How Rhemata Handles Sources

Transparency about where our content comes from, how it’s used, and how we protect the people who wrote it.

Rhemata is not trained on this content

Rhemata does not train an AI model on the books, sermons, and commentaries in its library. There is no model that has absorbed these works and learned to imitate them.

Instead, Rhemata is a retrieval and citation system. When you ask a question, Rhemata searches its indexed library for relevant passages, retrieves them, and presents an answer built on those passages — with every claim traced back to a named author and source. The original works remain intact, attributed, and pointed back to. Nothing is absorbed, blended, or anonymized into a model.

This distinction is the foundation of everything we do. AI tools that flatten a thousand voices into one synthetic answer erase the people who did the work. Rhemata exists to do the opposite: put named voices in front of you, with receipts.

Every claim is attributed, never invented

Rhemata does not put words in a teacher’s mouth. Every claim in an answer is a paraphrase of what a named source actually teaches — labeled as summary, never presented as the teacher’s own words — and it is always attributed by name, with a link back to where the real teaching lives. No AI-invented citations. No anonymous synthesis presented as someone’s words.

Verified, character-for-character verbatim quoting — software-confirmed against the source before a quote can ever be served — is on our roadmap. It is not live yet. Until it ships, Rhemata paraphrases and cites; it does not quote freely.

How content enters the library

Every source in Rhemata passes through the same pipeline:

  1. Classification. Before anything is served, each source is assigned a rights status: public domain, owned, licensed, or unverified.
  2. Hidden by default. New content is not visible to users until its rights status has been reviewed. Our system fails closed — if we haven’t cleared it, you can’t retrieve it.
  3. Attribution at ingest. Every document is tied to its author or rights-holder at the moment it enters the system, so attribution can never be lost downstream.
  4. Theological review. Featured voices are selected against our theological foundations (see Our Theological Lens) — not scraped indiscriminately.

What we prioritize

The heart of Rhemata’s library is public domain literature — the writers of past centuries whose works belong to everyone: R.A. Torrey, Andrew Murray, Charles Finney, the voices of the early Pentecostal outpouring, and many more. These works are freely quotable, and Rhemata’s mission is to put them back in front of a generation that has never read them.

For contemporary voices, we pursue direct permission from authors, estates, and ministries. Where content carries open licenses (such as Creative Commons), we honor the license terms, including attribution requirements.

Our interlinear and lexicon word-study data carries this required attribution: “Data created by www.STEPBible.org based on work at Tyndale House Cambridge (CC BY 4.0).”

For authors and rights holders

If you are an author, publisher, or rights holder and believe your content appears in Rhemata improperly — or if you’d like to discuss how your work is represented — contact us directly:

[contact email]

We respond to every rights inquiry. If content is identified as improperly included, we will remove it from retrieval promptly while the matter is reviewed. We would rather lose a source than misuse one.

Why we built it this way

Rhemata’s audience carries a healthy suspicion of AI — and we share it. An AI that invents quotes, launders sources, or speaks as an anonymous oracle is not a tool for serious theological study. Every safeguard on this page exists for one reason: so that when Rhemata shows you what Torrey or Murray said, you can trust that they actually said it — and go read them yourself.

Rhemata is not a substitute for the sources. It is a doorway to them.