Guides

Should you ask an AI chatbot about Islam?

A chatbot that writes its own answers can produce an āyah that sounds right and isn't. What researchers and Egypt's Dar al-Ifta have said, three questions to ask before trusting an answer, and where an assistant genuinely helps.

Muhammad Hasnain, founder and lead engineer of Nainnie Labs
Muhammad Hasnain6 min read

On this page

Ask a chatbot for "the āyah about patience in hardship" and it will give you something. It will be fluent, it will be in Arabic with a translation beneath, and it will have a surah name attached. There is a real chance some of it is wrong in a way you cannot see.

That is not a reason never to use one. It is a reason to know what kind of tool you are holding.

What a chatbot is doing when it answers

A general-purpose chatbot is built on a large language model. It has read an enormous amount of text and learned what tends to come next. When you ask it something, it writes the most plausible answer it can, one word after another.

Plausible and correct overlap a great deal of the time, which is why these tools are useful. But they are not the same thing, and the model has no separate step where it checks a quotation against the original. If the most likely-sounding āyah is a near-miss — the right theme, the wrong words, a reference one number off — it will be delivered with exactly the same confidence as a perfect one.

For most subjects, a near-miss costs you a little accuracy. For scripture, it costs you the one thing that mattered.

Why scripture is the worst case

A wrong date in a history answer is a mistake. A wrong word in an āyah is something else: text presented as the speech of Allah that is not.

A December 2025 paper, Islamic Chatbots in the Age of Large Language Models by Muhammad Aurangzeb Ahmad of the University of Washington Bothell, presented at the Muslims in ML workshop at NeurIPS 2025, surveys around twenty Islamic chatbots and sets out the risks plainly. Models trained on broad, uncurated text, it notes, can hallucinate Quranic verses, misquote prophetic traditions, and conflate weak narrations with authentic ones.

It makes a second point that is easy to miss. A chatbot's tone does not change with its certainty. A fabricated quotation, a weak hadith and a question on which scholars genuinely disagree can all come out in the same calm, finished prose. The reader is given no signal about which is which.

What scholars and researchers have said

The clearest institutional statement so far came from Egypt. According to a report from February 2026, Dar al-Ifta al-Misriyyah issued a fatwa the previous month holding that using artificial intelligence applications, including ChatGPT, to interpret the Quran is impermissible. The reasons it gave were that these systems lack real comprehension of the text, can produce factual errors and unverified material, and can generate interpretations that conflict with established teaching — without scholarly oversight or any way to verify the result.

Note what that ruling is about: interpretation. It is not a ruling on whether you may use a phone to look up when Maghrib is.

The research paper, for its part, does not argue that chatbots should be abandoned. It recommends two things that matter here. Answers should come with citations, so a reader can trace a response back to the canonical text — and where there is no citation, that should be said, not hidden behind fluent writing. And a system should be clear about its scope, distinguishing educational help from anything resembling a fatwa.

Three questions before you trust an answer

Whichever tool you use, these three take a few seconds and catch most of the danger.

1. Can it show me the source? An āyah should come with its surah and number. A hadith should come with its collection and number. If you get a quotation with no reference, or a reference you cannot find, treat the quotation as unverified. Check āyāt in a printed muṣḥaf or on quran.com, and hadith on sunnah.com.

2. Is the text quoted, or composed? There is a difference between a tool that retrieves the exact words of an āyah from a fixed copy of the Quran and one that writes out what it believes the āyah says. You cannot always tell by looking. A tool that tells you where its text comes from is usually the first kind.

3. Does it tell me what it will not answer? A good tool knows its edges. Asked whether something is halal, or what an āyah means for your situation, the honest answer from software is "ask a scholar". A tool that answers everything with equal confidence is not being more helpful. It is hiding the line it should be showing you.

Where an assistant does help

None of this means the idea is wrong. A great deal of what Muslims look up every day is finding, not interpreting:

  • the duʿā for anxiety, for parents, before sleep
  • Āyat al-Kursī, or the last two āyāt of Al-Baqarah
  • the first three āyāt of Surah Al-Mulk
  • when Maghrib is, and which way the qibla faces
  • how far you have read, and what today's portion is

For these, there is one right answer and it already exists in a fixed text. A tool that understands the question and hands back the exact text — with its reference — saves real time and introduces no risk of invention. The danger begins only when a tool starts writing the answer itself.

A note on the app I build

Disclosed interest: Qurbah is mine, and its assistant, Ask Qurbah, was designed around everything above.

A small model on the phone works out what you are asking — in Urdu, Roman Urdu, English, Arabic, French, Indonesian or Turkish — and the answer is then taken word for word from the verified library inside the app, or from your own prayer record. Nothing is generated. Each āyah comes with its surah and number, each duʿā with its reference, and the Quran translation is Kanz-ul-Iman, in Urdu and in English.

Ask it what something means, or whether something is halal, and it does not try. It says it can find an āyah or a duʿā, or help with your own prayers, and that meanings and rulings are for a scholar. If it cannot find what you asked for, it says that too, rather than offering its best guess.

It runs entirely on the phone, works with no connection, and if you ask by voice, your speech is recognised on the device and never sent to a server. It is free, with no ads.

It also does much less than a general chatbot. That is the point.

Frequently asked questions

Can I use ChatGPT to understand the Quran?

Egypt's Dar al-Ifta ruled in early 2026, according to press reports, that relying on AI tools such as ChatGPT to interpret the Quran is impermissible, citing the risk of errors, unverified material and interpretations that conflict with established teaching. Other scholars and bodies may phrase it differently, but the concern is consistent: interpretation needs knowledge, context and accountability that a text generator does not have. Ask a scholar.

Can an AI chatbot make up a Quran verse or a hadith?

Yes. A large language model writes the most plausible continuation of what it has read, and it has no built-in way to check that a quotation is real. A 2025 paper on Islamic chatbots from the University of Washington Bothell notes that such models can hallucinate Quranic verses, misquote prophetic traditions and blur weak narrations with authentic ones.

Can an AI give a fatwa?

A chatbot can produce text that looks like one, which is the problem. A fatwa depends on the person asking, their circumstances and a scholar who answers for what they say. A system that cannot be held responsible, and cannot tell you what it does not know, is not a substitute. Use it, at most, to find the text you want to ask a scholar about.

How do I check whether an āyah a chatbot gave me is real?

Ask for the surah and āyah number, then read that āyah in a printed muṣḥaf or a trusted Quran site such as quran.com. If the chatbot cannot give a number, or the number does not match the words, treat the quotation as unverified. Do the same for hadith: check the collection and number on sunnah.com.

Is any Islamic assistant safe to use?

The ones built to quote rather than compose are much safer for finding text. If an assistant shows you where each āyah and duʿā comes from, says plainly what it will not answer, and sends questions of meaning and law to a scholar, it is doing the part a tool can do well and leaving the rest alone.

More from the blog