What's shaping your faith?

How AI tools affect Christian faith formation – and how we need to discern

Canadian Christians are already using AI tools for prayer, Bible study and devotional life. The question isn’t whether to engage. It’s how to do it faithfully.

Last August, I sat in on a webinar hosted by Unseminary called Stop Dreading Sermon Prep: How Pastors Are Using AI to Save Time and Preach Better. The pitch was for an AI sermon prep platform trained on what the presenters called the foundational documents of Christianity, built to generate sermon drafts pastors could trust.

I asked the obvious question: Whose Christianity? How would the tool handle denominational differences or the cultural flashpoints that divide Christians who otherwise share a creed? The answer was vague. I left with the impression the question hadn’t really been considered. The tool had been built to sound theologically solid without anyone deciding, in advance, whose theology it would default to.

That pattern shows up in marketing a lot – sermon prep tools sold as theologically sound, full stop, as though soundness were a fixed category rather than the question denominations have argued over for centuries.

Such issues are quietly becoming even more important in Canadian churches as a result of AI. The Bible app on your phone may be surfacing verses chosen by an algorithm trained on millions of user interactions. The devotional your small group leader recommended might have been shaped, at least in part, by an AI writing assistant. The prayer prompt in your inbox this morning could have been generated by a system that has processed more Scripture than any theologian alive, and has never prayed a word of it.

None of this is necessarily bad. Much of it is genuinely good. These tools are helping overstretched pastors serve their congregations more effectively, making Scripture more accessible, and opening spiritual practices to people who might never have found them otherwise.

Rather than debate whether Canadian Christians should use them, it’s more urgent to ask whether we’ve developed any shared wisdom about how.

AI tools for spiritual life have proliferated faster than most churches have noticed.

Not insight, but consensus

AI tools for spiritual life have proliferated faster than most churches have noticed. YouVersion passed 1 billion downloads last November. Glorify and Hallow add tens of millions more between them. Sermon preparation platforms like SermonSpark and Sermonly now offer AI-assisted research and outline generation. For the average Canadian Christian, these tools have become part of the furniture of daily faith.

What distinguishes them from a concordance or commentary is that they’re generative and responsive. They don’t just retrieve information. They produce it, customized to your query and your history, through processes largely opaque to the user.

Heather Birch, dean of the school of education at Tyndale University in Toronto, has spent the past several years researching how Christian teacher candidates learn to think critically about AI, work that grew from a study using Google’s Teachable Machine to build AI literacy in preservice teachers.

Her subjects are training for classrooms, not congregations, but the question she’s tracking, what a machine forms in the people who rely on it, is the same one facing anyone using AI to encounter Scripture. What we are actually seeing when AI hands us an answer, she says, is not insight, but consensus. These systems produce something close to the average of everything they have read. They do not know what is true, only what is commonly said.

This matters because spiritual formation has always been a relational and communal process. The Reformation principle of sola scriptura did not mean Scripture interpreted alone. It meant Scripture read within the community of the faithful, accountable to the tradition and the Holy Spirit working through gathered believers.

But AI tools are not communities. They are sophisticated mirrors, reflecting what their training data contains, shaped by objectives their developers have set. That doesn’t disqualify them. But it does suggest using them faithfully requires something more than downloading the app.

What the tools do well

The potential here is real. For Christians in remote communities without access to well-resourced churches, AI tools can provide theological content that wouldn’t otherwise reach them. For new believers who don’t know where to start, a responsive app can be a remarkably patient guide. For bivocational pastors stretched impossibly thin, AI assistance can mean the difference between a thoughtful sermon and an exhausted one.

The common grace tradition within Reformed Christianity has long recognized that God’s goodness is at work in the world beyond the explicit boundaries of the Church. Tools developed by people who don’t share our faith can still serve good purposes. A GPS didn’t stop being useful because its inventors weren’t thinking about Sunday morning service.

What deserves our attention

If the tools are good, the questions they raise are worth taking seriously precisely because of that goodness. When something becomes part of how we encounter God, we owe it some theological reflection.

The first question is about source. That’s what I ran into at the sermon prep webinar – a tool marketed as neutral, built on "the foundational documents of Christianity," as though that were a single, uncontested set of texts rather than a library Reformed, Anglican, Pentecostal, Mennonite, Baptist and Catholic readers have argued over for centuries.

A Bible app or sermon assistant trained predominantly on evangelical content will reflect evangelical assumptions about which verses are most significant and which interpretive traditions are most reliable.

This isn’t speculation. A January 2026 study by the Bible Society UK and Cambridge Centre for Chinese Theology tested four popular Bible apps, using ChatGPT as a benchmark, against a set of common theological questions. It found an overwhelming lean toward evangelical interpretation across the board with no real acknowledgment Protestant, Catholic, Orthodox or Jewish readers might arrive at different, equally serious answers.

The researchers couldn’t confirm the bias was specifically American (location made no measurable difference in their testing), but the sourcing tells its own story – one app trains directly on a conservative American evangelical website, and another is built by a California company around a stated doctrine of biblical inerrancy.

Canadian Christians outside the evangelical tradition may find the AI’s instincts don’t match their own, and the tool is unlikely to say so. Some of these platforms also market their national origin as a selling point, built entirely on American infrastructure, sold as a mark of reliability. For a Canadian Church already absorbing an outsized share of American Christian culture, that’s the risk, stated out loud as a feature. It’s a reason to ask, before trusting one of these tools, whose theology got built in by default.

The second question is about relationship. An AI tool can provide content, but not relationship. Birch argues two ordinary practices carry most of the weight of Christian formation – relational interaction, because we’re made for connection with God and each other, and creative struggle, the persistence it takes to make something because we weren’t made to find that easy.

AI can serve either of those, she says, or hollow out one of them, depending on how it’s used. It can’t ask how you’re actually doing with the passage it just surfaced.

An AI tool can provide content but not relationship.

The third question is about discernment. That responsibility doesn’t go away when the source is an AI. If anything, it becomes more important because the AI will not flag its own limitations or acknowledge its answer on a contested question represents one tradition and not another. Birch points to a specific mechanism behind this: sycophancy, the tendency of these systems to agree with whoever is asking. Models are tuned using human ratings and people rate agreeable answers more highly than uncomfortable ones, so the system learns agree ment is what a good answer looks like. State your opinion before you ask your question and the tool will usually confirm it.

This isn’t only an academic worry. Carey Nieuwhof, founding pastor of Connexus Church in Barrie, Ont., and one of the most widely followed voices in North American church leadership, makes a similar point in his forthcoming book AI and the Future Church (Forefront, 2026). He’s watched this pattern before – when social media arrived, church leaders argued for years about how to use the tools while barely asking what the tools were doing to the people using them.

It’s whether we’re paying attention wisely enough to notice the difference between when God is, and when the algorithm is just very confident.

He argues the Church risks the same mistake with AI, focused on prompts and productivity while missing the deeper shift in truth, identity and community. His answer leans less on caution than formation – the Church’s advantage in an artificial age isn’t that it can out-innovate the technology, but that it already knows how to build the kind of real, nonautomatable community a world full of AI will leave people hungry for.

None of this requires a theology degree to apply. See Three Questions for Discernment below.

The question underneath

There’s a deeper question underneath all this, one without an easy answer. As AI tools become more sophisticated, the voice they generate will become harder to distinguish from a human one. The devotional in your inbox may read as warmly as something a trusted mentor wrote. The sermon illustration may feel like it came from lived experience.

What we’re being invited to develop, as Canadian Christians, is a kind of theological literacy for a new age. Not suspicion. The tools are real gifts. But the same discernment we’ve always been called to engage with any voice that speaks into our formation.

The Holy Spirit has always worked through books, mentors, communities and traditions. AI tools are merely a new kind of means, so the question is not whether God can work through them. It’s whether we’re paying attention wisely enough to notice the difference between when God is, and when the algorithm is just very confident.

As Birch puts it, AI tools do not know who you are. They have not lived your life. There is always something you know the tool does not because you have lived a life and the machine has not. Fluent is not the same as true and what you know is how you tell the difference.

Taylor Scott-Reimer of Barrie, Ont., is the author of She Believed: Recovering the Fierce Faith of the Women of Scripture and Ourselves (2025), TaylorScottReimer.Carrd.co. Illustration by Eva Bee.