Our Commitment to Accuracy
Scripture accuracy is the highest priority in this project, ahead of features and convenience. This page explains exactly how that is enforced, so you can judge it rather than take our word for it.
Our promise to you
Scripture in WordForge is never written by an AI. It is fetched by reference, word for word, from the Berean Standard Bible, and quoted exactly as it stands.
- ◆No paraphrase. A verse appears word for word, or it is cited without being quoted.
- ◆No guessing. If the text cannot be verified, no verse is quoted in that answer at all.
- ◆No hallucination. Verse wording never comes from the model. It comes from the translation itself.
- ◆Never quoted from memory. This is prevented by how the app is built, not merely by asking the model nicely.
- ◆Never edited afterwards. Nothing rewrites generated text, so nothing can quietly alter a verse.
Two honest exceptions
Greek and Hebrew are not fetched. The translation this app uses publishes English only, so when the original-language tool gives you a Greek or Hebrew word, its pronunciation, or a Strong’s number, that comes from the model’s own knowledge and is not verified the way the English is. The tool says so in its own results, and the same applies to the “Key Words” sections of studies and sermon outlines. Check anything that matters to you against an interlinear or lexicon.
Exporting converts formatting. Saving to Google Docs turns markdown into a document, which means headings and bold text are rewritten as formatting. Quoted text is deliberately excluded from every one of those passes, so a verse is carried across word for word, but it would be wrong to say nothing is touched at all. What you read on screen is never altered.
This holds everywhere, with no exceptions: studies, breakdowns, topical answers, sermon outlines, original-language lookups, kids studies, prayer helpers, and saved content alike.
The translation: Berean Standard Bible
Every verse you read here is the Berean Standard Bible. It reads in modern English and was placed in the public domain on 30 April 2023, which means it can be used freely, including here, without licensing fees that would otherwise have to be passed on to you.
There is no translation selector. One trustworthy text, used consistently, is better than a menu of options where a study quotes one translation and the reader shows another.
How AI is prevented from misquoting Scripture
AI models paraphrase Scripture when they quote from memory. They are confident and fluent while doing it, which makes it hard to spot. This app never lets them try.
When you generate a study, it happens in two passes. First the model is asked only which verses it intends to cite, and it returns references, never text. Then those verses are fetched from the Berean Standard Bible, and wherever a quotation belongs the model writes a placeholder naming the reference. The fetched text is inserted into that placeholder after the model finishes writing, so the words you read never passed through the model at all.
So the model decides what is worth quoting. It never supplies, and never even carries, the words. If verse text cannot be verified at all, it is told it may not quote anything, and must discuss passages in its own words instead.
Nothing rewrites a verse after the fact, either. Automatic “tidying” of generated text is deliberately not used, because any rule that edits output can eventually edit Scripture.
The words of Jesus, shown in red
In the Bible reader, verses containing words spoken by Jesus are shown in red and marked with a red rule down the left, so the distinction does not depend on being able to see colour.
One limitation, stated plainly
The red is applied to the whole verse, not to individual words. So red here means “Jesus speaks somewhere in this verse,” not “every word here was spoken by Jesus.” Where a verse begins with narration, that narration turns red too. Matthew 4:4 is a clear example:
But Jesus answered, “It is written: ‘Man shall not live on bread alone...’”
The words “But Jesus answered” are Matthew writing, not Jesus speaking, yet the reader colours them red along with the rest of the verse.
This is not fixed because fixing it would require guessing. The Berean Standard Bible does not publish markers for where a quotation starts and stops inside a verse, and the only translation that does mark them worded its sentences differently, so its boundaries cannot be laid over this text. Working them out by following the quotation marks would be right most of the time and confidently wrong the rest, which is worse than a limitation you can see and account for. Marking a whole verse claims less than pretending to know exactly where the red should stop.
That list covers 2,019 verses. It was originally derived from the CrossWire Bible Society’s markup of the King James Version, then reconciled to the Berean Standard Bible, because markers taken from one translation cannot simply be trusted on another. Two checks run against it:
- ◆Every listed verse must actually exist in this translation. Eight did not and were removed.
- ◆Every listed verse is cross-checked against the World English Bible, a separate translation whose editors marked Jesus’s words independently. All 2,019 are confirmed by that second source.
One verse was added: Mark 10:49, where this translation has Jesus speaking directly (“Call him”) while the King James Version reports it indirectly, so the original markup had missed it.
Where editions honestly differ
Red-letter Bibles do not all agree, and it would be misleading to pretend otherwise. The World English Bible marks 40 verses that this app does not. Seven of those simply do not exist in this translation. The remaining 33 fall into three groups: verses where someone else repeats or recalls Jesus’s words, the voice speaking to Peter in his vision in Acts, and passages in Revelation where the speaker is described as the Lord God or the one seated on the throne.
Those verses are left unmarked here, deliberately. Colouring a verse red asserts that Jesus said it, and where careful editors disagree, the safer error is to leave it in black.
What is verified, and what is not
Checked automatically, every time this app changes: verse text, character for character; the red-letter list; every quotation that appears inside any prose anywhere in the app, which must match the source translation exactly or be declared as ordinary speech on a reviewed list; and every citation, which must point at a chapter and verse that actually exist. That covers the dictionary, the character and location pages, the timeline, the quiz, and every pre-written study, including all of the children’s studies.
Audited by people, with corrections: the factual claims in the reference pages were fact-checked against the Bible’s own text, and the errors that were found were fixed. Timeline dates that could not be supported were removed rather than kept, and the era datings that remain show exactly what they rest on. The children’s studies were reviewed the same way, before any child saw them.
What no machine can verify: interpretation and application, what a passage means and how to live it, and historical background drawn from standard reference works rather than from the Bible itself. That is true of every study Bible ever printed, and it is true here. Treat the surrounding material as a helpful study aid, not as authority, and let Scripture itself settle anything that matters.
If you ever find a verse quoted incorrectly anywhere in this app, that is a bug and worth reporting. It should not be possible.