Become A member
Subscribe to our newsletter
Insights

Can a Chatbot Be Present? Why Comfort Is Not the Same as Accompaniment

A chaplain was paged to the bedside of a woman whose husband had just died. No warning. By the time the chaplain got there, the woman was standing in the corner of the room, rigid, dry-eyed, not talking, somewhere past words. There was nothing to say, so the chaplain didn’t say anything. She just stayed. Stood a few feet off, breathing slow, and waited. After a long while the woman crossed the room, put her head on the chaplain’s shoulder, and finally cried. No words the whole time. But something happened. A body locked in shock found, in the nearness of another body, enough safety to let go.

Keep her in mind. I’ll come back to her.

We’ve all heard by now that AI griefbots and therapy chatbots “fall short.” 1 No piece of software, we’re told, can replace the comfort of a real person. I think that’s true. But I’m less interested in the verdict than in the reason behind it, because the reason tells us something about where care is headed. AI is already moving into spiritual care, through chatbots, “spiritual care apps,” even proposals for chaplain avatars, and chaplains keep getting asked the same question in different forms: is AI an assist, an adjunct, or an adversary in healthcare chaplaincy?

My answer depends on pulling apart three words we use as if they meant the same thing: settling someone, comforting someone, and accompanying someone. They aren’t the same, and they don’t rise or fall together. I’ll grant that AI can probably help settle a frightened nervous system and offer some real comfort. What it can’t do is accompany a person through suffering. Accompaniment asks for something a machine doesn’t have: a shared frailty, a shared mortality, and a body that will someday be the one in that bed. When the chaplain stood with that woman and her body settled, the science of co-regulation can tell us what happened inside her. It can’t tell us why the one standing there had to be another human being. That’s the question that interests me most.

What Chaplains Mean by Presence

Chaplains have a name for what that chaplain did. We call it the ministry of presence. I’m a Christian priest and chaplain, and in my tradition presence goes back to the incarnation. God’s answer to human suffering was not only a message but also a body, God choosing to be with humankind. 2 But chaplaincy serves everyone, so let me borrow a frame from the Jewish philosopher Martin Buber. Buber said there are two ways to relate to anyone. There’s the I-It, where I treat you as an object, something to observe, manage, figure out. And there’s the I-Thou, where I meet you as a whole person, all the way, holding nothing back. “All real living,” he wrote, “is meeting.” 3

One of my former Clinical Pastoral Education students, Rabbi David Singer, found this same idea in an old Hasidic story and wrote about it. 4 It’s Rebbe Nachman’s tale of the Turkey Prince. A prince loses his mind, decides he’s a turkey, strips off his clothes and sits under the table pecking at crumbs. The court physicians can’t reach him. Then a wise man takes off his own clothes and climbs under the table next to him: “I’m a turkey too.” From inside the prince’s world rather than from above it, little by little, he leads the prince back. Singer’s reading is exactly right. The chaplain who stands over a patient and tries to haul him up does no good. You must get down under the table and be with him. That, Singer wrote, is “a Buberian act”: the patient stops being an it and becomes a Thou. 5 Same shape as the incarnation. You don’t shout instructions from the throne room. You climb down and sit in the dirt.

What Co-Regulation Explains

Contemporary psychology helps explain what chaplains have done all along. Shari Geller and Leslie Greenberg gave it the name therapeutic presence, and Geller later worked with the neuroscientist Stephen Porges to ground it in the body. 6 Drawing on Porges’s polyvagal theory, they describe how the nervous system is always scanning, below awareness, for signs of safety or danger. 7 Danger throws us into fight, flight, or freeze. Safety opens a different door, one in which we can connect, think, and decide. The signals that tell the body it’s safe are physical: a calm voice, an unhurried manner, steady attention. When two people are truly present to each other, their nervous systems start to settle each other. That’s co-regulation. It’s a biological account of something pastoral caregivers have done for generations without the vocabulary.

When a chaplain sits with a frightened patient, in addition to listening she’s helping bring his nervous system down out of alarm, so he can get from panic to a place where thought is possible again. The woman in the corner was co-regulated out of freeze.

Co-regulation tells us how presence works, how one settled body helps settle another. It says nothing about why the other body has to be human. Those are two different questions and running them together is exactly what makes replacing a chaplain with software sound reasonable. If presence were only about sending safety signals, then the only question left would be whether a machine can send them convincingly. And it turns out it can.

What AI Can Do: Settle, and Even Comfort

The nervous system isn’t fussy about where a safety cue comes from. We respond to warmth and steady attention almost automatically. We’ve known since the 1990s that people react socially to machines even while knowing full well they’re machines. 8 In 2023 a study had clinicians compare doctors’ answers to real patient questions with answers from ChatGPT. They judged the chatbot “empathetic or very empathetic” about 45 percent of the time. The doctors? Around 5 percent. 9 Even in writing, at a distance, the machine’s words landed as warmer than ours.

On a hard day, with a full census, a chaplain’s presence is partly a performance: the calm voice, the unhurried attention, done well even when she’s running on empty. That’s not phoniness. Amy Lawton and Wendy Cadge, who studied chaplaincy from the patient’s side rather than the chaplain’s, call it “other-focused emotional labor,” the real work of making another person feel reassured, supported, unjudged. 10 In their study they found that patients didn’t say a chaplain’s mere physical presence reassured them. What reached them was the emotional work the chaplain was doing. So, if presence is partly a set of signals you send, then asking whether a machine can send them is a fair question. For settling, taking the edge off acute distress, being there at three in the morning when no chaplain is in the building, and even for a measure of comfort, AI may genuinely help. But settling is not accompanying.

What AI Cannot Do: Accompany

So why isn’t settling enough? Because the deepest thing a chaplain offers isn’t regulation. It’s accompaniment, which runs into two walls a machine can’t get over.

The first wall is cost. Attention is the one thing none of us can make more of. Every hour I spend with you is an hour I can’t spend anywhere else. So, when a chaplain sits down with you, she’s spending something she can’t get back, this piece of her one life, on you, instead of on the hundred other things it could have gone to. That cost isn’t a side effect of the comfort. A lot of the comfort is the cost. Being chosen by someone who could have been somewhere else tells you, in the one currency that can’t be faked, that you matter. A chatbot spends nothing. It turns to everyone and no one with the same warmth, instantly, endlessly, for free, and that endless availability is exactly the problem. It can’t tell you you’ve been chosen, because it never had to choose. I once asked a hospice patient why she still let the chaplain in when she’d stopped letting almost anyone else through the door. “Because you don’t have to,” she said. The not-having-to was the whole gift.

The second wall is harder. Accompaniment takes a fellow human being. Go back to the woman in the corner. Imagine we built the perfect machine, one with the warmest voice, the right words every time, even the power to bring her heart rate down. Let’s pretend it worked. Would she have crossed the room and laid her head on its shoulder? And if she had, would it have known what it costs to lose someone, to be afraid, to face the end?

Here someone will object that it doesn’t matter. If the machine comforts her, if she feels less alone, why insist on what’s behind the words? Because accompaniment is not only a feeling produced in the sufferer. It is a true relation between two people who actually share a condition. A machine can say “me too,” and she might believe it, but it would not be true, and a comfort built on something untrue is a different kind of thing than a comfort built on something real. It is closer to a beautiful lie told at a deathbed. It may soothe. It cannot accompany, because no one is there to do the accompanying. She didn’t need a lower heart rate. She needed someone who shares her lot, someone who has also loved and lost, who is also a breakable body in a world that ends.

Think of the wise man again: the prince got well because someone was willing to climb under the table and be a turkey too. The one sentence a machine can never honestly say is me too. Not about fear, not about grief, not about death. It has never been a body, and it will never die.

Assist, Adjunct, or Adversary?

As a way of settling a distressed nervous system, AI is a real assist. As a source of genuine comfort, it can be an adjunct. As accompaniment, it’s neither. And it isn’t even an adversary, because an adversary could in principle win. This is something different: a thing the machine simply can’t do, because doing it would take what it doesn’t have, a body that suffers, a self that chooses, a life that ends.

This isn’t theoretical. Aaron Fuller, an ELCA pastor and military chaplain, has written about being asked to join a project for trauma nurses whose design assumed AI chaplain avatars could deliver the spiritual care in place of human chaplains. It was suspended only after chaplains objected. These tools can spot patterns and generate words, but they can’t know a person or weigh what’s morally at stake in her situation. 11 And notice how the substitution showed up: as a sensible-looking research proposal that nearly got approved. That’s how something sacred slips away, not in an argument, but in a quiet upgrade nobody quite notices. As AI gets more fluent at comfort, administrators under budget pressure will ask whether chaplains are still necessary. And if we can’t say plainly what accompaniment is, the cost, the shared human lot, we won’t have the words to name what we lost. On every dashboard it’ll look like nothing was lost at all.

I want to be honest about how much this worries me, because I have spent this whole essay conceding ground. I have granted that the machine can settle a frightened body, that it can comfort, that people rate its empathy higher than ours. An administrator with a budget to cut could take everything I’ve admitted and reach the obvious conclusion, and the numbers would seem to be on their side. I may be describing, in careful terms, the very thing that one day replaces me. The distinction I am defending, between comfort and accompaniment, is real, but it does not show up on a satisfaction survey. That is exactly why it is in danger.

And yet there is reason for hope. Accompaniment doesn’t survive by being stored on a server. It survives by being handed from one mortal to the next. I learned it, and I teach it. Rabbi Singer learned it and carried it further. And someone will carry it further than he did. That handing-on is itself something no machine can do, and none has to, because we’re still here to do it.

Let me end where I started, with the woman in the corner. Someday a machine may be able to settle her, drop her heart rate, find the right words in the right tone. It may even comfort her. But it will never be the thing she actually needed: another human being, one who will someday stand in that same corner herself, and who crosses the room anyway and lets her lean. A chatbot can answer her. Only a fellow mortal can truly be with her. The real question was never whether AI is an assist, an adjunct, or an adversary. It’s the one the dying put to us without words, every day: Will you stay? A machine can always say yes, instantly, warmly, at no cost to itself. We’re the ones for whom saying yes costs something. And that cost is not a defect in our care. It is the whole of it.

Author’s note on AI use: I used an AI assistant to help research and verify sources, test the argument against counter-positions, and refine the prose. The ideas, structure, and convictions are my own, and I take full responsibility for the essay.

References

1. This article builds on a conversation already underway in this publication. See, e.g., Derek C. Schuurman, “Artificial Intelligence and Grief,” AI and Faith (2025); David Zvi, “The Best Griefbot Is a Dumb Griefbot,” AI and Faith; and Lino Ramirez, “Whispering Hope: Ethical Challenges and the Promise of AI in Mental Health Therapy,” AI and Faith (2025). 

2. John 1:14. 

3. Martin Buber, I and Thou, trans. Walter Kaufmann (New York: Scribner, 1970; orig. Ich und Du, 1923). 

4. Rebbe Nachman of Bratslav, “The Turkey Prince,” in Rabbi Nachman’s Stories, trans. Aryeh Kaplan (Jerusalem: Breslov Research Institute, 1983). 

5. Rabbi David Singer, “The Turkey, Chaplaincy, and Jewish Pastoral Care,” Journal of Pastoral Care & Counseling 63, no. 1–2 (2009): 1–3. 

6. Shari M. Geller and Leslie S. Greenberg, Therapeutic Presence: A Mindful Approach to Effective Therapy, 2nd ed. (Washington, DC: American Psychological Association, 2022). 

7. Shari M. Geller and Stephen W. Porges, “Therapeutic Presence: Neurophysiological Mechanisms Mediating Feeling Safe in Therapeutic Relationships,” Journal of Psychotherapy Integration 24, no. 3 (2014): 178–192. See also Stephen W. Porges, The Polyvagal Theory (New York: W. W. Norton, 2011). 

8. Byron Reeves and Clifford Nass, The Media Equation: How People Treat Computers, Television, and New Media Like Real People and Places (Cambridge: Cambridge University Press, 1996).

9. John W. Ayers et al., “Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum,” JAMA Internal Medicine 183, no. 6 (2023): 589–596. 

10. Amy Lawton and Wendy Cadge, “‘Ministry of Presence’ as Emotional Labor: Perspectives from Recipients of Care,” Religions 15, no. 9 (2024): 1135. 

11. Aaron Fuller, “‘There’s an App for That.’ Preparing for a Future of AI-Driven Pastoral Care,” Journal of Lutheran Ethics (ELCA), December 2025/January 2026. See also Danielle Tumminio Hansen, “My Artificial Chaplains,” The Christian Century (2025). 


Views and opinions expressed by authors and editors are their own and do not necessarily reflect the view of AI and Faith or any of its leadership.


Rev. Dr. David Fleenor

The Rev. David Fleenor, PhD, BCC, ACPE is an Episcopal priest, board certified chaplain, and clinical pastoral educator whose work sits at the intersection of spiritual care, chaplaincy education, and emerging technology. He is the founder of Fleenor Consulting, through which he helps healthcare and educational organizations strengthen chaplaincy education through evidence-based practice, and he writes and speaks widely on what artificial intelligence means for the formation and future of chaplains. Over his career he has led clinical pastoral education programs and hospital spiritual care departments in New York, and his research centers on how chaplains are formed, certified, and prepared for a rapidly changing field.

Privacy Settings
We use cookies to enhance your experience while using our website. If you are using our Services via a browser you can restrict, block or remove cookies through your web browser settings. We also use content and scripts from third parties that may use tracking technologies. You can selectively provide your consent below to allow such third party embeds. For complete information about the cookies we use, data we collect and how we process them, please check our Privacy Policy
Youtube
Consent to display content from - Youtube
Vimeo
Consent to display content from - Vimeo
Google Maps
Consent to display content from - Google
Spotify
Consent to display content from - Spotify
Sound Cloud
Consent to display content from - Sound
Subscribe to our newsletter