No AI in religious content — the line we hold

Most apps are coy about AI. Sakinah shouldn't be. If you are going to read Qur'an, hadith, tafsir, or du'a inside this app, you deserve to know exactly which parts of that experience were written by a person and which were touched by a machine. So here it is, plainly.

What AI is not allowed to do in Sakinah

No religious text in the app is generated, paraphrased, summarised, translated, or re-ordered by AI. Specifically:

• Qur'anic Arabic — comes from Quran.Foundation's Uthmani text. It is not retokenised, corrected, or regenerated.
• Qur'an translations — come from named translators (Saheeh International, Mufti Taqi Usmani, and others selectable in Settings). AI does not edit them.
• Tafsir — six named works: Ibn Kathir, Ma'arif al-Qur'an, Tazkirul Qur'an, al-Baghawi, al-Sa'di, al-Muyassar. They are rendered verbatim from their source data.
• Hadith matn and isnad — fetched from Sunnah.com data, preserved word-for-word. Arabic sanad is never "cleaned up" or condensed.
• Gradings — shown as the source attributes them. We never re-grade a hadith or move it between collections.
• Du'a — pulled from Hisn al-Muslim, rendered as the compiler published it.

What AI is allowed to do

AI tools are used strictly for engineering tasks that do not touch the text you read:

• Code review and refactoring suggestions
• Writing tests and build scripts
• Catching TypeScript errors and accessibility issues
• Drafting internal documentation

Think of it as a junior engineer who is allowed near the plumbing but never near the Mushaf.

Why this line matters

Religious text is not like product copy. A hallucinated hadith, a misattributed grading, or an AI-paraphrased verse carries weight far beyond a bad user experience. People act on what they read in a Muslim app: they pray with it, they memorise it, they quote it.

Large language models are fluent. They are not scholars, and they cannot be held to a chain of transmission. Using one to produce scripture-adjacent text — even as a "light edit" — is how bida' quietly enters a product. So the line is drawn where it has to be drawn: the scholarly source, the named translator, the classical mufassir. Nothing regenerated in between.

What about emotion guidance and action steps?

The emotion definitions and the short action steps shown alongside verses are written by hand. They are grounded in classical works — Ihya Ulum al-Din by al-Ghazali, Madarij al-Salikin by Ibn al-Qayyim, Riyad as-Salihin by al-Nawawi — and they are attributed where attribution makes sense. They are not fatwa. They are reflection prompts. The app frames them that way in every surface.

If the line is ever crossed

If you ever see text in the app that looks paraphrased, regenerated, or otherwise off from its scholarly source, please tell us: arahim4956@gmail.com. The app will be corrected, and if it reveals a process gap, the process will be fixed too.