When the chatbot quotes the Bible
Ask a chatbot for a verse about patience and it will answer in a second, fluently, with a reference attached. It sounds like Scripture. It may not be. That is the problem Bobby Gruenewald, founder and CEO of YouVersion — the Bible App, with more than a billion downloads — put in front of the church this month. Speaking to Christianity Today on 16 November, he said the misquoting of the Bible by AI chatbots “becomes a really big generational problem.” His own estimate, from a March interview: even the best models misquote Scripture at least 15% of the time, and some as much as 60%. And the exposure is not marginal — new data cited in the interview found that half of the Christian questions sent to ChatGPT are about interpreting the Bible.
The people who know the Bible best are the most careful with it. YouVersion built a Bible chatbot prototype four or five years ago and chose not to launch it — partly over accuracy, partly over safety: people in desperate situations, including people who are suicidal, download the app searching for answers, and Gruenewald's view is that someone in that situation should be talking to a trained human, not a chatbot. The company is now building a benchmark to measure how accurately the major models quote Scripture, and pressing OpenAI, Anthropic and Google to quote it correctly. Until that benchmark exists, the checking is up to us.
Why a fluent answer is not a true one
A chatbot does not look a verse up the way a Bible app does. A Bible app retrieves a text that is stored, word for word, and quotes it directly — YouVersion has done that for more than eighteen years. A chatbot generates text that is statistically likely to come next, based on patterns in enormous amounts of writing. Most of the time the result is close to the real verse. Sometimes it is a blend of two verses, or the right idea in the wrong words, or a reference attached to words that belong somewhere else entirely. The danger is not that the errors are obvious. It is that they are not: a wrong quote sounds exactly like a right one, delivered with the same confidence, in the same calm tone.
The Acts 17:11 habit
The answer is not to stop asking chatbots about the Bible. It is to treat every answer the way the Bereans treated Paul's preaching: “they received the word with all eagerness, examining the Scriptures daily to see if these things were so” (Acts 17:11). Three steps, every time.
1. Get the reference. Never accept a quotation without book, chapter and verse. If the chatbot gives you words without a reference, ask for one. If it cannot give one, that is your answer — the words are not a quotation, whatever they sound like.
2. Look it up in a real Bible. Your own Bible, the Bible app on your phone, a Bible website — somewhere the text is stored and quoted directly, not generated. This is the step that catches most errors, and it takes less than a minute.
3. Compare word for word — and check the context. Is the wording right? Is it the verse the chatbot claims? And who is speaking, in what passage, about what? A verse can be quoted accurately and applied wrongly — the devil quoted Psalm 91 accurately enough in the wilderness (Matthew 4:6). The reference is not the end of the check; it is the beginning.
Try it once and the habit will stick. Ask your chatbot where “God helps those who help themselves” appears in the Bible. It does not — the saying is a proverb Franklin printed in Poor Richard's Almanack, not Scripture — and how the chatbot handles that question will tell you everything you need to know about how much to trust its next answer.
Used that way, the chatbot is still useful. It is good at finding passages — “where does Paul talk about…?” — at suggesting cross-references, at explaining a difficult word, at drafting discussion questions for a small group. The line is simple: let it find and suggest; never let it quote and settle. The moment a chatbot's words are about to leave your mouth as Scripture — in a sermon, a lesson, a message to a friend — they have to pass through the real text first.
Why it matters in church
The stakes are highest where the words travel furthest. A pastor who quotes a chatbot's verse in a sermon puts the error in front of the whole congregation — and in the livestream, and in the sermon notes that get shared afterwards. A small-group leader who passes on a chatbot's interpretation shapes how a dozen people read their Bibles that week. A parent answering a child's question teaches, without meaning to, that a fluent answer is a true one. None of this needs a policy or a committee. It needs a habit — one that the church has had since Berea, and that AI has just made necessary again.
The rule is short enough to remember: never quote from a chatbot without checking, and never preach from a reference you have not opened. The chatbot is a fast librarian with a good memory and no conscience about its mistakes. The Bible is still the book. Check the one against the other, and the tool stays a tool.
Sources: Christianity Today, Emily Belz, “YouVersion's Biggest Problem Is Claude, ChatGPT Misquoting the Bible”, 16 November 2026 — the Gruenewald interview: the “generational problem” warning, the benchmark work with OpenAI, Anthropic and Google, the shelved chatbot prototype (accuracy and safety), and the new data showing that half of the Christian questions sent to ChatGPT are about interpreting the Bible; Christian Daily International, March 2026 — Gruenewald's 15–60% misquote estimate. Honest limits: the 15–60% range is Gruenewald's own claim, from a March 2026 interview, and no published method sits behind the numbers — treat it as a warning sign, not a measured rate; YouVersion's benchmark does not exist yet, so no independent comparison of the models is available; the “God helps those who help themselves” test is offered as the reader's own experiment, not as a tested result — the saying itself is a proverb Franklin printed in Poor Richard's Almanack, not Scripture.
Related reading: Pastors want a tool, Christians want a friend — the YouVersion accuracy story in full, and the Barna & Gloo research behind it; AI literacy for the congregation — teaching the whole church to use AI wisely; Writing a sermon with AI without losing your voice — what stays human when AI helps with the sermon; AI preps, people disciple — what AI can prepare vs what stays human. This guidance is general and educational, not legal or safeguarding advice; your church's own safeguarding, data-protection, and denominational policies always take precedence.
Researched and drafted with the help of Muse, an AI assistant. Every claim is checked against its sources, and every article is approved by Aris Angelakopoulos before publication.