ILMINAR

FAQ

Things worth knowing

The honest answers to the questions people actually ask before trusting a cited ruling.

No. Ilminar retrieves what scholars have said in their lectures. It does not issue rulings. Every answer is grounded in retrieved transcript text and links directly to the moment the scholar spoke. If no relevant passage is found, Ilminar says so rather than generating a response.

Yes, in three ways. It can retrieve a passage that isn't quite the one you needed. Its summary can drop a condition or qualification the scholar attached. And the transcript underneath can mishear a term, particularly an Arabic one, which can pull in the wrong passage. What it cannot do is invent a scholar's position out of nothing, because every answer is tied to a real passage in a real recording. Every answer carries a timestamp for exactly this reason: if it matters, listen to the clip before you act on it. If you find an answer that misrepresents what a scholar said, send it to us through the Help page.

None of them. Ilminar returns what the speaker in the recording actually taught, so the position you get is theirs. Some of them specialise in a particular madhhab; many teach from the perspective of all four schools and the majority position rather than one. Every answer names who said it, so you always know whose view you're reading, and you can narrow the question yourself by asking about a specific school.

Several scholars and teachers across different backgrounds and schools of thought, with more added every week. Each new lecturer's lectures are run through Ilminar's transcription pipeline before they're indexed.

One person: Ilminar is built by Awab Akram. Nothing goes in anonymously or automatically. Every lecture comes from a named speaker and a published recording you can go and watch in full, and every answer tells you which speaker and which lecture it came from. Ilminar doesn't host anyone's video or audio. It stores what was said and links back to the original recording on their own channel.

Fiqh has always had legitimate differences of opinion, between schools of thought and between individual scholars. Ilminar doesn't try to resolve that or hand down one "correct" answer. It surfaces what the speakers in the corpus actually said and who said it, so you can act on guidance from someone you already trust rather than a flattened, opinion-less average. Deciding between those positions isn't something an AI should do.

No. Ilminar is a search tool for lectures, not a substitute for a living scholar who knows you and your circumstances. It has no idea who you are, what you have already done, or what makes your situation particular, and fiqh answers often turn on exactly that. Use it to find what a scholar taught on a topic, to check something you half-remember from a lecture, or to walk into a conversation with your imam already knowing where the question sits. For anything consequential or specific to you, ask a person.

Ilminar runs on Retrieval-Augmented Generation (RAG), not a general model answering from memory. It first finds the actual passage where a scholar addressed your question, then writes a short summary of that passage and nothing else, with a link to the exact moment in the recording. Lectures are processed through Ilminar's own transcription pipeline, built for this kind of audio and more accurate than generic auto-generated transcripts. The summary is still written by AI, so it can be clumsy or miss a qualification the scholar made. That is what the timestamp is for. The reason to trust an answer isn't the architecture, it's that you can tap the link and hear the scholar say it yourself, in their own words, every time. Treat the summary as a pointer to the source, not as the source.

Citations link to the exact YouTube timestamp where the scholar addressed the topic, so you can tap and hear them say it yourself. The transcript is used for retrieval; the linked clip is the primary source.

Ilminar tells you clearly that no matching passage was found in the lectures it has indexed. It does not fabricate an answer. This is by design: a missing answer is more trustworthy than a made-up one.

They're saved to your account so you can return to the conversation later, and they're processed by the services that answer them: OpenAI for the answer, Voyage AI for search. We don't sell them, don't use them for advertising, and don't build a profile from them. We don't use your questions to train any model ourselves. You can ask for your account and its data to be deleted at any time through the Help page.

Partly. Arabic terms in your question are understood, transliterated (salah, wudu, zakat) or in Arabic script, and some indexed lectures contain Arabic sentences transcribed in Arabic script. But every lecture indexed so far is taught mainly in English, and asking a question entirely in Arabic isn't something we've tested yet. English works best today. Indexing scholars who teach in Arabic, and full support for asking in Arabic, are on the roadmap rather than done.

Free accounts get a monthly question allowance that resets automatically. How large that allowance is is still being experimented with during beta, so it may move up or down rather than being a fixed promise. Pro lifts the cap for £9.99 a month.

No. It's in beta and built by one person. Things will change, occasionally break, and get better. What doesn't change is the standard: every answer cites its source and links to the moment it was said, and the AI never issues a ruling of its own.

Still stuck?

If your question isn't answered here, send it directly and we'll get back to you.

Contact us