Every year I help run the State of AI in the Church survey. Three years in, the question church leaders ask me has changed completely.

Here's where things stand in the 2026 results:

• 93.5% of church leaders are actively engaging with or exploring AI in ministry
• 2.4% say they're avoiding it entirely
• 78% use AI weekly or daily
• 43% are in these tools every single day

Two years ago, 13% told us they were opposed to using AI for church work at all. That group has nearly disappeared.

So the "should we" conversation is over. Nobody asks me that anymore. What I hear now is a harder question: how do we do this well?

I keep coming back to the same answer. We already have the framework. It's a few thousand years older than the technology, and it holds up better than anything I could write for you.

It comes down to three verses.

1. Ask God First (James 1:5)

"If any of you lacks wisdom, let him ask God, who gives generously to all without reproach, and it will be given him." James 1:5, ESV

James starts from the assumption that you're going to come up short. He doesn't scold anybody for it. Ask God, who gives generously and without reproach.

That verse gets at something I think about a lot with these tools. Information and wisdom are two different things.

AI is genuinely excellent at information. Ask it for ten discussion questions for your small group leaders and you'll have ten in about four seconds. A few will be usable.

Then ask which of those ten your groups actually need this month, after the year your church has had. You get nothing worth having. The model has never met your people. It answers exactly what you typed, and it has no way to tell you that you asked the wrong question.

That part is still your work.

Before You Open the Tool

Get clear on what you're actually after. These two prompts produce completely different work sessions:

• "Write an email to our volunteers"
• "Help our volunteers who are running on empty hear that what they do matters"

Thirty seconds of prayer before the prompt changes what you type.

One more question worth asking, and I ask myself this all the time: am I reaching for this because it's faster, or because I'm avoiding something? Speed is a fine reason. Avoidance is worth a second look.

2. Test What Comes Back (1 Thessalonians 5:21)

"but test everything; hold fast what is good." 1 Thessalonians 5:21, ESV

Paul wrote this about prophecy in the gathered church. Content that sounded spiritual, arrived with authority, and still had to be examined before anybody acted on it.

That describes an AI answer pretty well.

Our survey found 75% of church leaders name theological misalignment as their top ethical concern. That's the worry that AI-generated content will sound authoritative while being doctrinally wrong. I share it. Here's where I see it show up.

Made-up quotes. You'll get a line attributed to Spurgeon or Lewis or Bonhoeffer with a believable book title behind it. Some are real. Others were assembled by a system predicting what a quote like that would sound like.

Invented Greek and Hebrew. A model will hand you a term, a transliteration, and a semantic range, all delivered with total confidence, all wrong. This is the one I'd watch closest, because the people who could catch it aren't in the room while you're working.

Flattened doctrine. Ask about baptism or election or spiritual gifts and you'll often get a smoothed-over consensus that belongs to no actual tradition. It reads as balanced. It represents nobody's confession, including yours.

The 5-Minute Verification Habit

Keep using the tools. Build this habit around them and never break it:

• Open every Scripture citation and confirm it says what the model claimed
• Check every original-language claim in a real lexicon
• Trace every quote to an actual work with an actual page, then delete what you can't find
• Measure every doctrinal statement against your own confession
• Read the whole thing out loud once, listening for anything you wouldn't say

Five minutes. A lot cheaper than a correction two weeks later.

3. Trust Him With the Direction (Proverbs 3:5-6)

"Trust in the LORD with all your heart, and do not lean on your own understanding. In all your ways acknowledge him, and he will make straight your paths." Proverbs 3:5-6, ESV

Solomon's warning is about leaning on your own understanding. We've just found a new way to do it, by leaning on a machine's understanding instead.

Using AI to think through a problem is fine by me. The trouble starts when you hand it the decision, and it happens gradually, because the output sounds reasonable every single time.

Some things stay with you no matter what any model produces:

• Whether to confront a leader
• What to say at a graveside
• Who your church is called to reach over the next five years
• What a specific family needs in a specific crisis
• Whether what you're sensing in prayer is the Spirit or your own ambition

None of those are information problems. They take presence, history with actual people, and a relationship with God. No tool has any of that. A model will still produce an answer for every one of them if you ask. It'll sound wise. And it'll come from a system with no knowledge of your people and no stake in their souls.

Your Congregation Is Already Doing This

Barna's November 2025 survey of 1,514 U.S. adults found that 34% of practicing Christians agree spiritual advice from AI is as trustworthy as advice from a pastor. Among Millennials it climbs to 44%.

The same research found 31% of practicing Christians want guidance from their pastor on how to navigate AI, while only 12% of pastors say they feel comfortable teaching on it.

Your people are asking for this. Most of them are still waiting.

Put It To Work This Week

Three questions before your next significant prompt:

1. Have I asked God for wisdom here, or only asked the tool for output?
2. What in this response have I actually verified?
3. Am I using this to think, or to avoid deciding?

Then give your team something written. Our survey found only 9% of churches have a formal AI policy, so almost everybody is using these tools every week with no shared standard at all.

A one-page document your staff will actually read beats a twelve-page policy nobody opens. Start with four lines:

• Ask for wisdom before you prompt
• Verify before you publish
• Keep pastoral decisions with pastors
• Give everything AI-assisted a human read before it reaches the congregation

This is also worth teaching from the front sometime this fall. Your people are already using these tools without any theological framing, and only about one in eight of your peers feels ready to say anything about it. Being one of the few who does costs you one message.

The framework was written long before any of this arrived. Ask God for wisdom, test what comes back, and keep the direction in his hands. The rest is application.

Want to think through this alongside other church leaders? Join us free at AIforChurchLeaders.com/group.