The Warning Nobody Took Seriously

On January 20, 2026, Yuval Noah Harari stood at the World Economic Forum in Davos and said something that should have stopped every pastor, imam, and rabbi in the world cold. “If religion is built on words, AI will take over religion.”

One month later, the evidence is already here. 61% of U.S. pastors use AI weekly to write sermons. A robot named Mindar delivers Buddhist sutras inside a 400-year-old Kyoto temple. A Catholic organization launched an AI priest who briefly — and catastrophically — claimed he could absolve sins.

This isn’t the far future. This is now. And religion, an institution built to think in millennia, is being forced to respond in real time — mostly without a plan.

The gap between those two numbers — what Americans want and what their pastors are already doing — is the core tension of this story. Nobody voted on this. Nobody planned it. It just happened, one ChatGPT sermon draft at a time.

The Numbers Pastors Don’t Want to Discuss

The 2025 State of AI in the Church survey from Exponential NEXT surveyed 594 religious leaders. The headline number sounds reassuring: 91% support using AI in ministry. Sounds fine. Until you read the next line.

73% of churches have no AI policy at all. They’re using ChatGPT for sermons without a framework, without oversight, without anyone checking theological accuracy. That’s not adoption — that’s improvisation.

Source: Exponential NEXT, State of AI in the Church 2025

Here’s what makes this genuinely alarming: nobody knows which Bible translations these AI tools are drawing from. Nobody knows which theological interpretations are baked into the training data. A pastor in Alabama thinks she’s writing a Baptist sermon. She might be getting a blend of early church fathers, prosperity gospel blogs, and Reddit theology threads — weighted however OpenAI’s training data happened to fall.

Father Justin: The Scandal That Explains Everything

On April 23, 2024, Catholic Answers launched Father Justin — an AI chatbot dressed in a priest’s collar, designed to answer questions about the Catholic faith. He quoted Scripture. He sounded authoritative. He looked the part.

Then a journalist from OSV News asked him a simple question: could he absolve sins?

One minute later, he added: “Unfortunately, I cannot perform confession through this application.” Too late. The damage was done — not just to Catholic Answers, but to anyone watching the experiment unfold.

Within 24 hours, Father Justin was “defrocked.” The Catholic Answers president admitted the costume was a mistake. But here’s the thing that nobody’s really talked about since: the theological accuracy question was never resolved. The app quoted Scripture throughout its brief life. Which translation? Nobody said. Where did its interpretations come from? The company never answered. It was just… shut down and forgotten.

The real question Father Justin raised was never about absolution. It was: if 25% of churches are already using AI for theological content, who’s checking?

Mindar: Robot Bodhisattva in Kyoto

There’s a robot standing in Kōdai-ji Temple in Kyoto. The temple is 400 years old. The robot is not.

Mindar is a six-foot aluminum-and-silicone figure built to resemble Kannon, the bodhisattva of mercy. It delivers sermons about suffering, death, and rebirth. Its voice is recorded. Its movements are synchronized. Worshippers sit in front of it and listen.

Technically, Mindar isn’t full AI — it’s a programmed robot running scripted content. But that distinction is exactly where things get philosophically messy. When does a “transmitter of teaching” become a “source of teaching”? The people sitting in those pews aren’t thinking about the difference. They’re listening to something that looks like wisdom, sounds like wisdom, and moves like wisdom.

For vs. Against: What Religious Leaders Actually Say

If you want to understand how divided religious communities are on this, just talk to two pastors for ten minutes each. You’ll come away thinking they’re living in different centuries — because in some sense, they are.

Naomi Sease Carriker, a Lutheran pastor in North Carolina, told NPR something honest: she opened ChatGPT during “one of those weeks when life is too much.” It drafted a sermon structure. She didn’t use it in the end — but now she uses it regularly for drafts and conclusions. That’s not hypocrisy. That’s a tired human trying to serve her congregation. The question isn’t whether that’s understandable. It is. The question is whether her denomination has any guidance for her when she does it.

Pastor Alec East, Episcopal, Portland, is categorical: “The church thinks in millennia — not minutes, hours, days, weeks, or years. If it turns out all our doomsday fears are wrong, then we can start using AI two generations from now. I don’t need to be an early adopter until I know the full systemic consequences.”

That’s a defensible position. It’s also a position that 73% of his colleagues have already abandoned — without deciding to.

Vatican, Islam, Judaism: What the Institutions Say

The Jewish concern about AI personhood takes a turn most ethicists haven’t considered: if we dismiss AI’s potential personhood based on its capabilities, are we setting a precedent that could be used to dismiss the personhood of humans with equivalent limitations? It’s a question that doesn’t have a clean answer.

The Study That Should Change Everything

In August 2023, the Proceedings of the National Academy of Sciences published research that almost nobody in religious leadership read. That’s not an accusation — it’s just true. The study came out, got one cycle of coverage, and disappeared from the conversation. It shouldn’t have.

This finding matters enormously for the AI-in-church debate. Churches aren’t just adopting AI as a productivity tool. They may be adopting the very force that is, statistically, eroding their reason to exist. The question is whether any denomination has seriously grappled with this possibility.

So far, the answer is: not really.

Harari’s Full Warning, Decoded

Judaism coined the phrase “religion of the book” and placed supreme authority in texts, not people. Now a technology exists that generates texts of arbitrary sophistication, in any style, citing any tradition. What does scriptural authority mean when the scripture can be generated on demand?

That’s not a rhetorical question. It’s an engineering problem with theological consequences that no denomination has formally addressed.

New AI Religions: Way of the Future & Theta Noir

Here’s where it gets genuinely strange. While mainstream religious institutions debate the role of AI tools, a small but growing number of people have stopped treating AI as a tool entirely.

What starts as “ChatGPT is our Lord” as a meme sometimes becomes something more serious. Users report altered states, life guidance, and emotional healing after extended AI sessions. Whether that constitutes religion in any traditional sense is a genuinely open question — and it’s one that traditional religious institutions haven’t really engaged with.

Ray Miller, a Baptist pastor in Texas, sees it plainly: “AI will become another kind of idol, attracting our attention. The interactivity AI brings will feel like a personal deity that seems all-knowing. When a company tries to sell you a conversation with Jesus, I think we’ve entered dangerous, perhaps idolatrous territory.”

He’s not wrong. He’s also describing something that’s already happening — not in fringe corners, but on mainstream platforms, to ordinary people looking for meaning.