People sometimes ask which AI is “most biased.” The honest answer is that they all have a point of view. They just have different ones.
That’s not a conspiracy. It’s how these systems are made.
Where a model’s worldview comes from
Every major AI model gets its outlook from two places:
- Training data. Models learn from enormous amounts of text, much of it from the internet. That text includes Scripture and sound theology. It also includes heresy, mockery, half-truths, and every opinion anyone ever posted. The model absorbs patterns from all of it.
- Its makers’ choices. After training, companies shape how a model behaves: what it will and won’t say, how cautious it is, what tone it takes. Those are human decisions reflecting human values.
So when a model answers a question about marriage, sexuality, the exclusivity of Christ, or the reliability of the Gospels, it isn’t consulting the truth. It’s producing a likely answer shaped by its data and its makers.
Different tools, different leanings
Some models are tuned to be careful and balanced, which can come across as noncommittal on questions Christians think have clear answers. Grok has been marketed as less filtered and more willing to say things other models won’t, which some people read as more honest and others read as a different set of biases. Each company makes trade-offs, and each trade-off is a worldview decision.
None of them is a Christian. None is answerable to a church. None can say “thus says the Lord.”
What this means for believers
Our church’s AI policy says it directly: AI reflects its training data, not God’s heart. Every AI-generated theological claim must be tested against Scripture and the wisdom of the historic church.
A few practical habits:
- Ask for sources, then read them. If AI says “most scholars believe,” find out which scholars.
- Ask the question both ways. “What’s the case for…?” and “What’s the case against…?” You’ll quickly see the model is assembling arguments, not discerning truth.
- Notice the framing. Pay attention to what the answer assumes, not just what it says.
- Bring it to people. If AI gives you an answer about faith that unsettles you, take it to a pastor, mentor, or small group. Don’t sit alone with it.
- Don’t let it set the terms. A model may describe biblical convictions as “some traditions believe.” That’s its framing, not a fact.
“The simple believes everything, but the prudent gives thought to his steps” (Proverbs 14:15). That was good advice for every source before AI. It’s even better advice now.
For parents and youth leaders
Teenagers are asking AI the questions they won’t ask adults. That’s not all bad. But it means we need to talk with them about it. Ask them what they’ve asked. Ask what it said. Read it together. You may find it’s the best open door to a real conversation you’ve had in months.
I help churches teach their people to use AI with discernment, including sessions for parents, youth workers, and leadership teams. I’m a pastor with more than twenty years in business and technology. Here’s how I work.
Enjoyed this article?