A one-page church AI policy: which tools are approved, what data never goes in, who owns accounts, how the congregation is told, and when you review. Practical governance for executive pastors.
Somewhere on your staff, someone is already using AI — drafting the newsletter, summarizing a meeting, brainstorming sermon illustrations. The question in front of an executive pastor is not whether the church will use AI; it is whether the church will decide how, or find out how afterward. That's all a policy is: deciding once, in writing, before the incident that would have made you wish you had. Here is a one-page policy's worth of decisions, in the order they matter.
Most AI trouble in churches will not be a chatbot saying something weird in public. It will be a staff member pasting something private into a tool that shouldn't have it. So the first line of your policy is about data, not tools:
No private member information goes into AI tools not explicitly approved for it. That means: names attached to pastoral situations, counseling and prayer details, giving records, health information, anything about minors. If it wouldn't go in the church-wide email, it doesn't go in the prompt.
This single rule prevents the majority of realistic harms, and it costs nothing.
Keep a short living list: which tools are approved, for what, and who approved them. Three columns on one page. The point isn't bureaucracy — it's that "are we allowed to use this?" gets an answer other than a shrug. Anything customer-facing (a website assistant, anything that speaks to your congregation in the church's name) gets the highest bar: it should be grounded in your own content, cite its sources, and have published behavior for crisis situations. Staff-internal drafting tools get a lower bar but still live on the list.
AI accounts should belong to the church, not to whoever's personal email signed up first. Decide: who owns the accounts, who pays, who can add users, and what happens when a staff member leaves. Churches already know this lesson from social media — the youth account nobody can log into anymore. Same rule, new tools.
Decide what you disclose and say it plainly. The visible case: if your website has an AI assistant, people should be able to tell — and with AskMyChurch they can, since every answer cites the church content it came from, and hands off to a named human path when it doesn't know. The quiet cases matter too: a simple line like "our staff uses AI tools for drafting; everything published in the church's name is reviewed by a person" costs one sentence and buys durable trust.
Write down what AI never does at your church: it doesn't counsel people in crisis, it doesn't make pastoral decisions, it doesn't handle discipline or membership issues, it doesn't communicate hard news. Any AI touching your congregation should hand off to a person the moment content runs out or stakes rise. (This is the line we build to: the assistant's crisis check runs before any AI does, routes acute distress to 988, the Crisis Text Line, and your care team, and never attempts a pastoral response in a crisis.)
Date the policy and put a review on the calendar — every six months is plenty. The list changes, the tools change, and a dated policy signals it's maintained, not fossilized.
A new staffer should read it in five minutes and know: what data never goes in, which tools are fine, who to ask, what we disclose, and what stays human. If your draft does that on one page, stop writing — you're done. And when you evaluate a congregant-facing tool under it, ask each vendor for their answers in writing. A grounded, cited, crisis-gated assistant is what makes that a short conversation; it's the standard we built AskMyChurch to meet.
If anyone on staff uses AI tools — and someone already does — yes. A one-page policy beats an unwritten rule the day something goes wrong, and writing it forces decisions (what data, whose account, who approves) that are cheap now and expensive later.
The data line: no private member information — names tied to pastoral situations, giving records, counseling notes, kids' details — goes into any AI tool that isn't approved for it. Most AI mistakes in churches are data mistakes.
Yes, and it's easy: disclose the visible uses, like an assistant on the website, plainly. Trust survives disclosure; it doesn't survive discovery.
It's the easy case to govern: it answers only from your church's already-public website and sermons, cites its sources, and has published crisis behavior — so approving it doesn't require exceptions to your data rule.
Updated 2026-07-11 · AskMyChurch by Vision Genesis · Knoxville, TN
See it answer — try a live demo →