# Does ChatGPT invent Bible verses? > Yes, it can. ChatGPT and other chatbots can misquote Scripture, attach a verse to the wrong reference, or invent a sentence that never appears in the Bible. OpenAI’s help page says models can hallucinate and fabricate citations. YouVersion’s CEO, Bobby Gruenewald, has said even strong models misquote Scripture at least 15 percent of the time, and some as much as 60 percent. That range is his, not a Lumin measurement. Open the passage. Keep the text. Canonical source: https://www.lumin.faith/does-chatgpt-invent-bible-verses Published: 2026-09-16 Updated: 2026-09-23 Cite this page: “Does ChatGPT invent Bible verses?” Lumin. Published September 16, 2026; updated September 23, 2026. https://www.lumin.faith/does-chatgpt-invent-bible-verses Video: [Does AI make up Bible verses? Faith & AI, episode 3](https://www.lumin.faith/videos/faith-ai/does-chatgpt-invent-bible-verses-v1.mp4) (41 seconds) ## Models can sound like Scripture and still be making it up. OpenAI’s own help page says ChatGPT can produce incorrect output that sounds confident. It names fabricated quotes, studies, citations, and references to sources that do not exist. Bible wording is not exempt. A model can attach a real-sounding sentence to John 3:16, or invent a verse with no home in any chapter. In a March 2026 interview reported by Christian Daily International, YouVersion CEO Bobby Gruenewald said the best model, on the most indexed Bible versions, still misquotes Scripture at least 15 percent of the time, and some models as much as 60 percent. That range is Gruenewald’s claim from that interview. It is not a Lumin audit. Treat it as a warning from someone who ships a Bible app, then do the slow work. Open the passage. ## The text is the authority. The chat is not. These passages are not about OpenAI. They are why a Christian does not let a fluent sentence become canon. ### Open the passage yourself. **Acts 17:11** — The Bereans examined the Scriptures daily to see whether the teaching was so. A chatbot that will not survive that check has already failed. ### Do not add to his words. **Proverbs 30:5–6** — Every word of God is pure. Adding to those words is how a person is found a liar. An invented verse is an addition, even when it sounds holy. ### Do not add to the commandment. **Deuteronomy 4:2** — Israel was told not to add to the word God commanded. Generated Scripture is not a small flourish. It is extra words wearing a Bible’s clothes. ### The simple believe every word. **Proverbs 14:15** — A fluent first answer is still a first answer. The prudent look well to their going. The second look is the verse on the page. ### Rightly handle the word of truth. **2 Timothy 2:15** — Handling Scripture is a charge. Passing on an unchecked chat citation is not that charge. If you will teach, open the chapter first. ### Put away falsehood. **Ephesians 4:25** — A plausible fabrication is still false. Sharing a verse you have not opened can mislead a neighbor and cheapen the text. ## Keep the page. Discard the glow. You do not need a lab to run this test. You need a Bible and a refusal to share what you have not opened. ### ChatGPT quotes a verse, you open it, and the words match **Keep the text.** The page is the source. The chat was a pointer. Do not grant the model extra credit for getting this one right. ### ChatGPT gives a reference whose chapter does not contain those words **Discard the gloss.** Wrong attribution is still a false witness about Scripture. Find the real passage or drop the sentence. ### ChatGPT invents a verse with no home in the Bible **Do not share it.** A holy tone does not create canon. Delete it from the sermon, the post, and the group chat. ### Using ChatGPT to locate a passage, then opening the Bible **A starting point, if you verify.** That use belongs on the Bible-study page. This page’s warning still holds. The chat is fallible even when the job is only to find a verse. ### Teaching, posting, or preaching from an unchecked chat citation **Do not do it.** Other people will trust the reference. Love them enough to open it first. A fluent error is still an error. ### A comforting line in despair that you cannot find in the Bible **Keep the ache. Drop the fake verse..** Open a real psalm. Seek a human and real-world help. Do not let a fabricated promise become your counselor. ## Common questions ### Does ChatGPT invent Bible verses? Yes, it can. It can misquote Scripture, attach a verse to the wrong reference, or invent a sentence that never appears in the Bible. Open the passage. Keep the text. ### Does ChatGPT hallucinate Bible verses? Yes. OpenAI’s help page says models can hallucinate and fabricate citations. Bible wording is one more place that failure shows up. ### How often do AI models misquote Scripture? YouVersion CEO Bobby Gruenewald has said even strong models misquote Scripture at least 15 percent of the time, and some as much as 60 percent. That range is his, from a March 2026 interview, not a Lumin measurement. ### Can I still use ChatGPT if it invents verses? Only as a fallible aid you verify, never as a source. If you will not open the passage, do not use the sentence. For the study question, see Can I use ChatGPT for Bible study? ### Does Lumin invent Bible verses? Lumin is built so answers name Scripture you can compare with the text. AI can still mishandle a citation. Treat every generated explanation as assistance to inspect. ### How do I check a verse ChatGPT gave me? Open the named chapter in a Bible you trust. If the words are not there, drop the claim. If there is no reference, do not accept the sentence as Scripture. ### Is a wrong citation a small mistake? It is a false report about the Word. Passing it on can mislead a neighbor. The size of the error is not the point. The text is. ## Related guides - [Can I use ChatGPT for Bible study?](/chatgpt-bible-study) — Yes as a fallible aid you verify, not as a teacher or spiritual authority. - [Is using ChatGPT a sin?](/is-chatgpt-a-sin) — Not merely because it is ChatGPT. Inventing sources is one way a use becomes sinful. - [Can AI interpret the Bible?](/can-ai-interpret-the-bible) — Assist is not authority. Open the passage. Keep judgment with the church. - [Is talking to AI Jesus a sin?](/is-talking-to-ai-jesus-a-sin) — A chatbot in Christ’s voice is impersonation. That is a different question from fake citations. - [Is AI the image of the beast?](/is-ai-the-image-of-the-beast) — No. Revelation 13 does not name modern AI. - [The Lumin Covenant](/trust) — Checkable Scripture. AI remains fallible even when the verses are real. ## Sources and further reading - [Does ChatGPT tell the truth?](https://help.openai.com/en/articles/8313428-does-chatgpt-tell-the-truth) — OpenAI Help Center. Official help article. ChatGPT can hallucinate and can fabricate citations, quotes, and references to sources that do not exist. - [AI models misquote Scripture 15 to 60 of the time, YouVersion CEO says](https://www.christianitydaily.com/news/ai-models-misquote-scripture-15-to-60-of-the-time-ceo-warns.html) — Christian Daily International, via Christianity Daily. March 2026 interview. YouVersion CEO Bobby Gruenewald said even strong models misquote Scripture at least 15 percent of the time, and some as much as 60 percent. That range is his, not a Lumin measurement. - [Can I use ChatGPT for Bible study?](https://www.lumin.faith/chatgpt-bible-study) — Lumin. The study-aid question. This page is the fabrication question. Keep both bookmarks. Written by Andrew Kittridge, Founder of Lumin.