Where church Happens. Title by Grego Peña-Camprubí. Design by OpenAI.

When I started looking at how congregations use generative AI, I expected a story about a new tool. What I found, in addition to that, was a live theological argument about where the divine is allowed to move.

We asked more than 7,400 American congregational leaders how often their congregation uses generative AI.* Among the roughly 6,000 who answered that question, about two in three said: not at all. And for most non-users, AI simply hasn’t come up as a subject in their congregations at all.

We asked a smaller group to describe their congregation’s use of and attitudes toward AI in their own words. Reading those answers, theological objection turned out to be the rarest reason given for staying away. Environmental and ethical concerns outnumbered it many times over.

Congregation size predicted AI use more clearly than any other factor we tested. Among congregations with 50 or fewer people at weekend worship, 79% report no generative AI use whatsoever. Above 250, that drops to 24%. Size behaved like a dial, climbing steadily with every step up, whereas every other factor we looked at behaved more like a switch, on or off.  For example, the study found that church leaders under 60 were more likely to adopt AI (with adoption rates between 41% and 45%), while leaders aged 60 and older were less so (with an adoption rate at 27%).

Congregations using AI most heavily name the widest range of uses: newsletters, graphics, meeting minutes, translation, background research, and so on. While this is an interesting – if predictable – finding, what I’d like the reader to sit with is this: those same heavy-user congregations are, by a clear margin, also the most likely to name a limit — and the limit is theological. One leader shared: “We do not, and do not plan to, use AI to write sermons or to replace study. Those areas remain rooted in prayer, scripture, and human discernment.” Another: “Staff and elders use it a lot. We do not do any sermon work with gen AI.” A third said it is never allowed “to think for us or become the Holy Spirit in any way.“

“Heavy-user congregations are, by a clear margin, also the most likely to name a limit — and the limit is theological.”

So why not sermons? Is it fear of a confident, hallucination-prone machine? Perhaps — though it’s worth noting that a model trained closely on a tradition’s own texts might well produce something more doctrinally careful than a pastor’s quick Saturday-night draft. Is it about creative ownership, that is, the sermon as handiwork? Maybe, though I’ve heard no version of that concern raised about AI-generated bulletin visuals — something graphic designers might have a thing or two to say about.

To me, that third quote gets to the core of this reflection: where can the Holy Spirit move? Can it move through AI-generated content? These questions point to a larger, fascinating anthropological observation. The use of generative AI in congregations, particularly in liturgy and prayer, may be opening up a real opportunity for leaders and parishioners to interrogate and negotiate whether and how a technological tool can assist in the spiritual life and vitality of a congregation. It appears that AI might be more relevant than congregational leaders give it credit for; it’s showing itself, in real life, to be redrawing the relational map between the human, the divine, and the machine.

Perhaps it’s time to ask some new theological questions in the congregations you observe?

 

*Data: FACT 2025 / EPIC 6 Common Questionnaire, fielded Fall 2025, in partnership with Exploring the Pandemic Impact on Congregations.