sign in

Does AI make up Bible verses? Why it happens and how to check any verse in a minute

Updated · By the Versiah team

Yes, general-purpose AI chatbots can make up Bible verses, attach real words to the wrong reference, or blend and paraphrase passages while presenting them as exact quotations. It happens because language models generate likely-sounding text rather than looking words up. You can catch it in about a minute by opening the reference in a Bible and reading the verses around it, and tools that ground answers in the actual verse text avoid most of the problem.

The short answer

AI can and sometimes does produce Bible verses that are not in the Bible, or real verses with the wrong reference. Treat any quotation you have not checked as unverified. Open the reference, compare the words, and read the paragraph around it.

This is not a reason to panic or to avoid AI altogether. Well-known verses like John 3:16 or Psalm 23:1 are almost always quoted correctly, because they appear so often in the writing these systems learned from. The trouble tends to show up with less familiar passages, exact wording, verse numbers and anything that sounds like a proverb.

Why AI gets verses wrong

Most chatbots are built on large language models. A model does not open a Bible when it answers you. It writes one word at a time, choosing what is most likely to come next based on patterns it absorbed during training. Most of the time the most likely words are the right ones. Sometimes they are merely plausible.

  • Scripture has a recognisable style. Thee and thou, “blessed is the one who,” “the LORD says.” A model can imitate that style perfectly while producing a sentence no biblical author wrote.
  • Numbers are fragile. Chapter and verse numbers carry no meaning of their own, so a model can easily attach the right words to Proverbs 3:6 when they belong to 3:5, or to the wrong book entirely.
  • Translations get mixed. English has dozens of translations. A model may start a verse in one wording and finish it in another, producing a quotation that matches no real Bible.
  • Confidence is not a signal. A model sounds equally sure whether it is right or wrong. The tone of an answer tells you nothing about its accuracy.

The five most common mistakes

1. The invented verse

A sentence with a biblical feel and a real-looking reference, but no such words exist at that address or anywhere else. This is the rarest error, and the most serious.

2. Right words, wrong address

The quotation is genuine but the reference points somewhere else, often a neighbouring verse or a similar passage in another book. Harmless if you only want the idea, confusing if you go looking for it.

3. The paraphrase in quotation marks

The meaning is roughly right, but the words are the model’s own, presented as if they were a translation. Small shifts in wording can matter, especially for verses people memorise or preach from.

4. Two verses stitched together

Half of one verse and half of another, joined into a single smooth sentence. Each half may be real, but the combination says something neither author said on its own.

5. A real verse used for something it does not mean

The quotation is perfect, but it is cut away from its context and made to promise something the passage never promised. This is the hardest mistake to catch, and humans make it all the time too.

How to check any verse in about a minute

  1. Find the reference. If the answer gives none, ask for it. A quotation with no address is unverified.
  2. Open it in a Bible. A printed Bible or any Bible app will do. Check that the words match, allowing for translation differences.
  3. Read three verses either side. Who is speaking? To whom? Is it a command, a promise, a prayer, a poem, or someone’s mistaken opinion?
  4. Compare a second translation. If the wording differs a lot, that is normal. If the meaning differs, slow down and look closer.
  5. Search the words if the reference fails. Paste a distinctive phrase into a Bible search. If it appears nowhere, it is not Scripture.

Doing this a few times trains your instincts. You start to notice when something sounds more like a greeting card than like Paul or Isaiah. That is a good instinct to have with any source, not only AI.

How grounding fixes most of the problem

The reliable fix is simple in principle: do not let the model quote Scripture from memory. A grounded tool first retrieves the relevant passages from an actual Bible text, gives those exact verses to the model to work with, and then checks that every reference in the finished answer exists and matches. Words that fail the check do not reach you as Scripture.

Grounding does not guarantee a wise answer. A tool can still choose passages poorly or lean too hard on one verse, so reading in context still matters. But it removes invented and misquoted verses, which is the error that does the most damage because it puts words in God’s mouth.

This is how Versiah works: answers cite exact verses, those references are checked against the Bible text, and you can tap any of them to read the passage in context and in another translation. If you use a different tool, it is worth testing it with a few obscure questions and checking every reference it gives you. Our guide to the best AI Bible apps looks at how the main options handle this.

Why accuracy matters so much here

Misquoting a novel is careless. Misquoting Scripture is more serious, because people shape their prayers, choices and sense of God around what they believe the Bible says. The Bible itself warns against adding to God’s words.

Do not add to His words, lest He rebuke you and prove you a liar.
Proverbs 30:6 (BSB)

The warning is aimed at people who put their own ideas in God’s mouth. A study tool, human or digital, should be held to the same standard.

King James VersionAdd thou not unto his words, lest he reprove thee, and thou be found a liar.

The good news is that checking is easy, and the habit of checking makes you a better reader. For a broader look at the question, read is it okay to use AI for Bible study?, and for a step-by-step method of reading passages in context, see how to study the Bible.

Questions people ask

Does ChatGPT make up Bible verses?

It can. General chatbots usually quote famous verses correctly, but they sometimes give the wrong reference, blend two verses, paraphrase in quotation marks, or produce a line that is not in the Bible. Check every reference in a Bible before relying on it.

Is “God helps those who help themselves” in the Bible?

No. It is a proverb from outside Scripture that is often mistaken for a verse. The Bible more often speaks of God helping those who cannot help themselves, as in Psalm 46:1.

How can I tell if a Bible verse is real?

Look up the reference in a Bible or Bible app and check that the words match. If there is no reference, search a distinctive phrase in a Bible search tool. If it appears nowhere, it is not Scripture, however biblical it sounds.

Why does AI get Bible references wrong?

Language models generate text by predicting likely words rather than looking them up. Verse numbers carry no meaning of their own, so they are easy to swap, and dozens of English translations make it easy to mix wordings.

Are AI Bible apps more accurate than general chatbots?

Some are, depending on how they are built. Apps that retrieve the actual verse text and check their references against it avoid most invented or misquoted verses. Ask how a tool handles quotations, and test it with a few less familiar passages.

Read next