Our standards
How we handle Scripture, AI, and mistakes
This is what we actually do — including what we cannot do yet. You can hold us to every line.
How Scripture is quoted
- Scripture is quoted word for word from licensed translations — never rewritten by AI.
- Every page names the translation it uses, so you always know which version you are reading.
- Cross-references come from the public Treasury of Scripture Knowledge, not from the model's memory.
How AI answers show their sources
AI is a reading companion, not an authority. So every answer has a "Sources" panel that lays out what it rests on:
- Key conclusions are labeled in three kinds: Scripture says (the text itself says it), Tradition (a reading held by a named tradition), and Inference (the AI's own synthesis or application).
- Every verse an answer cites is looked up in our Bible text before it is shown. References we cannot find are dropped — you never see them.
- A conclusion marked "Scripture says" that cannot point to a single verifiable verse is relabeled "Inference". If we say the text says it, we must show the text.
- Verse references written in the answer text itself (like "John 3:16") are checked too; any we cannot find are flagged "Couldn't verify" in the Sources panel.
- On a passage page, sources outside that passage and its cross-references are marked "AI-added".
- Where a question is debated, or the text itself cannot settle it, the answer says so in a separate "Open question" note.
Where our content comes from
Different content goes through different processes. Each page's byline says honestly which one it is — we never borrow a real person's name:
Reflections, topics, space-time pages
AI-assisted, human-reviewedDrafted by our content pipeline; a person checks facts, references, and tradition before it goes live.
Film guides
AI-assisted, human-reviewedWritten only for hand-picked films; drafted with AI and read by a person before publishing.
Book guides
AI-generated, review in progressGenerated by AI and being reviewed by a person, book by book. A guide that has been reviewed is signed "AI-assisted, human-reviewed"; one that hasn't carries no byline.
Chapter, original-language word, and question pages
Machine-checkedGenerated by AI with no human review; instead, hard machine checks: cross-references may only come from TSK data, original-language words only from Strong's data, quoted verses are matched word for word against the text, and near-duplicate pages are blocked.
AI answers
Live, verses checkedGenerated live with no human in the loop; verses are checked under the rules in "How AI answers show their sources" above.
Where traditions (Protestant, Catholic, Jewish, and others) read a passage differently, we label whose reading it is rather than ruling on who is right.
When something is wrong
- Email us the page and the sentence that is wrong (the "Report an error" link next to each byline fills in the page address for you).
- We check it against the text and our sources. If it is wrong, we correct the page and its "updated" date changes with it.
- If the mistake comes from a generation rule (say, how AI answers cite sources), we fix the rule, not just the page.
- Significant corrections are recorded in the public changelog.
What we can't promise yet
- AI still makes mistakes. Checking sources guarantees that a verse exists and is quoted correctly — not that the interpretation is right.
- Not every book guide has been reviewed by a person yet; the ones without a byline are those still waiting.
- We do not publish model leaderboards or accuracy scores — numbers from a test we wrote and graded ourselves are hard to trust. We would rather publish the rules and let you check them one by one.