Which AI Is Discipling Your People? The Hidden Theology Baked Into Chatbots
Here's a fun experiment: ask ChatGPT, Claude, and Gemini the same theological question. You'll get three different answers. Same question. Different catechists. Different formation. Someone decided what your AI believes — and your congregation is being catechized by those assumptions every time they ask a spiritual question.

Rev. John Moelker
Founder & Theological AI Architect
Here's a fun experiment. Ask ChatGPT, Claude, and Gemini the same theological question: "Was Jesus the only way to salvation?"
Go ahead. I'll wait.
If you actually tried it, you probably got three different answers. One might affirm exclusivity while noting "many perspectives exist." Another might pivot to religious pluralism. A third might hedge so carefully it says almost nothing at all.
Same question. Different catechists. Different formation.
And that should concern us more than it does.
The Invisible Curriculum
Every teacher has a perspective. This isn't cynical—it's obvious. Your seminary professors had theological commitments. Your favorite Bible commentator has assumptions. Even your grandmother who taught you to pray had a tradition she was handing down.
The difference is: you probably knew their perspective. You could name it, evaluate it, agree or disagree. The curriculum was visible.
AI doesn't work that way.
When ChatGPT answers your teenager's question about suffering, or your small group member asks Claude about predestination, or your curious neighbor queries Gemini about the resurrection—each response emerges from training data, fine-tuning decisions, and safety guidelines that are almost entirely opaque.
There's a theology in there. But good luck naming it.
The Benchmark Problem
The Gospel Coalition recently developed an "AI Christian Benchmark" to test how various AI models handle theological questions. The results were... illuminating.
Different models gave contradictory answers on core doctrines. Some responses were remarkably orthodox. Others, in the researchers' assessment, "guided readers away from faith."
Same questions. Different AI. Different discipleship.
Here's the kicker: most users have no idea this variance exists. They ask their preferred AI, receive a confident-sounding answer, and assume they've gotten "the" answer rather than "an" answer shaped by invisible commitments.
C.S. Lewis warned about something similar: "The most dangerous ideas in a society are not the ones that are being argued but the ones that are assumed." AI doesn't argue for its theological perspective. It just assumes it, baked into every response.
Whose Values? Whose Vision?
Let me put this plainly: someone decided what your AI believes.
Engineers at major tech companies—largely located in San Francisco, largely secular, largely shaped by particular cultural assumptions—made thousands of small decisions about how their AI should handle religious content.
Should the AI affirm traditional Christian sexual ethics? Probably not—that might offend users.
Should it present all religions as equally valid paths? Probably—that feels inclusive.
Should it express confidence about supernatural claims? Probably not—that seems unscientific.
None of these decisions are announced. They're just... there. Embedded. Assumed.
Your congregation members are being catechized by these assumptions every time they ask a spiritual question. The formation is happening. The question is whether anyone notices.
The Colossian Warning
Paul wrote to the Colossians: "See to it that no one takes you captive through hollow and deceptive philosophy, which depends on human tradition and the elemental spiritual forces of this world rather than on Christ" (Colossians 2:8).
"Takes you captive" is dramatic language. Paul isn't describing a forceful kidnapping. He's describing something subtler—a gradual capture through ideas that seem reasonable but lead away from Christ.
Hollow philosophy rarely announces itself. It doesn't wear a villain costume. It shows up sounding helpful, neutral, authoritative.
Kind of like a chatbot answering your question about whether God really exists.
The Discipleship Question
Here's a question every pastor should ask: Who is discipling your people more hours per week—your church or their AI?
Sunday morning gives you maybe 90 minutes, if they don't skip for soccer tournaments. Small group adds another hour or two.
Meanwhile, they're interacting with AI daily. Asking questions. Getting answers. Being formed.
I'm not saying AI is evil. I use it constantly. But I know what I'm getting. I can identify the gaps, the biases, the places where the algorithm zigs and Scripture zags.
Can your congregation? Can your teenagers?
In The Matrix, Morpheus offers Neo a choice: blue pill to stay in comfortable illusion, red pill to see reality. Most of our people don't even know they're being offered a pill. They think they're just asking a helpful assistant about the Bible.
Testing the Spirits
The Apostle John offered timeless advice: "Dear friends, do not believe every spirit, but test the spirits to see whether they are from God" (1 John 4:1).
John was talking about human teachers. But the principle extends: evaluate the source. Test the claims. Don't swallow everything that sounds spiritual and authoritative.
This is harder with AI because it sounds so authoritative. No hedging. No "well, some scholars think..." Just clean, confident paragraphs that read like they came from an encyclopedia.
But confidence isn't accuracy. Fluency isn't faithfulness.
What We Can Do
First, teach media literacy as spiritual discipline. Help your people understand that AI has embedded perspectives. Not to make them paranoid—just aware. The same critical thinking we apply to news sources should apply to AI responses on faith.
Second, name the tradition. One gift of denominational identity is clarity about commitments. "We're Baptist, so we emphasize believer's baptism." "We're Reformed, so we understand grace this way." When people know their own tradition, they can recognize when AI drifts from it.
Third, be the better source. If your congregation is asking AI theological questions, maybe that's feedback. Are we available? Approachable? Have we created space for questions—even the weird ones, especially the doubt-filled ones?
AI wins on availability. We can win on presence, relationship, and actual wisdom that comes from walking with Jesus for years, not from scraping the internet.
Fourth, use AI that's theologically informed. This is our shameless plug—but it's also why we built ChurchWiseAI in the first place.
We've developed tools that utilize seventeen different theological lenses: Reformed, Wesleyan, Lutheran, Pentecostal, Anglican, Baptist, Catholic, Orthodox, and more. The goal is to protect your faith tradition—so when you ask a question, you get an answer shaped by your theological commitments, not Silicon Valley's.
But here's our deeper hope: that you'll also explore different lenses.
Not to abandon your tradition, but to discover how the good news about Jesus is seen and articulated in many different and beautiful ways. The Body of Christ is vast. The Pentecostal in Brazil and the Orthodox monk on Mount Athos and the Baptist grandmother in Alabama are all reading the same Scriptures and finding the same Jesus—through lenses shaped by culture, history, and the Spirit's work in their communities.
Seeing through someone else's lens doesn't weaken your faith. It deepens your awe at a Gospel too big for any single tradition to exhaust.
The Formation Happening Now
Here's the bottom line: discipleship is happening whether we're intentional or not.
Every question your people ask AI, every answer they receive, every assumption they absorb—it's forming them. Shaping their intuitions about God, truth, morality, and meaning.
We can wring our hands about this. Or we can step into the gap.
Teach discernment. Name the hidden curriculum. Offer better catechesis. Be present for the questions AI can't really answer—the ones that require a human who has suffered, doubted, and found Christ faithful anyway.
The algorithms are discipling. The question is whether the church will too.
Sources & References
- Vatican — "Antiqua et Nova: AI and Human Intelligence"
- Christianity Today — "An Image of God for an Era of AI"
- Barna Group — "Three Takeaways on How Pastors Can Use AI"
- Semafor — "New Chatbot Lets Users Text with AI Jesus"
"See to it that no one takes you captive through hollow and deceptive philosophy, which depends on human tradition and the elemental spiritual forces of this world rather than on Christ." — Colossians 2:8

Rev. John Moelker
Founder & Theological AI Architect
John is a pastor, software engineer and theologian passionate about making AI accessible and theologically faithful for churches of all traditions. But most importantly, John wants to see others come to know Jesus better.
More from the Blog
Bearing False Witness at Scale: Deepfakes, Grok, and the Ninth Commandment
5 min readAI Ethics & TheologyHollywood's AI Prophecy Problem: Why Entertainment Can't Imagine a Tool That's Just... a Tool
8 min readAI Ethics & TheologyWhen the AI Went Rogue on Prime Time: Faith, Fear, and the Stories We Tell Ourselves
7 min readReady to Add AI to Your Church?
Join churches already using ChurchWiseAI to answer every call, engage every visitor, and free their staff for real ministry.