I originally wrote this essay in the fall of 2025 and never published it. The topic, however, is not going away, and it seemed appropriate to finally share it. Many thanks to Rev. Jayan Koshy for his input.

Commercial AI is ubiquitous these days—especially generative AI. We interact with customer service chatbots. Our phones let us speak to virtual “agents” who can answer our questions with some moderate degree of accuracy. And every other day we read a new pearl-clutching thinkpiece about someone who wants to marry Chat-GPT, or college students cheating on tests, or something along those lines.

Naturally, religious professionals are exploring AI. Some, like my Buddhist friend Alex Sarkissian, are thinking deeply about the concept of artificial intelligence and how (or whether) we as humans ought to pursue it. Most of us, though, are just trying to answer a more basic question: is it OK for us to use AI in our jobs? Can it improve our daily lives, give us more bandwidth, as we try to tend to the innumerable and ever-expanding needs of parishes, synagogues, and other religious communities?

The framing of this question usually equates “artificial intelligence” with commercial AI products, such as ChatGPT or Claude. It also assumes that AI is a tool, an object which can extend our natural human abilities. Lenses are tools that extend our eyes; the book is a tool which extends our memory; AI is a tool which potentially extends our mind, as well.

Zac Koons takes this position in his article in The Christian Century, critiquing the EpiscoBOT chatbot from TryTank, and I generally agree with his conclusions—as far as they go. AI can be a tool to extend our language composition skills, enabling us to skip the boring part (e.g. writing a letter to dispute a parking ticket) and get on with our day. However, when it comes to preaching or praying, the act of composition is more than an instrumental formality. Composing a prayer is part of praying it. Delivering a specific, targeted, original sermon is part of what we, as pastors, have agreed to do for our congregation week after week—and we may not even know what we want to say until we have labored at writing and rewriting it.

From this perspective, using AI as a tool to “skip the boring part” both misses the point of praying and preaching and sidesteps our responsibilities as spiritual leaders. It even has the potential to distort our own relationships with God, which is developed through our very thoughts. Allowing Google to suggest responses to an email (“Should we buy three or four dozen donuts for coffee hour?”) is one thing. Allowing it to suggest entire prayers or preach for us is another.

It’s easy to see, too, how AI might “creep up on us,” becoming more and more a part of our daily practice until we have forgotten how we used to think without it. As I have watched my toddler daughter grow, I have become more and more convinced that it is very important that we keep generative AI out of our children’s learning process for as long as possible—and that it would be good for adults to eschew it, too. People may joke about “frictionmaxxing,” but it’s true that “if you don’t use it, you lose it,” and I have no interest in losing my capability to write coherent texts!

There is, however, another way to think about AI and other text generation tools. This is as a medium, like paint, stained glass, or videography. Approaching AI as a medium means that we are not treating it as “transparent,” as merely a means to an end, but rather that we engage with AI to create things which could not be created any other way.

To understand what this looks like in a religious context, it may be sensible to think about something else—something simpler, something secular. The artist Dara Birnbaum made a classic work of video art called “Technology/Transformation: Wonder Woman,” reworking material from the 1970s Wonder Woman TV show. (It’s worth taking a few minutes to go watch it now.) Birnbaum could have made an original animation of Wonder Woman doing the same things. Or she could have recorded an actor playing Wonder Woman using a film camera. But instead, she chose to record an existing TV show using a videotape and recut those clips into her own work of art. In doing so, Dara Birnbaum provokes us to think about how women are sexualized and/or empowered (on TV, in particular) and to think about the way technology like the VCR lets us rewind, replay, and thus interact with motion pictures in a particular way. Video recording and replay here is not just a tool used to produce a representation of Wonder Woman. It is central to the meaning of Birnbaum’s work.

With “Technology/Transformation: Wonder Woman” in mind, I would invite us to think about an artwork by Kyle Booten: “To Pray Without Ceasing.” In many ways, this project is exactly the kind of thing that Zac Koons is critiquing. The project was a computer program which scraped Twitter for any desires or needs that people were posting about. The program then generated prayers for those people. It continued to do this day and night, with no human intervention needed, until changes to the Twitter API disrupted its operation.

There are some important ways that “To Pray Without Ceasing” is different from using AI to write a collect, though. “To Pray Without Ceasing” is an artwork and a personal spiritual project, not an attempt to replace a pastor’s prayer and care for their flock. It also makes no attempt to hide what it is and what it is doing. The entire point of “To Pray Without Ceasing” is that the computer is capable of writing infinite prayers, truly without ceasing, for people that it knows nothing about. These are capabilities humans do not have. Indeed, the artist says that he has “tried to make a computational system that—whatever its faults and limitations—is in some limited but objective sense a better person than I am. And one that invites me into its way of paying attention to the world, with constancy.”

Whether we think that the computer’s prayers are efficacious or not, Booten’s project invites us to consider what it would take for us to truly pray without ceasing. If the important part of prayer is our intention or orientation toward it—if the important part is that we are really allowing the Spirit to speak through us, in “sighs too deep for words,” then we might also consider the possibility that AI-generated text could serve as a medium that we engage as we give voice to our intention.¹ Whether we are praying the rosary under our breath on the bus or speaking a long, spontaneous prayer with and on behalf of other members of our Bible study, the intentionality of prayer already takes a variety of forms, drawing on a range of media. This includes computational technology! The medieval theologian Ramon Llull, for example, used combinatorics in his Ars Magna to generate all the logical qualities of God. Surely the Ars is just as prayerful a work as any manually composed sermon or prayer, even though its author relied on mathematics and spinning dials to fully express his vision of God!

In a similar way, we might be able to write programs, even train AI, to generate prayers which are ultimately the fruit of a human’s labor and thought, because a human created the inputs and parameters of the prayers’ creation, even though they did not directly arrange the words of the prayers. A human author could then read those prayers without feeling that the process of voicing their prayer intention had been shortchanged.

For this framework to work, religious leaders would need to approach the process of writing those programs, well…prayerfully. Thinking of the AI as a labor-saving tool, rather than a medium or a method, would be exactly the wrong attitude to take. We would need to be prayerfully engaged with the process of text generation, and this type of prayerful engagement is exceedingly difficult when using commercial AI products, like Claude or ChatGPT. These products are absolutely opaque to the end user—both by necessity (they are the work of many programmers and have been trained on essentially all the text on the internet) and by design (commercial products need to be slick, easy to use, and not complicated; otherwise the average person will find them too confusing). This opacity makes it impossible for a human to have sufficient control of any commercial AI product for it to genuinely function as a medium shaped by a human author.

Fortunately, there are alternatives to commercially available AI products. We might choose to spin up our own small-scale AI and train it on a corpus we have chosen, or even use a simpler method of text generation (Markov chains, combinatorial Python programs…).² Does this require learning basic programming skills? Yes! Does it make your life as a parish priest easier? No! But I believe it creates a result more pleasing to God.

Generative AI is a powerful extension of our minds. It can enable us to converse across language barriers, can reduce busywork, and otherwise save many kinds of labor. It can create artistic effects and lead us to understand the world in new ways, and it is not appropriate in every context. But the same is true for other media and technology. The book is also a powerful extension of our minds, and it is not appropriate in every context either. Sometimes it is fitting to pray a written collect from the Book of Common Prayer, but sometimes extemporaneous prayer is necessary; sometimes it is deeply important to recite memorized prayers, such as the “Our Father,” rather than reading them off a piece of paper. We need discernment to decide what tools to use, what media to use, in each circumstance. I hope that as religious professionals, we will commit to the discernment process, rather than deciding based on knee-jerk reactions in any particular direction.

Last modified on 2026-04-09