שShema

How We Use AI

Effective October 24, 2026

Shema uses artificial intelligence to transcribe sermons and to help you study them. That is a serious thing to do with someone's faith, so here are the rules we hold ourselves to.

AI serves the text. It never speaks for God.

AI can do useful, ordinary work: turn speech into text, translate a language, summarize a message you heard, surface historical and cultural background, suggest a passage to read next. What it cannot do is add to Scripture, replace the Holy Spirit, or tell you what God is saying to you. We build Shema with that line firmly drawn.

Test everything

Every AI-generated summary, study note, devotional, and answer in Shema is a study aid — not Scripture and not a substitute for your own reading. Treat it the way you would treat a study guide written by a stranger: useful, fallible, and worth checking. Read the passage yourself. Bring questions to your pastor and your church.

AI can be wrong

We do not claim the AI in Shema is always accurate. It can mishear a word, miss context, or state something confidently that is simply mistaken. We label AI-generated content as AI-generated, we show you the passage it drew from wherever we can, and we would rather admit this plainly than pretend otherwise.

Your content is yours

To do this work, Shema sends the audio, transcripts, and notes you choose to submit, what you write in your prayer journal, the questions you ask about a sermon or a passage, and how you answer the devotional's daily question, to OpenAI — it does the transcribing, and writes the summaries, devotionals, and study material. Narration is split: when Shema reads something aloud, the text to be spoken goes to either ElevenLabs or OpenAI, depending on the voice. And if you import a YouTube link, that link goes to Supadata, which sends back the video's captions. Your content is not used to train AI models. We ask for your consent before any of it happens, and you can delete any recording, note, or your whole account at any time. The full detail is in our Privacy Policy.

What we will not build

Human oversight

We review the prompts and behavior behind Shema's AI features, and we correct them when they drift. This is ongoing work rather than a finished guarantee, and we would rather say that honestly than overstate it. If you find AI output in Shema that is wrong or theologically off, please tell us at support@shemaai.com — reports from users are the most reliable check we have.


What we believe →

Shema · support@shemaai.com