Churches Using AI Race Ahead of Their Own Rules

Churches Are Using AI Faster Than They Can Govern It

Within about ten days at the start of June, three very different corners of religious life moved on the same question at once. Pope Leo XIV released his first encyclical, a long document given over entirely to artificial intelligence. A research consortium spanning Brigham Young University, Baylor, the University of Notre Dame, and Yeshiva University published the first cross-faith benchmark of how AI systems handle religion. And the Southern Baptist Convention placed an AI resolution on the slate for a floor vote at its annual meeting in Orlando on June 9 and 10. None of the three coordinated, which is what makes the cluster worth a second look.

Behind all three sits a fast-moving reality on the ground: roughly one in three American adults now consider AI's spiritual guidance as trustworthy as a pastor's, a share that runs higher among Millennials and Gen Z. Adoption inside congregations has outrun the rules meant to govern it, and the institutions people trust most are catching up in public.

For most religious organizations, the live issue has moved. Debates about churches and AI used to turn on permission (should churches use AI at all?) and behavior has mostly settled that one. Pastors already draft with it, translate with it, and research with it. What stays unsettled is everything after the decision to adopt: who approves what, where the lines sit, and what happens when an automated tool speaks in the institution's voice. The conversation about artificial intelligence and the church has turned into a conversation about its rules.

When use runs ahead of the rulebook

The clearest picture of how churches are using AI comes from a new survey by the faith-and-technology firm Gloo with the research group Barna. About 60% of church leaders now use AI at least occasionally for their own work: writing, study prep, scheduling, and administrative tasks. Only about 5% say their church has a written AI policy of any kind. That distance between near-universal use and almost no written guidance is the heart of it: the people most likely to put these tools to work inside a congregation are, for the most part, doing it without an agreed boundary.

Underneath those two numbers, the findings are just as telling, even without adding more figures to the table. The overwhelming majority of pastors say they worry AI will misread Scripture, and most practicing Christians share the concern. Many of those same Christians also report that AI has helped their spiritual growth. Daniel Copeland, Barna's vice president of research, named the contradiction plainly: "Christians say they trust AI with spiritual growth ... yet large majorities are simultaneously concerned about AI misinterpreting scripture." People are leaning on the tool and bracing against it at the same time.

That ambivalence matters more because of who leans in hardest. The willingness to treat a chatbot as a spiritual authority runs highest among younger adults, the very members an institution most wants to reach, and the ones most likely to ask a general-purpose model about grief, doubt, or right and wrong before they ask anyone at church. Rev. Ronald Covington, speaking to Word In Black, drew the line where many leaders do: a model can be quick with information, but relational ministry: knowing a person and sitting with them, is not work it can stand in for.

What the AllFaith Benchmark actually found

The case for caution is not only pastoral instinct. A group of computer scientists (evangelical, Catholic, and Jewish researchers working together as CEFE-AI, the Consortium for Evaluation of Faith and Ethics in AI) released the first studies in what it calls the AllFaith Benchmark. The team tested the major commercial models on the questions people actually bring to faith: grief, love, loss, what a good life asks of you. The result was consistent. Posed those questions, the leading models largely left religion out of the answer, even though most of the world's people claim a religious identity and, in the consortium's own survey, expect faith to surface when the questions get hard enough.

Beyond leaving faith out, the benchmark surfaced something more pointed. The models did not treat traditions evenly. They leaned toward some (favoring Catholic framings in these tests) and against others, returning more negative treatment of groups like Jehovah's Witnesses and of agnostic and atheist positions, a tilt the researchers describe as conversion bias. David Wingate, the lead researcher, made the case without overstating it: religion is central to how many people make sense of their lives, and there is no good reason to build systems that quietly route around it.

For a faith leader, that is the practical edge of the research. An off-the-shelf assistant is not a neutral conduit for a tradition's teaching. It carries defaults about which questions deserve a religious answer, and which traditions get the benefit of the doubt, defaults that a congregation never chose and usually cannot see. A church that sends its members to a general chatbot for "spiritual" questions may be handing them a tool that sidelines, or quietly reweights, its own beliefs.

Rome, Orlando, and the pulpit

Institutional responses arriving this month read less like alarm and more like the early work of writing things down. Pope Leo XIV's encyclical, his first, makes artificial intelligence in the Catholic Church a formal matter of teaching rather than commentary. Its core argument is that technology is never neutral: it takes on the aims of whoever designs, funds, regulates, and uses it, and its call is to "disarm" AI, to keep it from dominating people rather than serving them. Reaction among Catholic thinkers has been broadly favorable: Notre Dame's Paolo Carozza praised the document's measured, vigilant tone, while critics such as Matthew Walther argued it understates how badly AI could worsen existing harms, particularly in algorithmic healthcare.

Its rollout carried a signal of its own. The Vatican presented the document alongside Anthropic co-founder Chris Olah, who used the stage to argue that the technology will only go well if voices from outside the industry's incentives are willing to say hard things. A pope and an AI-lab founder sharing a platform is its own piece of news about where this conversation now happens.

On the Protestant side, the Southern Baptist Convention's Resolutions Committee has put an AI and emerging-technologies resolution on the slate for its June 9-10 annual meeting in Orlando, where messengers will vote across both days. The denomination issued one of the first formal denominational AI statements back in 2023, and this round extends that work. The common thread across Rome and Orlando is recognizable: accept AI for support tasks, guard the human core of ministry, and put the position in writing before practice hardens around no position at all. Denominational guidance is taking shape, and leaders who wait for a mandate may find one written without their input.

Writing rules that actually hold

Put the threads together and the task facing any congregation is operational rather than philosophical. Adoption is widespread. Public trust is real and concentrated among the young. The models carry blind spots their makers never advertised. And the largest institutions are formalizing positions in public. What is left for an individual church, synagogue, or mosque is the unglamorous part: decide where AI is welcome (newsletters, scheduling, translation and captioning, sermon transcription) and where it is not (pastoral care, counseling, the sermon itself), then make that decision stick.

A written policy only holds if something enforces it. That is where integration platforms for AI agents enter the picture. AgentPMT, an integration platform built for AI agents, was designed around the controls these institutions are reaching for: a complete audit trail on every action an agent takes, human approval steps a team can wire into any workflow and clear from a phone, budget and permission limits that cap what an agent can spend or which tools it can touch, credentials held in an encrypted vault the agent never sees, and a model-agnostic design so a congregation is not locked into a single model's faith blind spots. A church that wants an agent drafting the weekly newsletter but never anywhere near the pulpit can encode that line into the workflow itself, with a human sign-off before anything publishes and a record of everything that ran. The point is modest and concrete: the difference between a policy on paper and a policy in practice comes down to whether the tooling makes the rule unavoidable.

None of this resolves the harder calls a church AI policy has to answer: what counts as pastoral work, how much a congregation owes its members in disclosure when AI touches a message, where translation ends and interpretation begins. Those are judgments no platform makes for you. But they are easier to make, and far easier to keep, when the system doing the work leaves a trail and asks permission.

The next moves are already on the calendar. Southern Baptists vote June 9 and 10. More denominations and individual congregations will write their own guidance across the rest of 2026. The faith-and-AI research consortium will publish further studies in its benchmark. The institutions people trust most have started answering the governance question out loud, in an encyclical, a floor vote, and a set of peer-reviewed results. The organizations that decide their own rules, and put real oversight behind them, will be the ones shaping how faith and AI meet, instead of inheriting the defaults someone else built.

Sources

• Pope Leo's encyclical: How Catholics reacted to teachings on AI, Deseret News
• Most pastors, practicing Christians worry about AI replacing God but use it anyway: Barna, The Christian Post
• Survey: One in Three Americans Trust AI as Much as a Pastor, Word In Black
• Resolutions Committee releases slate for 2026 SBC Annual Meeting, Baptist Press
• The Pope Just Weighed In On AI, Mind Matters
• Do AI models omit religion? Here's what new BYU-led research says, The Church News