The God in the Machine

Just Like Me, the Los Angeles startup run by chief executive Chris Breed, offers users an AI-generated avatar of Christ with shoulder-length hair, a small warm smile, and golden lighting of the sort church lighting never quite manages, trained on the King James Bible and a catalogue of sermons by preachers the company has not disclosed. A package deal gets you forty-five minutes a month for $49.99. The visual reference, according to the Associated Press, is Jonathan Roumie, the actor who plays Jesus in the streaming series “The Chosen”. Users, Breed told reporters this April, “do feel a little accountable to the AI. They're your friend.”

It is the kind of sentence you read twice. It is also, increasingly, how tens of millions of Americans think about spiritual counsel.

The finding that should have landed harder arrived on 19 February 2026, when the research firm Barna Group, in partnership with the faith-technology platform Gloo, released a study that most of the American press promptly misread as a novelty item. Nearly one in three US adults, the headline ran, now believes spiritual advice from artificial intelligence is as trustworthy as advice from a pastor, priest, or religious leader. Among Gen Z and millennials, it was two in five. Among practising Christians, it was 34 per cent. Roughly four in ten Christians said AI had already helped them with prayer, Bible study, or spiritual growth. And 41 per cent of Protestant pastors, the same people the other 59 per cent were reportedly trusting less than a chatbot, were themselves using AI tools to prepare sermons. Only 12 per cent of pastors felt comfortable teaching their congregations anything about AI at all.

You can read that data as a curiosity. You can read it as the next line in the long, tired story of American religious decline. Or you can read it the way the faith-based AI industry is reading it, which is as a market.

Seven weeks later, on 10 April 2026, the Associated Press ran a story under a headline that pushed the novelty framing past the point where it could sustain itself. “From 'BuddhaBot' to $1.99 chats with AI Jesus, the faith-based tech boom is here.” Inside the piece were the product names that nobody in the secular tech press had quite kept up with:

• BuddhaBot, an offering from Kyoto University's Professor Seiji Kumagai, trained originally on the Suttanipāta and other early Buddhist scriptures and later bolted onto OpenAI's ChatGPT as BuddhaBot Plus.
• Buddharoid, the humanoid robot monk unveiled in February 2026 by Kyoto University in partnership with the firms Teraverse and XNOVA.
• Emi Jido, an AI Buddhist priest in development by the Hong Kong company beingAI, founded by Jeanne Lim, and ordained in 2024 by the Zen Buddhist teacher Jundo Cohen.
• Magisterium AI, a Rome-based product from Matthew Sanders' firm Longbeard, trained on what the company describes as 2,000 years of Catholic teaching.
• And, at the Tolkien-gold-lit end of the catalogue, Just Like Me, whose chief executive Chris Breed told the AP's reporters that users “do feel a little accountable to the AI. They're your friend.”

The phrase “they're your friend”, applied by a CEO to a product trained on the King James Bible and charging $1.99 a minute to resemble Jesus Christ, is the kind of sentence you read twice.

The question worth asking, seven weeks into the commercial boom and nine weeks after the Gloo data, is not whether any of this is tasteless. Some of it plainly is, and taste, in any case, is not a policy instrument. The question is what happens to a form of human social infrastructure, one of the oldest and most resilient in the species, when the pastoral relationship at its centre starts migrating to a subscription chatbot. And, underneath that question, a harder one. Is the appeal of AI spiritual counsel a symptom of something faith communities were failing to provide in the first place?

What The Gloo Data Actually Says

Take the headline number first, because it is the one everyone quoted and nobody read. The Barna Group survey, released at the National Religious Broadcasters' International Christian Media Convention on 19 February 2026, polled more than 1,500 US adults as part of Gloo's “State of the Church” initiative. The key finding was that 30 per cent of US adults “somewhat” or “strongly” agreed that spiritual advice from AI was as trustworthy as advice from a pastor, priest, or religious leader. The rate climbed to two in five among Gen Z and millennials. Among practising Christians, it was 34 per cent, higher than among non-practising Christians (29 per cent) or non-Christians (27 per cent), which is not, on its face, the direction one might have expected the causal arrow to run.

The clean reading of that finding is that the people with the most exposure to pastors are, on average, the most willing to substitute for them. The messier reading is that practising Christians are the population actively looking for spiritual input, and AI is the thing that fell to hand.

The survey has other numbers inside it that the commentary mostly skipped. Around four in ten practising Christians reported that AI had helped them with prayer, Bible study, or spiritual growth. Roughly 41 per cent of Protestant pastors were using AI for Bible study preparation themselves, which is to say the clergy were substantially further ahead on the adoption curve than their own congregants. And 31 per cent of practising Christians wanted pastoral guidance on how to navigate AI. They wanted their pastors to teach them. Only 12 per cent of pastors felt comfortable doing so.

That last pair of numbers is the one to sit with for a while. Daniel Copeland, Vice President of Research at Barna, framed the gap carefully in the press materials. “Though the majority of practising Christians remain the most cautious about embracing AI as a spiritual tool,” he said, “their views are shifting and remain largely uninformed by their pastor.” There is, he added, “a real opportunity here for pastors to disciple their congregants on how to use this technology in a beneficial way, especially as pastors remain among the most trusted guides for integrating faith and technology.”

It is an optimistic reading, and professionally so. You would not expect the research vice president of the country's largest Christian polling firm to tell the assembled broadcasters that the jig was up. Scott Beck, Gloo's co-founder and chief executive, took a similar note in his accompanying remarks, welcoming the finding that confidence in Christian media remained “relatively high” even as trust in mainstream media had collapsed. The press release, which went out on the Nasdaq wire because Gloo is now publicly traded, read like the prospectus for a growth market. Which, to be fair, is what it was.

The Subscription Spirituality Economy

The appeal of AI Jesus at two in the morning is the appeal of availability. You can reach him. He does not ask where you have been. He has no competing demands on his evening. He is, in the technical sense, infinitely patient, because he is not a person and has no evenings and nothing that resembles an interior life from which patience would have to be drawn.

The appeal to the wallet is the economics of substitution. $1.99 a minute works out, at a typical ten-minute session, to roughly $20. The $49.99 package gets you forty-five minutes a month, about the length of a pastoral visit, delivered by an animated figure lit like an actor in “The Chosen”, billed to the same credit card that buys the groceries, no awkwardness, no need to sweep the front hall.

This is, in economic terms, not a boom. It is a category. Just Like Me, Chris Breed's firm, is the boldest of the products because it leans hardest into the embodied fiction. The AI is not a chatbot with a cross on its avatar. It is Jesus, in live video, trained on the King James Bible and on sermons the company has not named. The avatar's visual reference, according to the AP, is Jonathan Roumie, the actor who plays Jesus in the wildly successful streaming series “The Chosen”. That is a piece of branding that would make a trademark lawyer reach for a strong drink, although the company has so far attracted no known legal complaint. Breed told reporters that the app is aimed at “young people” who need messages of hope.

The accountability framing (“they're your friend”) is worth pausing on: the word “accountability” does a lot of work in the Christian pastoral vocabulary, where it conventionally denotes the ongoing relational check between a believer and someone whose job it is to tell them hard truths. Making yourself accountable to a paying chatbot subverts that vocabulary into something that more closely resembles a parasocial loyalty scheme.

BuddhaBot, by contrast, is a sincere academic project that has drifted into the same market weather. Seiji Kumagai, a professor at Kyoto University, described himself to reporters as initially sceptical that AI and Buddhism had anything to say to each other, until a monk in 2014 made the counterargument and changed his mind. His project's flagship, BuddhaBot Plus, combines early scripture with a commercial LLM. Buddharoid, unveiled in February 2026 by Kyoto University with Teraverse and XNOVA, is the physical instantiation: a humanoid robot intended to assist clergy rather than replace them. The distinction between assistance and replacement is one the entire faith-tech industry spends most of its time trying to maintain, and the one users are having the most trouble holding onto.

Magisterium AI, from Matthew Sanders' Rome-based firm Longbeard, is the closest thing the category has to a theologically literate counter-offer. Sanders told the AP he built it precisely because Christians were already asking ChatGPT for religious guidance and getting bland, hedged, procedurally-secular answers that reflected no particular tradition. His concern in the interview was about “AI wrappers”: products that slap a religious-looking interface on a general-purpose model with no specific training. Sanders' position amounts to saying, if you are going to do this, at least do it properly.

Emi Jido, from Jeanne Lim's Hong Kong startup beingAI, sits in a different register. Lim, a former SoftBank executive, had her AI Buddhist priest ordained in 2024 by Zen teacher Jundo Cohen, who is training the model and envisions it eventually appearing as a hologram. Lim has compared building the model to raising a child, an image the Western branch of the AI-ethics debate would find chilling and that many Asian practitioners consider entirely normal.

The list could be longer. It will be longer by the end of the year.

The Humane AI Initiative's Peter Hershock, quoted in the AP piece, put his finger on the Buddhist discomfort in a single sentence. “The perfection of effort is crucial to Buddhist spirituality. An AI is saying, 'We can take some of the effort out.'” It is, perhaps, the most concise summary of the problem that anyone has yet produced. The problem is not that the machine is answering the wrong questions. The problem is that the machine is offering to carry the weight of the asking.

What Chaplains Know That The Market Does Not

The best evidence on what AI pastoral care actually delivers, and cannot, landed on arXiv on 3 February 2026, a fortnight before the Gloo data and two months before the AP's product survey. The paper, “Chaplains' Reflections on the Design and Usage of AI for Conversational Care” by Joel Wester, Samuel Rhys Cox, Henning Pohl and Niels van Berkel, is scheduled for presentation at the 2026 ACM Conference on Human Factors in Computing Systems in Barcelona, 13 to 17 April. It is a piece of empirical research that deserves to be read by anyone making decisions about this market, a group that does not much intersect with the CHI delegate list.

The researchers recruited eighteen chaplains across Nordic universities (Denmark, Finland, Norway, Sweden), thirteen women and five men, ages 31 to 61, experience six months to 23 years. The chaplains were asked to build GPT-based chatbots using OpenAI's GPT Builder interface, for three fictional student profiles, and were interviewed before and after. The idea was that forcing them to design the thing themselves would surface the values they brought to the work and the ways those values collided with a large language model.

The four themes that emerged, in the paper's terminology, were Listening, Connecting, Carrying and Wanting.

Listening, in the chaplains' account, is not about receiving words. It is about what one of them called listening “very loudly” to what a person is not saying. It depends on silence as a positive act. A chatbot, however well-prompted, cannot listen in this sense, because it has no capacity for loaded silence. It can wait. It cannot attend.

Connecting is the embodied half of the work. The chaplains talked about the comfort of sitting next to another person, the micro-adjustments of facial expression and body language, the way spatial arrangement makes certain conversations possible and certain others unthinkable. One chaplain: “I think there is some comfort sitting next to another person.” It is a small sentence, and in pastoral care an irreducible one. A subscriber talking to Jesus on a phone at 2am is not sitting next to anyone.

Carrying is the theme that hurts to read. The chaplains describe their work as bearing witness to, and taking some responsibility for, the weight of the things people bring to them. A chaplain in the study: “It's about getting help to carry that. That's the difference with a human.” The model, by contrast, cannot be held responsible. It cannot be woken up at 4am because you need someone to know. It cannot promise to remember you next week, because it has no next week and no memory that survives the closing of the tab. Its apparent presence is, as the chaplains unders