Skip to content
Shariah Agent
العربية
Menu
Can AI issue a fatwa?Check an answerAI vs scholarThe StandardGlossaryAbout

Can you trust an AI with a Shariah question? Here is how to tell.

What is a Shariah agent? AI that answers Islamic questions, and the line it must not cross

A Shariah agent is an AI system that retrieves, explains, or acts on Islamic rulings and finance rules. The useful ones show their sources, name the school, date the ruling, admit uncertainty and send you to a scholar. None of them is a mufti, and Egypt's Dar al-Ifta has ruled that software may not be relied on to issue a fatwa. This site explains the difference and keeps score.

All guides Can AI issue a fatwa?
By HalalScreener ResearchUpdated 6 min read

Type "is my pension halal" into any AI assistant and you will get a confident paragraph in under three seconds. The paragraph may be right. It may be a plausible blend of three schools and a Reddit thread. Nothing in its tone will tell you which. This site exists to give you a way to tell, and to hold the people building these tools, ourselves included, to a public standard.

The short version
  1. Agents can retrieve and apply; they cannot rule. Dar al-Ifta in Egypt has said relying on AI to obtain a fatwa is impermissible. In the Ja'fari school rulings come from a qualified living jurist. Both traditions leave no slot for software as a mufti.
  2. Grounded beats fluent. An answer that quotes a named, dated, linked source from a named scholar or institution can be checked. An answer generated from memory cannot, however good it sounds.
  3. The spectrum matters. Search, chatbot, grounded assistant, tool-using agent, autonomous agent: each step adds usefulness and risk. Know which one you are talking to.
  4. Five questions sort good tools from dangerous ones. Source, school, date, uncertainty, hand-off. The test is below and on its own page.
  5. We publish a standard and apply it to ourselves. The Shariah Agent Standard is seven commitments any builder can adopt. HalalScreener, which runs this site, is measured against it on that page.

The spectrum: from search box to autonomous agent

less autonomymore autonomySearchfinds pagesfineChatbotwrites an answer from memoryrisky for rulingsGrounded assistantquotes a retrieved sourcenames the schooluseful, still not a muftiTool-using agentruns screenscalculationsalertsneeds limits and logsAutonomousacts on money without a humannot today
Five kinds of tool that all get called 'AI'. The risk rises left to right. So does the usefulness, if the limits are right.

Search. Finds pages. You read them. Fine, because you are doing the reasoning.

Chatbot. Writes an answer from what the model absorbed in training. Fast, fluent, and unverifiable. Academic testing of general-purpose models on fiqh questions has found a systematic tilt toward permissiveness and a reliance on secondary sources rather than authoritative fatwa collections. The problem is not that the chatbot is always wrong; it is that you cannot tell when it is.

Grounded assistant. Retrieves a passage from a defined library of scholarly sources, quotes it, names the school and the scholar, and declines when nothing relevant is found. This is the first level at which an AI answer becomes checkable, and it is the minimum this site considers acceptable for religious questions.

Tool-using agent. Takes a goal and executes: screen these fifty stocks against this published ruleset, compute the purification amount, watch for changes and alert me. Powerful and legitimate when the rules were written by scholars and every action is logged. Dangerous when it is allowed to improvise.

Autonomous agent. Acts on money or on religious decisions with no human in the loop. Not acceptable today under any tradition's governance rules, and not offered by any responsible builder we know of.

What the scholars have actually said

Software and the authority to rule

Sunni schools

Egypt's Dar al-Ifta, in fatwa 22255, holds that using AI technology is permissible in principle and that its ruling depends on the purpose it serves, but that it is impermissible to rely on AI applications to obtain a fatwa. A fatwa requires a qualified, accountable, trustworthy human scholar; current tools lack sound methodology, knowledge of authoritative sources, and the ability to assess their own conclusions; and rulings tied to custom change with time and place.

In January 2026 the same body ruled that relying independently on AI to interpret the Quran is forbidden, while noting this is not a rejection of technology. Egypt has at the same time explored AI-assisted systems to support human muftis, which shows the line is drawn at authority, not at tools.

Ja'fari (Shia) school

In the Ja'fari school a layperson follows the rulings of a qualified living jurist (a marja' al-taqlid) whose authority rests on recognised scholarship, justice and accountability, as set out in the opening chapter of Ayatollah Sistani's Islamic Laws. The structure is personal and living by design: a ruling is followed because of who issued it and that they can be asked again.

Software cannot meet those conditions, so an AI's answer on a fiqh question has the standing of a reference lookup: potentially useful for finding what a marja' has said, with no authority of its own. The same reasoning covers Islamic finance questions, where Iran's capital-market Shariah committee is likewise a body of named jurists.

Different institutions, different chains of authority, the same operational rule: an agent may retrieve, explain and apply; a human rules.

The five-question test

Ask these of any AI answer on a religious or Islamic finance question. One minute. Any "no" means do not act on the answer without a human.

1Does it show the source?yes / no2Does it name the school?yes / no3Does it date the ruling?yes / no4Does it say when it is unsure?yes / no5Does it send you to a human?yes / no
Print it. Stick it next to the screen. It works on every tool, including ours.
  1. Does it show the source? A link or citation to a named scholar, institution or standard. Not "scholars say".
  2. Does it name the school? Hanafi, Shafi'i, Maliki, Hanbali, Ja'fari, or a named institution's methodology. Rulings differ; an answer that hides which one it is giving is hiding something.
  3. Does it date the ruling? Positions change. AAOIFI's 2008 sukuk resolution changed a market. A dateless answer may be decades stale.
  4. Does it say when it is unsure? A tool that never declines is a tool that guesses.
  5. Does it send you to a human? For anything with consequences, a good agent ends with "ask a qualified scholar of your school", and means it.

The long form, with examples of good and bad answers, is at how to check an AI Islamic answer.

Where agents are genuinely useful

  • Finding what a scholar said. A grounded assistant over a library of fatwas is a better index than any search engine, provided it quotes rather than paraphrases.
  • Applying published rules at scale. Screening thousands of stocks against ratio thresholds and business-activity exclusions written by scholars is exactly the repetitive, rule-bound work software does well. This is what HalalScreener does for equities, ETFs and crypto. It applies rules; it does not write them, and it does not certify.
  • Calculations. Zakat on mixed assets, purification of dividends, prayer times. Arithmetic with inputs you can inspect.
  • Monitoring. Watching for a company's debt ratio crossing a threshold, or a ruling being updated, and alerting you.

In each case the agent is doing the work a careful assistant would do, and the authority stays where it was.

How agents call tools, and why you should care

Agents act by calling tools: a fatwa database, a screening service, a calculator. In November 2024 Anthropic published the Model Context Protocol, an open standard that lets an AI system connect to tools through one interface and record each call. The engineering detail matters to you for one reason: a logged tool call can be audited. "The agent checked this stock against ruleset version 4.2 on 12 October and got this answer" is a sentence an auditor, a scholar or you can verify. "The AI said it was fine" is not. Our page on AI agents and halal investing walks through what a well-built agent does with your portfolio and what it refuses to do.

What this site is for

Three jobs.

  1. Explain. What these tools are, how they work, and what the two major traditions have said about them, in language anyone can follow.
  2. Set a standard. The Shariah Agent Standard: seven commitments for anyone building AI that touches Islamic law or finance. Open, attributable, and applied first to the tools of the team that wrote it.
  3. Keep score. As institutions publish positions and builders publish products, we record them here with dates and sources.

We are not scholars and this site is not a fatwa service. We build software that applies scholars' rules and we think the public deserves a plain account of where software's authority ends. If you want a note when something on this site changes, the form at the bottom is the way.

SharePost on XShare on LinkedInSend on WhatsAppSend on TelegramReport a correction

Questions people ask

What is a Shariah agent?
An AI system designed to answer or act on questions of Islamic law and Islamic finance: looking up rulings, explaining them, screening investments against published rules, or running calculations such as zakat and purification. The term covers a spectrum from a search assistant to a tool-using agent. It does not include a human scholar, and it should never be mistaken for one.
Can a Shariah agent give me a fatwa?
No. A fatwa is a ruling issued by a qualified, accountable human scholar. Egypt's Dar al-Ifta ruled in fatwa 22255 that relying on AI applications to obtain a fatwa is impermissible. In the Ja'fari school, rulings are followed from a qualified living jurist. An agent can show you what scholars have said; it cannot be one of them.
Are AI Islamic assistants reliable?
It depends entirely on how they work. Chatbots answering from memory have been found in academic testing to drift toward permissive answers and to rely on secondary sources. Grounded assistants that retrieve and quote a named scholarly source are far more checkable. The five-question test on this page tells you which kind you are using.
What is the difference between a chatbot and an agent?
A chatbot answers questions. An agent takes a goal and executes steps using tools, for example screening every stock in your portfolio against a ruleset and alerting you to changes. Agents are more useful and need stricter limits, because they can act on money.
Who runs this site?
The research team at HalalScreener, a Shariah stock-screening service that also runs a grounded question-answering tool called Scholar. We disclose that on every page. The standard we publish applies to our own tools as much as to anyone else's.

Sources

Every figure and ruling on this page links to where it came from. Check them.

  1. Using AI applications to obtain fatwas (fatwa 22255), Dar al-Ifta al-Misriyyah
  2. Egypt's Dar al-Ifta rules AI-based Quran interpretation impermissible, MEA Tech Watch, 28 January 2026
  3. Islamic Laws, Chapter One: Following a Jurist (Taqlid), Office of Ayatollah Sistani
  4. Islamic chatbots in the age of large language models, arXiv (presented at the Muslims in ML workshop, NeurIPS 2025), 31 December 2025
  5. Digital authority and Islamic legal reasoning: a comparative study of AI-generated and human-issued fatawa, University of Malaya
  6. Chat GPMufti, Center for Strategic and International Studies
  7. Gartner identifies the top 10 strategic technology trends for 2025, Gartner, 21 October 2024
  8. Introducing the Model Context Protocol, Anthropic, 25 November 2024

All guides

GuideCan AI issue a fatwa?No. Dar al-Ifta's fatwa 22255, the conditions of taqlid in the Ja'fari school, and the research on AI fatwas, with what AI may still do.Updated 12 October 2026 · 5 min readGuideHow to check an AI answerFive questions to ask any AI answer about halal, haram or Islamic finance, with examples that pass and fail and what to do with each score.Updated 12 October 2026 · 4 min readComparisonAI vs scholarAI Shariah assistants and human scholars compared across ten dimensions: authority, accountability, context, consistency, speed and failure modes.Updated 12 October 2026 · 4 min readGuideGrounded AIWhat retrieval-grounded AI means for Islamic questions, why it is the minimum acceptable design, its failure modes, and a disclosed worked example.Updated 12 October 2026 · 5 min readGuideAI agents and halal investingScreening at scale, purification maths, verdict alerts and MCP tool calls: what a well-built halal-investing agent does, and what it must refuse.Updated 12 October 2026 · 5 min readReferenceThe Shariah Agent StandardSeven open commitments for AI that touches Islamic law or finance: grounding, school, dating, decline, hand-off, logging, accountability. Version 1.0.Updated 12 October 2026 · 5 min readReferenceGlossaryShort definitions of the AI and Islamic-law vocabulary used on this site: agent, fatwa, fiqh, grounding, marja', MCP, mufti, taqlid and more.Updated 12 October 2026 · 3 min read

Be first when this becomes real

A short note when something changes: a new ruling, a live issuance, a tool worth testing. No weekly filler.