I need to tell you about the moment that broke something in me.
A pastor I know, a guy I respect, someone who genuinely loves his people, sent me his sermon outline and manuscript. He wanted feedback. I started reading, and within a few paragraphs I knew something was off.
The theology was tight. The illustrations were vivid. The application was specific. But it wasn't him. I've heard this guy preach. I know his voice, his rhythms, the way he circles back to a point three different ways because he wants to make sure it lands. This wasn't that.
It was too polished. Too even. Too perfect.
"Did you use AI for this?" I asked.
Long pause. "Yeah. I gave it my notes and asked it to write the manuscript."
My heart sank. Not because he used a tool. I use AI tools every day. But because I could tell. And if I could tell, his congregation could tell. They just might not know what they were sensing.
The Uncomfortable Question
Here's what kept me up that night: if AI can deliver a sermon that's theologically sound, emotionally resonant, and practically applicable, what exactly are we paying pastors for?
I don't ask that flippantly. I'm an ordained pastor. I've sweated through sermon prep at 2 AM. I've rewritten introductions six times. I've stood in a pulpit with a lump in my throat because the text hit me before it hit anyone else.
And now a machine can do it faster, with fewer grammatical errors, and without the nervous throat-clearing.
But Here's What AI Can't Do
It can't sit with you in the hospital at 3 AM when the doctors say there's nothing more they can do.
It can't notice that you've been sitting alone in the back pew for three weeks and something's wrong.
It can't break bread with you and look you in the eye and say, "This is the body of Christ, given for you."
It can't cry at your mother's funeral.
The word the church has always used for this is incarnation. God didn't send a memo. He didn't post a blog. He showed up. In flesh. In a specific place, at a specific time, with specific people.
That's what pastors do. Not perfectly (Lord knows, not perfectly) but we show up.
The Delivery Trap
Silicon Valley wants you to believe that content is king. That if the information is right, the medium doesn't matter. That a perfectly crafted AI sermon is functionally identical to a human one.
I think that's a lie from the pit of... well, from the pit of Palo Alto.
Here's why: a sermon isn't information delivery. A sermon is an encounter. It's a particular person, formed by a particular life, standing before a particular community, speaking a particular word that the Holy Spirit makes alive in that particular moment.
The stuttering, the tears, the awkward pause when the pastor loses their place? Those aren't bugs. They're features. They remind us that this word comes through flesh.
That's not a weakness. That's the whole point.
So What Do We Do?
I'm not saying ignore AI. I use it. I've written a whole book of AI prompts for pastors. The tools are genuinely helpful for brainstorming, research, finding illustrations, drafting outlines.
But there's a difference between using AI as a tool and outsourcing the spiritual work of preparation to a machine.
The difference is this: when you sit with a text all week, wrestle with it, pray over it, let it read you before you try to preach it to anyone else, that process changes you. And a changed preacher preaches differently than an AI ever could.
People don't just need good sermons. They need a shepherd who has been with God that week and comes bearing the marks of it.
The Real Question
The question isn't "Can AI preach better than my pastor?"
The question is: "What am I actually looking for on Sunday morning?"
If it's a polished TED talk, yeah, AI wins.
If it's an encounter with the living God through the stumbling, beautiful, incarnate presence of another human being who has been broken open by the same text? That's something else entirely.
And I think that's what the church has always been about.
AI can write a sermon. It can't be a pastor. And until we understand the difference, we'll keep asking the wrong questions about both.