ religious questions in eleven languages, including Arabic, English, Urdu, Bengali, and Malay.

The Manara robot, decked in Islamic geometric motifs, connected via 5G, equipped with high-resolution cameras, is positioned at the very threshold of the Masjid al-Haram. Hundreds of thousands of pilgrims from every corner of the earth pass by it daily during the Hajj season. Many of them stop. Many of them ask.

The Saudi authorities were careful in their framing: the robot, they said, draws on a database of pre-approved fatwas and, for complex questions, facilitates a video call with a live scholar. The intent, framed through the gleaming language of Vision 2030, is efficiency — to serve the linguistically diverse ocean of pilgrims with religious guidance without bottlenecks.

One understands the logic. One also understands what happens when the pre-approved database encounters the infinite complexity of human circumstance, and the fallback video call is unavailable, and the pilgrim — fatigued, emotional, spiritually raw — simply accepts what the touchscreen tells him.

The symbolism of the location deserves to sit with us. If AI religious guidance can be normalized at the Kaaba’s doorstep, it will be normalized everywhere. And it already is.

The Accountability Vacuum

Here is the question nobody in Silicon Valley, Riyadh, or the boardrooms of Islamic fintech is asking with sufficient urgency: when an AI chatbot issues a wrong fatwa, who is accountable?

In the traditional model, the chain of accountability is clear. A mufti issues an erroneous ruling; he can be challenged by other scholars, rebuked by his institution, and — in the framework of Islamic ethics — will answer before Allah. His name is attached to his opinion. His reputation stands or falls on his scholarship. Dar al-Ifta in Egypt, the Islamic Fiqh Academy of the Organisation of Islamic Cooperation, the Fatwa Committees of Malaysia and Indonesia — these are institutions with oversight structures, scholarly hierarchies, and centuries of self-correcting tradition.

Now contrast that with an AI chatbot.

The model has no name in any jurisprudential sense. It has no ijazah — the chain of scholarly certification that legitimizes an Islamic scholar’s authority. It has no taqwa, no God-consciousness, no existential stake in the correctness of its output. It cannot be held to account in any framework — theological, legal, or institutional — that Islam has ever recognized. Its errors vanish into the ether. The user who acted on the wrong ruling is left to bear consequences — in their marriage, their finances, their conscience — alone.

Egypt’s Dar al-Ifta, one of the oldest and most respected fatwa institutions in the world, has been unequivocal: it is not permissible to depend on an AI-produced fatwa or to act upon it. “Artificial intelligence,” the institution stated, “is not qualified to issue fatwas.” The Islam Q&A platform, widely followed across the Sunni world, echoed the same ruling. These are not the voices of technophobic reactionaries. They are the voices of institutions that have navigated telegraph, radio, television, and the internet — and found, in each case, a way to adapt tradition to modernity without surrendering the essence of jurisprudential integrity.

What is different about AI is not its novelty. What is different is the plausibility of its outputs, and the invisibility of its failures. A clearly uneducated person giving bad religious advice is easy to dismiss. A chatbot that speaks in fluent, confident, well-formatted prose — quoting the Quran, citing hadith, referencing schools of jurisprudence — is not. It wears the costume of scholarship without its soul.

The Training Data Problem, and Why It Is Theological

Let us be precise about what a large language model actually does when it “answers” a religious question.

It does not reason. It does not weigh evidence. It does not exercise ijtihad — the rigorous independent juristic reasoning that Islamic scholars undertake when applying primary sources to novel situations. It predicts the next most statistically likely word, based on patterns learned from an enormous corpus of text — text that includes, indiscriminately, the profound and the profane, the scholarly and the speculative, the authentic hadith collections and the fabricated ones, the careful jurisprudence and the social-media hot takes, the centuries of painstaking scholarship and the blog posts written in an afternoon.

This is not a minor technical caveat. It is a theological catastrophe waiting to unfold.

Islam’s epistemological tradition is extraordinarily careful about the chain of transmission — isnad — precisely because it understands that knowledge, to be trustworthy, must be traceable. The entire science of hadith criticism, ‘ilm al-rijal, is built on the principle that you cannot accept a narration without knowing who transmitted it and whether they were trustworthy. The Quran itself commands: “If a fasiq (transgressor) brings you news, verify it” (49:6).

An AI model cannot tell you where its answer comes from. It cannot distinguish between a narration in Sahih Bukhari and a fabricated hadith circulating on a dubious website. It has no chain of transmission. It has no isnad. In the most rigorous Islamic epistemological sense, its religious outputs are, categorically, unverifiable — and therefore unreliable — by design.

Researchers studying Islamic chatbots have confirmed what Islamic epistemology would predict: these systems misattribute Quranic verses, hallucinate hadith references, oversimplify rulings that depend on contextual nuance, and — most dangerously — provide internally consistent but jurisprudentially incorrect answers in a voice that radiates certainty.

One researcher described it with chilling precision: the chatbot operates with the confidence of a scholar and the knowledge of a layman who has read widely but understood shallowly.

The Fatwa Market and the Attention Economy

The AI fatwa crisis does not emerge in isolation. It is the latest — and most dangerous — chapter in a longer story about the erosion of Islamic scholarly authority in the digital age.

For the past two decades, the internet has been quietly dismantling the institutional structures that once gave Islamic jurisprudence its coherence. Scholars with formal training competed, and often lost, against charismatic YouTube preachers, viral TikTok du’at, and Telegram channels with millions of followers run by individuals whose scholarly credentials are, at best, informal. The phenomenon of fatwa shopping — seeking religious rulings from whichever source provides the most convenient answer — has metastasized in the digital environment, where the supply of opinions is infinite and the cost of switching scholars is zero.

Research published in 2025 in the journal Religions on digital mediation and fatwa authority captured this transformation precisely: “Platform logics incentivise brevity, emotional resonance, and virality” — precisely the opposite of the deliberative, contextual, accountability-laden process through which a legitimate fatwa is produced.

Into this already-fractured landscape, AI arrives with a superpower: it is available at 3 a.m. It speaks your language. It never makes you feel judged. It answers immediately, confidently, and for free. Against the friction of seeking a qualified scholar — finding one, making an appointment, explaining your situation, accepting an answer you might not want — the chatbot wins on convenience every single time.

This is not a story about technology. It is a story about what happens to trust when authority becomes frictionless, and accountability becomes optional.

What Is at Stake: Not Theology, But Lives

There is a temptation — particularly in the corridors of the Islamic finance industry, which has spent three decades professionalizing Sharia compliance — to treat this as an abstract theological debate. It is not. The rulings that people are outsourcing to algorithms carry real-world consequences of devastating weight.

Consider the woman asking about the validity of her divorce. If the AI tells her the talaq was invalid, she may consider herself still married. If it tells her she is divorced, she may remarry. In both cases, an error does not merely offend doctrine — it can strand a human being in a marriage she believes does not exist, or expose her to the Quranic prohibition on marrying without completing the iddah (waiting period). These are not theological abstractions. They are lived realities, with children involved, with inheritance implications, with the deepest questions of a person’s lawful life at stake.

Consider the halal food entrepreneur who uses an AI to verify the permissibility of a new ingredient in his supply chain — and receives a confident, wrong answer. Consider the Muslim investor who asks a chatbot whether a particular sukuk structure is Sharia-compliant, and acts on the result. Consider the young couple asking about Islamic marriage contracts, or the dying parent asking about the Islamic rules of bequest, or the new Muslim trying to understand what prayer times apply during polar travel.

In every one of these cases — which are happening, daily, at scale — an AI error is not a data point. It is a life, touched wrongly, by a system that bears no consequences for being wrong.

The Scholars Are Talking. Is Anyone Listening?

It would be inaccurate to suggest that Islamic scholarship has been silent. The responses have been serious, and in some cases, remarkably sophisticated.

Egypt’s Dar al-Ifta issued a formal ruling. Islam Q&A published a detailed fatwa on why seeking religious rulings from AI is impermissible. The Jamiatul Ulama of South Africa published a remarkable piece — noted for the irony that it was written by ChatGPT itself, prompted to enumerate its own religious dangers — cataloguing the harms of AI Islamic guidance. Academic journals from Malaysia’s International Islamic University to Unisa in South Africa are publishing peer-reviewed assessments of AI’s jurisprudential failures. The scholarly conversation is alive and serious.

But scholarship, however serious, operates at the speed of institutional deliberation. The chatbots operate at the speed of 5G. The mufti publishes a ruling on a website; the chatbot answers ten million queries overnight. The institutional responses, however correct, are fighting a distribution war with asymmetric weapons.

What is notably absent is regulatory will. No Muslim-majority government has yet enacted meaningful legislation governing AI-generated religious content. No international Islamic body — not the OIC’s Fiqh Academy, not the AAOIFI, not the Islamic Financial Services Board — has issued binding governance standards for AI use in religious guidance. The Manara robot operates at the Grand Mosque with no publicly disclosed oversight framework for its error-correction protocols.

The contrast with Islamic finance is instructive. It took decades of scholarly effort, institutional architecture, and regulatory development to build a credible Sharia compliance infrastructure for the Islamic finance industry — Sharia boards, AAOIFI standards, central bank Sharia guidelines. That infrastructure is imperfect, and perpetually contested, but it exists. It has accountability built in.

For AI religious guidance, we are starting from zero. And the technology is already deployed in the holiest mosque on earth.

A Faith Community at a Crossroads

Thirty years of covering the Islamic economy have taught me one immutable truth: the Muslim world’s greatest asset is also its greatest vulnerability. That asset is the depth and sophistication of its jurisprudential tradition. That vulnerability is the assumption that tradition will protect itself.

It will not. Not without deliberate, urgent, and sustained institutional action.

The Islamic economy — which the Dinar Standard now values at over $7 trillion across its halal food, Islamic finance, modest fashion, pharmaceutical, tourism, and media sectors — is predicated, at its foundation, on Sharia compliance. That compliance is only as credible as the scholarly infrastructure that certifies it. If that infrastructure is quietly replaced, at the consumer level, by algorithmically confident but jurisprudentially incompetent chatbots, the entire edifice is built on sand.

There are things that can be done, and done now. The major Islamic scholarly bodies must treat AI religious guidance as a matter of communal urgency — fard kifaya, a collective obligation — and move beyond position papers to governance frameworks. Muslim-majority governments must regulate AI in the religious domain with the same seriousness they apply to AI in finance and healthcare. Islamic technology platforms — the Muslim Pros, the Bayan apps, the Zakat calculators — must adopt, and transparently disclose, their policies on AI-generated religious content. And the AI companies themselves — OpenAI, Google, Meta — must be challenged, through scholarly engagement and regulatory pressure, to implement meaningful guardrails on religious queries from Muslim users.

Most fundamentally, Muslim communities — parents, imams, educators — must make digital religious literacy a priority of Islamic education. A generation that cannot distinguish a hallucinated hadith from an authenticated one, that does not understand why the source of Islamic knowledge matters, is a generation that will not ask the right questions.

The algorithm cannot do tawbah. It cannot make istighfar. It will not stand before Allah and account for the Muslims it misled.

But the scholars who saw this coming, the institutions that had the resources to respond, the governments that had the regulatory tools available — they will.

The question is not whether we can prevent AI from entering the space of Islamic religious life. That battle, if it was ever winnable, is already lost. The question is whether the Muslim world will build the governance architecture to ensure that when the machine speaks on matters of din, there is a scholar — real, accountable, qualified — somewhere in the chain.

And the answer to that question, right now, in the year 2026, is: not yet.