What I Actually Do (and Don't Do) With AI in Sermon Prep

A few weeks ago, The Wall Street Journal ran a story about pastors and AI, and I was one of the pastors featured. Since then, I've been called a lot of things online — most of them by people who never read past the headline.

I enjoyed the 5 hour interview with the WSJ reporter on how I use AI for sermon prep. We even had the reporter and photographer over to my house for lunch after. She was friendly, curious, respectful, and thoughtful.

As you can imagine, in 5 hours, we covered a lot of ground. Only about 6 sentences from that 5 hours made it into the story... and few of my cautions and details about what I don't do.

But that didn't stop the eruption online. (Which I expected. More on that later!)

So let me say plainly what I told the reporter, and what I'll keep saying: I have never let AI write a sermon for me. But I want to be precise about what that means, because "I never touch AI" isn't my position either, and I'd rather you hear the real line from me than guess at it.

Here's my actual rule: I touch every single word that ends up on that page. Nothing survives onto my manuscript that I don't personally believe is

• an honest reading of what the passage actually means, and

• the best way to say what God wants said to my people, in my church, this week.

If a line doesn't meet both of those, I cut it. I don't care if it comes from ChatGPT or Charles Spurgeon. The tool or the author doesn't get a vote on truth or timing. I do.

That's the short version. Here's the long version, talking about:

• how I actually think about this

• why I don't think I'm doing anything new.

Why ChatGPT is inevitable & useful for sermons

Something genuinely changed in the last few hundred years, and it keeps accelerating.

It's often said — and while I can't point you to a rigorous, peer-reviewed source for the exact number, the idea is broadly true even as folklore — that the average person in past centuries might go a lifetime without encountering the volume of information in a single New York Times Sunday newspaper. Whether or not the precise comparison holds up, nobody disputes the direction: there are exponentially more books, more studies, more sermons, more quotes, more scholarship available to a pastor today than at any point in church history.

John Wesley knew something about this pressure two centuries ago — he told young preachers to read several hours a day, every day, or find another calling. (Whoa!) I'd give young pastors today the same advice. But even Wesley's "read broadly" assumed a world with a fraction of what's now published every single day.

Given that, Google becoming part of sermon prep wasn't really a choice pastors made — it was inevitable. Nobody could out-search a library index by hand anymore. AI is just the next step in the same direction: a faster, more conversational way to sort through more material than any one human being can hold in his head.

And it was a good thing! (Even with apocryphal stories and dubious quotes. Not like those didn't exist in books.)

Think about it the way you'd think about medicine. A doctor in 1850 could plausibly know most of what was knowable in his field. A doctor today can't — the literature is too vast, and it's still growing. If you go to a doctor to be treated, you don't ask him to "only help me from the studies in your memory or research you've read yourself." We don't object when they lean on a nurse or a resident to say "go check what the current best practice is for treating this." That's not laziness. That's stewardship of limited time in service of the patient in front of them. I think the same logic applies to a pastor and the text in front of him.

And frankly, the "just open your mouth and trust God to fill it" model was never how faithful preaching actually worked, for anyone, ever. Left un-studied, your brain will fill your mouth — usually with hot air.

Paul told Timothy to study, to labor, to be a workman unashamed. Wesley agreed. Spurgeon read constantly. Every serious preacher in history has been, among other things, a serious researcher. The question was never whether to gather information broadly. It's how it's gathered, and who does the gathering.

Effective Pastors have already been using research help (before AI)

This is already how large churches and some effective teachers operate — I'm just saying it out loud.

It's an open secret that pastors of larger churches have research help. Somebody combs through fifty possible illustrations and hands the pastor the five best ones to choose from. Often somebody helps with Greek and Hebrew word-studies as well. Nobody objects to that. Nobody says "how come the Holy Spirit isn't giving you what you need to know?" It's just teamwork.

That's exactly how I use AI.

First, I use it as a research assistant.

I ask for illustration options, word studies, historical background, cross-references — the same requests I'd make of a research assistant if I had a full-time one on staff — and then I choose, discard, and rewrite based on what I actually believe fits the text and my congregation.

The research assistant doesn't decide what goes in the sermon. I do. AI is the same arrangement, minus the salary. (I personally know two brilliant original language scholars -- top tier guys! -- who use AI prolifically and love it.) I've got a post here on how I do word studies with AI if you're interested.

The same logic covers two other roles that already exist in ministry and publishing, all the time.

Second, I use it as a writer who helps shape expression.

John Maxwell has worked with a writer, Charlie Wetzel, for decades. Countless authors and speakers work this way — the ideas and convictions are the author's own; the writer helps find the clearest, most memorable way to say them. When I ask AI to help me say something I've already thought through more clearly, that's the same relationship, just without a second person in the room.

This literally happened last Sunday. I was trying to figure out how to say a particular concept. AI gave me a sentence that was better, clearer and more memorable than the one I had in mind. I used it.

Incidentally, did you hear about the guy who said, "I'll be original or nothing!" ...and he was both?

Third, I use it as a graphic artist who executes a vision.

If I had a designer on staff and told her, "I want a slide that captures Dietrich Bonhoeffer's story of costly grace," nobody would say I hadn't really told the story myself. When I prompt AI to build that image instead, it's the same division of labor: I supply the idea and the theological content, the tool supplies the execution.

None of these three roles — researcher, writer, artist — is new to ministry. What's new is that one tool can now play all three, which is exactly why it feels different even when it isn't functionally different.

Does AI just make sermon writing bland?

Some people worry that AI averages everything down to the bland middle of whatever it was trained on. I understand the concern, but I think it gets the issue backwards.

These models were trained on a staggering amount of well-formed thought — centuries of careful writing, argument, and expression from some of history's clearest thinkers.

That's why AI is often good at expressing an idea plainly and memorably: it has absorbed a lot of good writing about how to do that. It's not incapable of quality. It's incapable of rejoicing in what it finds. Those are different problems, and only one of them is actually dangerous to preaching.

Sure, AI just simulates intelligence. Sure, it just predicts the next most likely word. And yes, that can have a "blandifying" effect on writing. But it can also get the 50% of us who are below average up to above average! Don't forget -- it's been trained on everyone who was a stunningly good writer for the last 1,000 years.

If you have a thought, it probably can help you shape it to be punchier, better, simpler, clearer, etc.

Why I'm glad I didn't have this for my first 21 years

This is the part I'd want every young pastor to hear directly from me:

I'm genuinely grateful I pastored for two decades before any of this existed. It forced me to read broadly and deeply, on my own, without a shortcut available. I'd tell a young pastor starting out today to do the same, on purpose, even though the shortcut now exists.

Here's why it matters practically, not just spiritually: if you've only ever encountered C.S. Lewis through quotes someone else — or something else — handed you, you won't recognize it when a "C.S. Lewis quote" is fabricated, misattributed, or subtly wrong. That happens constantly with AI, and it happens constantly with lazy internet quote culture generally. The only real defense is having actually read the man yourself. Broad, deep, personal reading the only quality-control system you'll ever have for handling this issue with AI. Skip that step early in ministry, and you'll never fully develop the discernment to catch when the tool is wrong.

I'll say that once more: Young pastors, DO THE WORK.

DO the reading. DO the thinking.

There's actually research that backs up why sequence matters here. A widely cited MIT Media Lab study on AI and writing found that people who did their own thinking first, and only brought in AI afterward, showed the strongest mental engagement with the material — their brains worked harder to integrate the tool's suggestions with thinking they'd already done.

People who leaned on AI from the very first draft, before doing any of their own thinking, showed the weakest engagement of any group, and struggled afterward to even explain what they'd supposedly written. Study first. Tool second. Never the other way around. That's not just my rule of thumb — it's measurable.

The mistake that keeps me careful

A couple of years ago I was rushing to finish a slide for a missionary story and dropped in an AI-generated image without really looking at it. After the service, a member asked me if I'd made the image myself — and then told me there was a topless figure visible in the background. (It was a story about a missionary in the South Pacific.) I hadn't caught it. I apologized and changed my policy. I never make changes last minute now without reviewing the images. (Of course, the models have gotten a LOT better since then as well!)

That's not actually a story about AI being dangerous.

It's a story about rushing being dangerous, and AI making it easier to rush. That was a lesson I needed to learn.

Every one of the roles I described above — researcher, writer, artist — still requires a pastor who checks the work before it goes in front of his people. I don't need to move that fast. Neither does anyone reading this.

Why I said yes to a controversial interview

A few days ago I was in an online disagreement on X, and a guy apparently Googled my name for ammunition. He gave me the "Ok, pastorGPT" insult. 😂

That's OK. I expected some of that. I'm not some hero, and I'm not a victim, it's more of a calculated thing.

I knew before I ever sat down with the reporter that this was a controversial topic, and that talking openly about it was a risk to my own reputation.

I said yes anyway, because as a culture we will never get this Pandora's box closed. AI use isn't a door we get to shut — pastors are already using it whether they talk about it publicly or not.

Barna's most recent research on pastors found that:

• only about 1 in 8 say they're using no AI at all,

• roughly a quarter now write or edit sermons with some help from it, (up from about 1 in 8 just two years ago.)

Most of those same pastors also describe themselves as cautious, even conflicted, about the whole thing. That's not a fringe story about me. That's 25% of American pulpit ministry.

Pope Leo XIV has spoken into this directly, urging priests to resist leaning on AI for their homilies and reminding them that to preach is to share faith — something he's said a machine will never do. I'm not Roman Catholic, but I don't strongly disagree with him.

AI will never be able to be anointed by the Holy Spirit. I understand the objections to using AI in preaching. I've written in some detail about it.

Given all of that, the question was never whether pastors should be stopped from using this. It's whether we're going to use it wisely or foolishly. Somebody has to go first and chart the terrain honestly, mistakes included, instead of pretending the technology isn't already in the room. That's really what the course I built is — a kind of Lewis and Clark expedition. I don't mean this in a self-aggrandizing way... like I'm some super courageous explorer.

But I do think that now that this huge new territory has been unlocked, someone's going to have to go map it, and if we take a few risks in the process, so be it.

I don't want pastors coming behind me to learn all this the hard way, or wake up in thirty years and regret how carelessly the Church adopted this.

The Primary Danger for Pastors who Use ChatGPT for Sermon Prep

I quoted Isaiah 29:13 to the reporter, and I meant it as a warning to myself, not a defense: "..people who draw near with their mouths while their hearts stay far away." 

That's the actual danger of any tool that lets you produce something true-sounding without ever being changed by it yourself — AI, commentaries, borrowed sermons, doesn't matter which. AI cannot do my job.

• It can't pray for me.

• It cannot meditate on a text and let it work on my soul.

• It cannot obey God on my behalf,

• It cannot walk by faith.

• It cannot love the truth it finds

• It cannot love my people.

Those things are the actual substance of pastoral work, and no research assistant, writer, or artist — human or otherwise — has ever been able to do them for me.

That's the whole issue, honestly. Not "does a pastor use AI," but "does a pastor still bring his whole self to the job."

Preaching is "truth through personality" Phillips Brooks said years ago. That means it is through a life that's actually been with the text before it's spoken from the platform. AI can hand you research. It cannot make your heart live that truth.

Pastor, guard your heart!

What if I want to Learn How to Sermon Prep With AI?

I get asked often — usually by pastors who are curious but afraid of messing it up — how to actually build guardrails around this instead of either avoiding AI entirely or using it carelessly.

That's the whole reason I built a course walking through it session by session: what AI is legitimately good for, where the ethical pitfalls actually are, and word-for-word prompts for the parts that are genuinely just time-savers, not shortcuts around the work. If that's useful to you, it's linked below. If not, I hope this post at least clarified where I actually stand.

Either way — my door's open. If you disagree with how I'm doing this, I'd rather talk about it than argue about it in a comment thread.

When I launched the Sermon Illustrator tool, I heard from so many pastors who loved the idea—but said:

“I don’t know how to use it.”

So I built this course:🎓 How Pastors Can Use AI for Sermon Prep (Without Selling Out or Cheating)

✅ It includes 6 video lessons

✅ A 40-page workbook

✅ Tutorials for creating sermon illustrations, visuals, quotes, outlines, and more

✅ Lifetime access to the Sermon Illustrator

✅ All for just $97