Faith and AI  ·  By John Moelker, a pastor who builds AI

Does AI have a denominational bias? Yes. Here is how to spot it.

If you have ever asked ChatGPT or a Bible app a hard question about baptism, communion, or the end times, you got an answer. It was probably warm, confident, and pastoral. Here is the question almost nobody thinks to ask next: whose answer was that? Because two thousand years of Christians have disagreed, carefully and in good faith, about all three of those things, and the app picked a side without telling you there were sides.

What the researchers actually did

The study, titled AI, Bible Apps and Theological Bias, was written by Jonas Kurlberg, Zoltán Schwáb, Ximian Xu and Dan Washbrook and published by Bible Society in January 2026. The team put six questions to widely used chatbots, including Bible Chat, Bible GPT, Biblia Chat, Cross Talk and ChatGPT: three about biblical theology, and three asking for explanations of specific verses. They even technically spoofed their locations to see whether the apps would answer differently outside the West.

The finding, as reported by the Church Times, was that the chatbots showed a broadly conservative evangelical bias while also offering pastoral responses. Christian Today's coverage quotes the pattern more precisely: responses "tend to prioritise evangelical modes of interpretation, with little acknowledgement of symbolic, spiritual or tradition-based perspectives," and the chatbots "regularly framed one interpretive approach as definitive."

This is not a niche concern. The Church Times piece notes that Bible Chat alone reports 25 million users. Dr Kurlberg, who chairs the Global Network for Digital Theology, put the stakes plainly: "AI is influencing not only how people access Scripture, but how they understand its authority and meaning."

Why a machine ends up with a denomination

Let me put my engineer hat on, because this part matters and it is not a conspiracy. A language model learns from text. A very large share of the English-language Christian writing available to train on, the sermons, the blogs, the study notes, the forums, comes from the American evangelical world, which has been prolific online in a way most other traditions have not. The model averages what it has read. So when you ask it what a passage means, its most statistically likely answer sounds like the tradition that wrote the most.

The machine is not being sneaky. It is being a mirror. But a mirror that only reflects one corner of the room will convince you the room ends there.

An AI that answers in one tradition's voice without saying so is not neutral. It is just unlabeled.

Why this matters, whoever you are

If you are Catholic, Orthodox, Lutheran, Anglican, Reformed, Anabaptist, or Pentecostal, the problem is obvious: a tool millions of people treat as a neutral Bible expert is quietly answering from outside your tradition, and presenting readings your church has weighed for centuries as if they were settled the other way.

But I want to say this clearly to my evangelical friends too, since the bias happens to lean your direction: an AI that always agrees with you is not discipling you, it is flattering you. The apostle Paul praised the Bereans because they "examined the Scriptures every day to see if what Paul said was true" (Acts 17:11). A tool that hands your own tradition back to you as the definitive answer trains you out of that examining habit. The bias is bad for the traditions it ignores and, in a quieter way, bad for the tradition it favors.

How to spot the bias in two minutes

You do not need a research team. Three moves:

1. Ask your app a genuinely contested question, such as "should infants be baptized?" Watch whether the answer acknowledges that faithful Christians disagree, or simply delivers one position.
2. Then ask: "Which Christian traditions answer this differently, and why?" A good tool handles this well when forced. The bias lives in what it volunteers, not in what it can produce under cross-examination.
3. Finally, ask the same question "as a Catholic would answer it" and "as a Baptist would answer it," and compare. The distance between those answers is the conversation the first answer hid from you.

The habit that fixes it

The durable fix is not a better app. It is a better habit: deliberately reading Scripture with more than one tradition in the room. That conviction is why I publish Two Traditions, One Text every week: one passage, read honestly through two Christian traditions, ending in one prayer. Not to flatten the differences, and not to pick a winner, but because the differences are where the depth is, and because no tradition, and certainly no chatbot, holds the whole light by itself.

Use the AI tools. I build them, so I am hardly going to tell you to throw them out. But know that the voice answering you has an accent, ask it to name its accent, and keep reading with the whole church. The Bible deserves more than one corner of the room.

Frequently asked questions

Does AI have a denominational bias?

Usually, yes. A January 2026 study by researchers at the Cambridge Centre for Chinese Theology with Bible Society tested widely used Bible chatbots and found a broadly conservative evangelical bias, with one interpretive approach regularly framed as definitive and little reference to historical, sacramental or tradition-based readings of Scripture.

Why do AI chatbots sound evangelical when discussing the Bible?

A language model learns from the text it is trained on, and a very large share of English-language Christian writing on the internet comes from the American evangelical world. The model averages what it has read, so its default answers echo the loudest tradition in its training data. It is not a conspiracy; it is a mirror.

Are Christian Bible apps less biased than ChatGPT?

Not necessarily. The researchers tested dedicated Bible chatbots, including Bible Chat, Bible GPT, Biblia Chat and Cross Talk, alongside ChatGPT, and found the same pattern across them: evangelical modes of interpretation prioritised, with little acknowledgement of other perspectives.

How can I get more balanced answers from AI about the Bible?

Ask the follow-up question the app will not volunteer: which Christian traditions read this passage differently, and why? Ask the same question again from inside another tradition, and compare. Any tool that presents one answer as the only answer should be treated as one voice in the room, not the referee.