How we make the notes.

The reader includes five translations: the Berean Standard Bible (BSB), the King James Version (KJV), the World English Bible (WEB), the Bible in Basic English (BBE), and the American Standard Version (ASV). Search covers three of them: Berean Standard, King James, and World English. Basic English and American Standard readers search the King James words. The Berean Standard Bible, the World English Bible, the Bible in Basic English, and the American Standard Version are in the public domain. The King James Version is in the public domain in the United States. The text is checked against pinned copies of the source files. AI never writes or changes it.

Everything else is commentary: chapter context, introductions to people, story moments, reflection questions, journey notes, the year plan's section notes, the notes on Traces links, Timeline eras and places, and portraits. Each is labeled as AI-made and kept apart from the Bible text. Notes link the verses they rest on, and most carry the line AI-drafted and source-checked. Not yet reviewed by a person.

All 1,189 chapters have General, Historical and Literary context. 249 people have an introduction and their story told in a few moments: the 200 people Naravist is built around, each with a portrait, and 49 more. The people index lists all 3,054 people named in Scripture, most with their passages only.

60 guided journeys have a short note at each of their 560 stops, and every stop adds a reflection question. Traces has 388 reviewed links between 181 of those 200 people. The Timeline sets them in 9 eras, and Places maps 177 places.

The chapter context and introductions were drafted in September 2026 from packets of the King James text: 460 with OpenAI Codex and 983 with Anthropic Claude. The journey notes and questions, the year plan, Traces, the Timeline and Places were drafted with Anthropic Claude. Drafting happened before publication. Nothing is generated while you read.

The chapter context, introductions and journey notes work from the King James text alone. Historical context describes the setting the Bible itself gives. The notes do not add word meanings the text does not state, dates, outside history or archaeology.

These checks run automatically every time content is published:

• Every section must cite at least one verse, and each quoted piece of evidence must match the King James text of that verse word for word.
• A section fails if it opens in the first person, so no note speaks as a Bible character.
• A section fails if it uses stock claims of certainty, such as "all scholars agree" or "infallible".
• Each published note keeps a record of its source version, who drafted and reviewed it, the outcome, and fingerprints of its text and evidence. If the note or its evidence changes, it fails until it is reviewed again.
• Words in quotation marks inside a chapter or people note must be the wording of the King James, American Standard or Basic English text. Capital letters and the punctuation around a quotation are ignored. A pattern such as "As ... so" is checked piece by piece.
• Journey notes and questions must quote the King James text of their stop word for word and pass their own format checks.

After the automated checks, each kind of content had an AI review pass:

• Chapter context and introductions. Each note had a review pass that read every section against its cited verses, recorded whether they support it and why, and looked for unsupported claims, contradictions, mistaken identities and overstatement. For 1,425 of the 1,443 notes, the same AI session that drafted the note then reviewed it. The other 18 were reviewed by a second agent of the same model.
• Journeys. Each journey was drafted by a Claude agent. A second AI agent, one that had not drafted that journey, then reviewed it. It checked every claim in the stop notes, reflection questions and closing against the passage and the verses it cites, cut anything drawn from outside the King James text, such as word meanings, dates or outside history, and narrowed what the verses did not support.
• Traces. Each batch of links was drafted by a Claude agent. A second AI agent, one that had not drafted that batch, then reviewed it. It read the King James text of every evidence verse and of every verse where the two stories cross, and checked that each link is the right kind, points the right way and says no more than the verses.
• Timeline and Places. The Timeline and the places were drafted by Claude agents. A second AI agent, one that had not drafted them, then reviewed them. It checked that each "same time" claim is stated by a verse, that each era's dates follow the usual traditional ranges and say they are debated, and that each place's certainty label and visits hold up against the verses.
• The year plan. The same AI session that drafted the plan then reviewed it. It checked that the chapters, days and people add up, and that no section note claims an author or identity the text does not give.

Every review pass was made by the same AI model that drafted the content, even when it ran as a second agent. So none of it is an independent review, and none of it is a human review.

A note that fails a check is not published. So far every published note passed, and none has been held back.

The portraits are AI-generated artistic interpretations, made from written briefs based on the passages. They are not evidence of how anyone looked.

No scholar, pastor or editor has reviewed the notes, journeys or portraits. We plan to have a person review the introductions, the journeys and a sample of chapter notes, and this page will say when that has happened. Until then, treat the notes as a starting point. They describe what the passages say, they are not the last word on them, and traditions read some passages differently.

Open the verse and choose Report a problem at the bottom of the page. Your report comes straight to us, and we check it against the source text before changing anything. You can also write to help@naravist.com. Read where the Scripture text and people index come from.