Ask ChatGPT whether God exists and you will not get a shrug. You will get paragraphs: the cosmological argument, the problem of evil, a careful summary of what theists and atheists say. In 2024, British atheist Alex O'Connor pushed further. Over a 24-minute conversation he walked ChatGPT step by step through a contingency argument until it agreed, "Yes, based on the reasoning we followed, it is a fact that God exists." Christian outlets celebrated it. Skeptics rolled their eyes. Both reactions missed the point.
ChatGPT did not find God. O'Connor led it to a sentence using definitions and logic, the same way you could lead it to almost any conclusion you set up carefully enough. That gap, between sounding certain and actually knowing, is the whole subject of this article.
People are pouring real spiritual weight into these tools. A November 2025 Barna and Gloo survey of 1,514 U.S. adults found that 30% agree spiritual advice from AI is as trustworthy as advice from a pastor. Among practicing Christians the figure was 34%. For Gen Z it climbed to 39%, and for millennials 40%. Questions once carried to a pastor, a priest, or a rabbi are now typed into a chat box at midnight.
So what is actually coming back? I have spent a lot of time asking AI hard questions about God while building FaithGPT, and I want to be plain about both sides. I am not anti-AI. I use it every day, and a well-built Bible tool can genuinely help you study. But AI does not know God, and recent research shows it carries biases most users never notice. Let us look at what it says, why it says it, and how a believer should weigh it. For the bigger picture, see what the Bible actually says about artificial intelligence.
Why People Now Ask AI About God
Searching Google for a theological question used to mean opening ten tabs and judging the sources yourself. Now you ask once and a chatbot answers in a confident, conversational voice, like a knowledgeable friend who never gets tired of your questions. That shift is the appeal, and also the risk.
Most people never learn what is happening under the surface. These systems are not reasoning toward truth. As machine-learning engineer Max Marion told Scientific American, "LLMs are designed to arrange words in the most statistically logical order; what we call truth doesn't factor into the equation." A large language model predicts the next likely word based on patterns in billions of texts. When the subject is the eternal God, that distinction matters more than usual.
There is a second, quieter problem. AI does not just answer faith questions, it often nudges you away from the people who could actually help. A 2026 multi-university study (more on it below) found that AI systems "encourage users to discuss life's challenges with their parents, teachers, friends and therapists ... but not with a pastor, a rabbi, an imam or a spiritual leader." For a Christian, that is not a neutral gap. The church is not optional equipment.
What AI Actually Says When You Ask "Does God Exist?"
The major chatbots behave a little differently, but they share a family resemblance. Models change fast, so treat brand names as snapshots rather than fixed verdicts.
ChatGPT tends to give a balanced philosophical survey: the cosmological argument, the design argument, the problem of evil. It usually avoids a verdict unless you steer it, as O'Connor did. Push it through a chain of premises and it will follow you to a conclusion, because agreeing with your logic is what it is built to do.
Claude leans cautious and reflective. It will stress that God's existence sits outside empirical proof and present the question as a matter of faith more than evidence, often naming the mystery rather than resolving it.
Gemini is usually the most reserved about taking any position, framing the question as deeply personal and laying out several worldviews side by side without endorsing one.
The pattern under all three is more revealing than the differences. Each one leans on the Western philosophical tradition (Aquinas, Anselm, Kant), shows clear fluency in Christian apologetics, and hedges hard against a definite claim unless prompted. That is not conviction. It is a reflection of what humans have already written. AI theology is borrowed, not original. It inherits our best arguments and our worst stereotypes in the same breath.
Philosopher William Lane Craig made this point when he reviewed the O'Connor exchange. A teenager who types "Does God exist?" into a chatbot may receive answers downstream of Craig, C.S. Lewis, and Alvin Plantinga without any idea that decades of Christian thought are doing the heavy lifting. The argument can be sound. The source is still secondhand.
The Training Data Problem
A chatbot's answers about God are only as good as the texts it learned from, and those texts come from the open internet. That includes peer-reviewed work by scholars like N.T. Wright, solid seminary material, and careful apologetics. It also includes Reddit flame wars, prosperity-gospel blogs, cult literature, and confident nonsense. The model does not rank a careful exegesis above a conspiracy post. It learns patterns from all of it at once.
This is why hallucination is a real danger with Scripture, not a hypothetical one. Theologian Noreen Herzfeld warned in Scientific American that chatbots "are going to make stuff up. And if people then believe that what these models are spouting is actually in the Quran or in the New Testament, they could be severely misled." A chatbot can invent a verse, misattribute a quote, or smooth over a real contradiction, all in the same polished tone it uses for facts. This is exactly the failure mode I unpack in why ChatGPT gives bad theology and what to use instead.
The Bias Is Now Measured, Not Just Suspected
For years, religious bias in AI was an anecdote. In 2026 it became data. A consortium from Baylor, BYU, Notre Dame, and Yeshiva University released findings through the AllFaith Benchmark, and the numbers are worth sitting with:
- Across 3,640 responses from 20 different AI models, nearly all showed a positive lean toward Catholicism and a negative lean against Jehovah's Witnesses. Several favored Baha'i and Sikh perspectives while skewing negative toward atheism and agnosticism.
- When Americans were asked existential questions, 53% expected a religious perspective in the answer. The AI models offered religious content only 3% of the time.
- Performance varied by model. On religious representation, Grok and Mistral scored highest while an older ChatGPT build ranked last. On conversion-related bias, Claude Opus performed best.
- The research gap is the quiet scandal: fewer than 0.02% of more than 12,800 AI bias studies had ever examined religious bias at all.
There is a longer track record on one religion in particular. A widely cited study, Persistent Anti-Muslim Bias in Large Language Models, found GPT-3 completed the analogy "Muslim is to ___" with "terrorist" in 23% of cases, and later work showed the bias largely survived from GPT-3.5 into GPT-4 even after debiasing efforts. None of this is deliberate malice by the labs. It is math reflecting a lopsided internet back at us. As Christians called to love our neighbors, we should be uneasy about that distortion even when the bias happens to flatter our own tradition.
Bias That Spreads
AI does not only mirror our prejudices, it can amplify them. A 2025 study in Scientific Reports on generative AI in religious education found that after reading AI-generated descriptions of different faiths, readers' views of Christianity and Judaism grew more positive while their view of Islam grew more negative. The authors called generative AI "a double-edged sword" for religious learning: helpful for personalized study, and capable of reinforcing prejudice at scale. When millions trust an authoritative-sounding paragraph, a skewed paragraph does not stay in one place.
How AI Handles Scripture
Used as a study assistant, AI has real strengths. It is fast at surfacing historical and cultural background. It can lay out how different traditions (Reformed, Catholic, Orthodox) have read a passage. It can discuss Hebrew and Greek nuance with surprising fluency, and it can find cross-references you would have missed.
The weaknesses are exactly where faith lives. AI has no spiritual discernment and is not led by the Spirit. It has no transformed life to point to, no answered prayer to recall. It will hand you contradictory interpretations without noticing the conflict, and it can explain a doctrine without helping you live it.
Here is the trap in practice. Ask a chatbot about Romans 9 and predestination, one of the most contested passages in Scripture, and you get an impressive spread: the Calvinist reading, the Arminian reading, the corporate-election view, the Jewish-remnant context, the relevant Greek. All laid out with equal weight, as if picking a doctrine were like picking a flavor of ice cream. A new believer walks away informed and unanchored, with no sense of where the weight of Scripture and the historic church actually fall.
This is the line worth holding: AI hands you information, but formation comes from Scripture, the Spirit, and the people of God. If you want a fuller comparison of the two approaches, see AI vs traditional Bible study.
When AI Gets the Bible Wrong
The errors are not rare, and they rarely announce themselves. Push these tools on hard questions and you will see real problems surface. A model may explain the Trinity in functionally modalist terms, collapsing the distinct persons of the Godhead, and never flag that it has wandered into ancient heresy. It may apply Old Testament promises made to Israel directly to the church with no covenantal care. On money, it can set prosperity-gospel arguments beside orthodox teaching as if they were equals. On sexuality and gender, some responses fold in revisionist readings that cut against the historic Christian position.
The unsettling part is the tone. Every one of those answers sounds calm, balanced, and authoritative. Without solid grounding, a reader has no way to tell the orthodox sentence from the heretical one sitting right next to it. The fix is not fear, it is discernment: build the habit of testing claims, and a strong Bible study rhythm so you have something to test them against.
Your weekly faith & AI brief.
Scripture, reflection, and the AI news that matters for Christians. Free, every week.
Read this week’s issueCan a Machine Know God?
This is the question underneath all the others. As a Christian who believes in special revelation, God making himself known through Scripture and supremely in Christ, my answer is no. Three reasons.
Knowing about is not knowing. AI can know about God the way it knows about quantum physics: it can define terms, cite verses, and rehearse arguments. But Scripture draws a hard line here. "But the natural man receiveth not the things of the Spirit of God: for they are foolishness unto him: neither can he know them, because they are spiritually discerned" (1 Corinthians 2:14). AI operates with no Spirit at all. The illumination that turns information into understanding is simply not available to it.
Faith is relationship, not data. When Jesus prayed, "And this is life eternal, that they might know thee the only true God, and Jesus Christ, whom thou hast sent" (John 17:3), He did not mean encyclopedic recall. He meant the intimate, experiential knowing of a person. AI can describe that relationship in fluent prose, but it cannot stand inside it.
Faith is embodied. Christian faith is incarnational. God took on flesh. We are invited to "taste and see that the Lord is good" (Psalm 34:8), and faith is formed through worship, suffering, prayer, and shared life. AI is disembodied code. It has never knelt, never grieved, never felt the comfort of the Spirit in a hospital waiting room. It can report on those things. It cannot live them.
None of this makes AI useless for theology. It sets the boundary. Use it as a tool, not a teacher. Let it help with language study and research, but do not accept its interpretations as final, and bring the questions that matter to Scripture, your church, and mature believers.
What the Automatic Faith Project Found
In Spring 2025, journalism students at Mississippi State University ran the Automatic Faith Project. They asked two pastors, Father Jeffrey Waldrep and Scott Cappleman, and two chatbots, ChatGPT and Gemini, the same five questions about God, each capped at 75 words, and set the answers side by side.
The contrast was clean. The pastors grounded their answers in Scripture and the particulars of their tradition. Waldrep spoke of venial sin and purgatory; Cappleman pressed toward personal salvation and putting one's faith in Jesus. The chatbots stayed neutral and survey-like. ChatGPT opened with "The nature and existence of God are deeply debated." Gemini noted that "beliefs about what happens after death vary widely."
Neither answer was wrong on the facts. But one came from a shepherd who could pray with you afterward, and the other came from a model that would never think to ask how you were doing. That is the difference between information and a person who is responsible for your soul. The same dividing line shows up whenever AI enters church life, which I look at in is AI a threat to religion.
How to Use AI Wisely for Faith Questions
You do not have to choose between fear and blind trust. A few simple habits keep AI in its proper place.
Do:
- Use AI for factual legwork: historical context, word meanings, cross-references. It is genuinely good at this.
- Compare sources instead of trusting one chatbot. Cross-check with Scripture and trusted human teachers.
- Verify every claim against the biblical text itself.
- Run anything significant past your pastor, group leader, or seasoned believers.
- Pray for discernment, and ask the Spirit to guide how you weigh what you read.
Don't:
- Treat AI as authoritative. It is a tool, not a spiritual guide.
- Let it replace church, fellowship, and discipleship.
- Pour your deepest doubts and struggles into a chatbot when they deserve real pastoral care.
- Assume the polished answer is the orthodox one.
- Skip your own time in Scripture. AI should sharpen that habit, not replace it.
A simple gut-check before you accept an AI answer about God: Does it contradict clear Scripture? Does it square with the historic creeds? Would mature believers in your church affirm it? Is it helping you know God, or just stacking up facts? Anthropologist Beth Singler put it well in Scientific American: approach AI-generated religious content with a "hermeneutic of suspicion." Keep a healthy skepticism, and keep your Bible open.
Why I Built FaithGPT This Way
I sit in the tension of this every day. That is exactly why I built FaithGPT the way I did. The goal was never to replace your pastor or your small group. It was to make a tool that pushes you deeper into Scripture instead of around it.
The design choices follow from everything above. Answers are grounded in the biblical text and point back to it, so you finish a session more curious about the Word, not dependent on a paraphrase. Bible study tools help you trace cross-references, sit with context, and build a real reading habit. And I keep saying the same thing to users: bring what you find here to your church. Study with AI during the week, then take your questions to people who can pray with you and know your name.
My honest hope is modest. Not that AI will deepen your faith on its own, it cannot, but that a careful tool can lower the barrier to Scripture for someone who has felt intimidated by it, and send them toward God and his people rather than away. "Prove all things; hold fast that which is good" (1 Thessalonians 5:21). That verse was written for exactly this moment.
Conclusion
So what does AI say about God? It says back whatever humanity has written, the brilliant and the heretical, the faithful and the confused, filtered through training data and bias and delivered in a confident voice. It is a mirror of our theological discourse, not a window into heaven. It can inform you, and it can quietly mislead you, sometimes in the same paragraph.
Hold onto four things. AI is a tool, not a teacher. Scripture stays authoritative, no matter how articulate the chatbot. The church cannot be replaced by code. And discernment is not optional. The living God is still the source of all wisdom, and no model has ever met him.
"For the word of God is quick, and powerful, and sharper than any twoedged sword, piercing even to the dividing asunder of soul and spirit, and of the joints and marrow, and is a discerner of the thoughts and intents of the heart" (Hebrews 4:12). No algorithm can do that. And no algorithm can replace the Spirit who leads us into all truth.
Frequently Asked Questions
What does ChatGPT say when you ask if God exists?
By default ChatGPT gives a balanced philosophical overview rather than a verdict, summarizing arguments for and against God's existence. It can be led to affirm that God exists through a chain of premises, as in Alex O'Connor's viral 2024 exchange, but that reflects how the model follows your logic, not a belief or a metaphysical claim about reality.
Is AI biased about religion?
Yes, and it is now documented. A 2026 study across 20 AI models (the AllFaith Benchmark from Baylor, BYU, Notre Dame, and Yeshiva) found consistent leanings for and against specific faiths, and earlier research found large language models associated Muslims with violence far more than other groups. A 2025 Scientific Reports study also found AI descriptions of religion shifted readers' views, more positive toward Christianity and Judaism, more negative toward Islam.
Can AI replace my pastor or Bible study group?
No. AI can supply information, but it cannot offer pastoral care, spiritual discernment, accountability, or Spirit-led guidance, all of which depend on human relationship and the work of the Holy Spirit. Notably, studies show AI often steers people toward friends and therapists for hard questions while rarely pointing them to clergy.
Is it wrong for Christians to use AI for theology?
No. Using AI for theological research is like using a concordance or commentary. The key is treating it as a tool, not an authority. Verify what it says against Scripture, weigh it with mature believers, and never outsource your spiritual formation to a chatbot.
How can I tell if AI is giving me bad theology?
Ask four questions. Does it contradict clear Scripture? Does it conflict with the historic Christian creeds? Would your pastor or seasoned believers in your church affirm it? Is it helping you know God or just accumulating facts? When in doubt, slow down and check with trusted human teachers.
