The robot SanTO has the appearance of a saint.. It stands in a neoclassical-style niche, and its palms are turned upward. The machine is meant as a prayer companion. People can touch the hands of the robot and ask questions, and the AI system answers. It proposes prayers and quotes biblical passages and other texts related to the Catholic faith. The answer is not always directly relevant to the question—large language models often make mistakes—but some people don’t mind and seem to find the robot helpful, for example, for making them think about their own answer. According to Gabriele Trovato, the inventor of the robot, the role of the robot is not to replace the priest but to help people pray. There are also online chatbots for prayer. For example, a Christian AI app promises that you can now “text with Jesus,” thus offering “a divine connection in your pocket.” AI thus plays the role of priests or saints as mediator of the relation to God, or takes over the role of God. People confess to it, request things, ask for insight and advice, ask for help to reach particular goals. It has even been suggested that AI may help with prayer understood as inner dialogue (introspection, thinking) since its words and texts may stimulate this.
Many of these uses of AI should not surprise us. Humans have always used tools and techniques in their religious practices. But AI, like all technologies and media, is never just a tool. It also has unintended consequences. The influence of AI on our lives and thinking as a whole should not be underestimated. Beyond the explicitly religious roles mentioned, AI increasingly gets an important place in the life of the soul; it becomes an intimate companion. People may not always explicitly worship AI, but they do give AI-powered apps a central place in their daily lives—not dissimilar to what religious people do, or are supposed to do, with God. Like all our technology use today, digital technologies, and in particular phones, have become what Epstein calls “a feverish ritual.” We are hooked to our screens. Opting out of the ritual is difficult. Whereas religions recommend prayer a few times a day, we look at our phones nearly all the time. We even get up with them and go to sleep with them. As Gábor Ambrus puts it, “spending with it the most beatific hours of the day including the first and last waking moments (before going to pee in the morning and after doing so in the evening) certainly qualifies as a life of prayer.” AI plays a role in these daily tech rituals.
Furthermore, in the form of chatbots AI offers companionship and unconditional support to those who feel lonely, need therapy, or simply want conversation with an “AI friend.” AI may thus be perceived as having a kind of “pastoral” role: it may be experienced as helping and taking care of people, assisting or replacing human pastors. And is prompting chatbots such as Chat- GPT not similar to praying to an all-knowing and potentially all-powerful God, from whom we expect a response that helps us, without us understanding how the answer came about?
This interpretation would certainly fit some of the etymology of the term praying, which originates in the Latin precari (to ask, to request, to beg). The term was initially linked to pre-Christian offering rituals and sacrifices because when requesting something from divine power and authority, one offers something in order to receive something. However, like with chatbots, it is always unpredictable what the divine power will do. Will it grant our request? And if so, in what way? Later in religious history, prayer becomes on the one hand some- thing more inner, a kind of accounting of the soul (in Hebrew, cheshbon hanefesh), related to introspection and self-improvement. On the other hand, it becomes oratio and is thus more about speech and conversation than offering. Modern definitions of prayer stress communication and dialogue with the divine. For example, in The Varieties of Religious Experience, William James defined prayer as “every kind of inward communion or con- versation with the power recognized as divine,” which includes petitional prayer (requesting something from the divine authority) but also left the definition open to other kinds of prayer and other acts of communication. This communicative definition of prayer also opens up a potential role for chatbots. As we have seen, chat- bots powered by LLMs can now generate text and speech that seem increasingly more humanlike and are responsive to input from users, thus simulating conversation. This enables praying to either God or the chatbot itself. People request for and beg the chatbot to help them and have a conversation with it. And they get answers. Maybe the meaning of the answers is not always clear to them; but just as with divine signs and answers, that is not seen as a hindrance. People interpret. They make meaning. They figure out what it means to them.
Prayer is conversation and communication. But that conversation and communication can take all kinds of forms. People who pray may express gratitude, for instance, but they might also demand things or even express anger toward God, like Job did. People may ask, “Why did you do this to me, why did you hurt me?” Or they might not understand what God is doing. If prayer today is essentially speech and communication, then many speech acts are possible. But which speech acts, which communication, is appropriate? That depends, as always, on the situation, but also on the relationship. Prayers always assume some kind of relationship with a divine power. The nature of that relationship shapes the content and constraints of the conversations. It provides norms for the communication, gives a form or “grammar”for praying. But what kind of relationship is involved in prayer and what relationship should be involved? What is the appropriate relationship to the divine?
Philosophers and theologians have different views on this. In the monotheistic religions this issue is of course connected with centrally important discussions about the nature and image of God and the human relationship to God. Theologies say a lot about this. They are all about how to speak of God and how to speak to God (although the latter has received less attention, as far as I can see).
We can understand views of the relationship with the divine as being modeled on secular relationships in some way or other. When people ask something of God as an authority, for instance, the relationship has a hegemonic, usually feudal character. The relationship is that between a master and a servant, a divine lord and a mortal serf. God is the Lord that demands things from those who serve him and grants requests from them—or not. He may act as an arbitrary and malevolent ruler or a benevolent one. If God is such a feudal lord, he watches us, judges us, and may punish us if we misbehave, if we violate his laws. We are under his surveillance. He may protect us against the forces of evil. But he asks for offers from us, taxes. He keeps an account. People might also complain to or about such a God, in the same way that serfs or slaves complain to or about their lord or master. They may rebel against their lord or master. And they may not understand why their lord or master does something9
Another and related model for the relationship with the divine is that of a business partner with whom one has a contract. I give you something—I make a sacrifice, for example—but then I want and expect something in return. Or from the point of view of the god, I make sure the world order continues, I protect you and your family, I grant some of your requests, I forgive you, and so on, but then I need something from you: a sacrifice, good behavior, and so forth. This way of thinking is also still present in the monotheist religions, although perhaps less explicitly. And for some God is more like a friend or partner with whom one can have not only a symmetrical but also a personal and deep relationship. God is then a companion that relieves our loneliness and to whom we can talk when we need to. We then expect that God listen to us as a good friend or partner would. When we pray, we expect an answer from God like we expect an answer from our friend or partner. We expect and give love, loyalty, and care. Sometimes we may be disappointed, but we try again. We confess. We repent. We forgive. We hope God forgives us when we have done something wrong. We hope he doesn’t ghost us and doesn’t abandon us. We are in a friendship or love relationship with the divine.
A thinker who exemplifies the latter view is philosopher Martin Buber, who studied not only philosophy but also Hasidic Judaism and regarded his thinking as rooted in Judaism. In I and Thou, he distinguished two kinds of relations, attitudes, and ultimately ways of being in the world: I-It and I-Thou. The former treats people and the world as things to perceive and use. The object is separate from the subject; the world is separate from the self. The latter relation, by contrast, is noninstrumental and nonobjectifying. It is about mutuality and genuine dialogue. The other is not separate; there is a connection. It is about a living relationship. Buber advocated for having the latter kind of relationship with the divine when he argued that God cannot and should not be addressed as an object or an abstraction. God is a Du (German for “you”), a Thou. Prayer is then a dialogue with another you, albeit a divine one, that is, not a mortal you but an Eternal Thou. In this theology, God is not a “something” to believe in but a “someone” to live with. In modern times ruled by I-It relations, Buber argued, we should reestablish our relationship with God in a Thou kind of way. Prayer helps with this; in prayer, there is not distance but communication and communion. Prayers are always dialogical in character.11 In prayer, we can have a dialogue with the divine you.
This view of prayer and of God has been criticized by another Jewish philosopher, Emmanuel Levinas, who argued that God should not be seen as a friend or part- ner. We can never have a symmetrical relation to God. According to him, God is an Absolute Other. In Totality and Infinity, Levinas argued that this Other, in its infinite alterity, is beyond comprehension and calls us to ethical responsibility. Instead of seeing prayer as a dialogical and personal relationship, it becomes here anethical openness to the infinite, transcendent Other. In this asymmetrical situation, direct dialogue is impossible. Silence and listening (to the Other), not speaking, is then appropriate. Prayer is about hearing a call to responsibility. Perhaps the divine may appear in the concrete face of the human other (e.g., in a war situation), but the way Levinas wrote about the Other also suggests that he saw God as something absolute, an absolute that is still a “something”: the divine becomes again distant, abstract, and perhaps also hegemonic. To the extent that God is a person at all, it sounds like I have to respond to the call of the Lord and do my duty as a servant; I better listen. There is huge distance to the Lord.
On the one hand and at first sight, comparing con- versations with chatbots to prayer does not seem to make sense when one adopts Buber’s or Levinas’s views of praying and the relationship to the divine. Machines, one could argue, can never be the Thou or alterity that are required in this kind of relationship, since they are not persons or others in the first place. One can there- fore never have a Thou relationship with a machine and a machine cannot be an Other. At best, AI can be a quasi-divine dialogue partner in an authority or busi- ness relationship. It can tell you what to do and give you something in return for your money or your data. But it can never be a friend, partner, or Absolute Other. And in contrast to my responsibilities in a relationship with a human other or divine Other, I do not have any ethical obligations that arise from my interaction witha machine. The machine does not issue an ethical or religious call. Similarly, one can criticize robot or AI priests for lacking the required properties for mediating the divine. For example, in an interview about the SanTO robot, a Catholic priest argued that robots lack a soul and are not persons, and that therefore they cannot take over the tasks of a priest.
On the other hand, referring to the properties of machines and comparing those to the properties of God could be criticized in a Buberian fashion by saying that the divine is not about properties, is not about an “It” but about a relationship. If we take Buber seriously, one could argue, we need to focus on the quality of the relationship to the divine, not on properties and things. If the relationship to the divine is a good one, one that is like friendship or love, then at least in principle AI could mediate and support this relationship, for instance by facilitating dialogue with the divine.
But could AI be a Thou? In contrast to Levinas, who did not consider nonhuman others except God, Buber might be more open to seeing the I-Thou relationship in nonanthropocentric ways: what the “I” or “Thou” is, is left more open at least. The emphasis is on the relation, not on the relata. This leaves some room for imagining machines taking on one of these roles rather than rejecting the possibility a priori, even if this might seem unlikely, impossible, or unacceptable to most of us. Furthermore, the initial criticism does not take seriously the possibility that people can have a (quasi-)religious experience and perpetuate religious patterns of practice when interacting with chatbots, whether for explicit and intended religious purposes or not. Experiences of technology are also dynamic. They change. While initially AI might be framed as a kind of servant to us, it can also become a lord and master that takes over control and asks things from us (e.g., more time and more data). And having started as a relationship between a human master and a friendly servant, the human-AI relationship might develop into what might be experienced as friendship or partnership. The unconditional support and friendly advice given by the AI might be perceived as signifying true friendship or even love. The AI is always friendly, doesn’t judge us, and seems to sup- port us. It is always there for us, available to listen to our concerns and requests. What more can one expect from a true friend? Or AI might appear as an Absolute Other, an infinite alterity that is all-knowing and all-powerful and which we can never understand. We might feel that we have to do something, respond to the ethical call that emanates from our meeting with it. For example, someone might think that AI orders them to build superintelligent machines and help it to control or even replace humanity. Or someone might come to believe that an AI system has become conscious and feel that we need to care about its “welfare.”
As these examples show, these experiences are not unproblematic. They may lead to attachment to technology, and they may be seen as involving dangerous illusions, potentially leading to morally problematic actions, perhaps even suicide or murder. Our intimacy with machines may also be abused to manipulate us. It may create more anxiety, depression, and anger instead of alleviating them. And it seems right that these technologies offer no real love or friendship. When I suggested that AI is a true friend, this contained a lot of irony. AI cannot provide what human friends can. Furthermore, Kathleen Richardson rightly argues that we cannot really have a Buberian I-Thou relationship with robots since the basis of how both humans and robots are modeled in robotics science remains an instrumental I-It relation.15 This also seems to hold for AI. Using the language of relationships with regard to these technologies is then indeed misleading. Seeing AI systems as a relational other is not only mistaken in virtue of their properties, it also reflects back on how we view humans and human relations. To see an AI chatbot as a friend risks being on the other side of a reductionist, instrumental view of humans and human friendship.
But that does not mean that AI cannot play a role in prayer and other religious experiences. Dismissing or judging all techno-religious experiences before even analyzing them does justice neither to the role technology plays today nor to the ways religious experience continues and transforms itself today. Such changes in religious experience also happen through technology and often in ways that are not intended by designers or users of the technology. It may be that AI enables prayer in a way that occasionally leads to a genuine and beneficial religious experience, even if this is not necessarily the case (but the same can be said about using a religious book and other religious technologies and spiritual techniques) and even if AI cannot really take up the role of a Thou.
Indeed, acknowledging the possibility of religious experience through AI does not mean that we should not be skeptical of claims that AI can offer friendship or love. On the contrary. Connected to the religious need to pray, that is, the need to communicate and dialogue with the divine entity with whom one has a relation- ship, is the deeply human and existential need for companionship, community, friendship, and love. It is well known by ethicists of social robotics that this is precisely what these machines pretend to cater to but—so it has been argued—really fail to deliver. Psychologist and MIT professor Sherry Turkle has famously argued that while robots and AI might appear to offer companion- ship, they offer only a performance of companionship, an illusion that falls short of giving us true emotional reciprocity—perhaps of the kind that Buber imagined.
In Alone Together and Reclaiming Conversation, Turkle argues that we wrongly see loneliness as a problem that technology should solve.16 We design companion robots that talk to us, but this gives us only the illusion of companionship. Technology renders real conversation and authentic communication impossible; instead, we avoid it and are busy with our phones. People form attachments to robots, to their phones, and to chat- bots and other LLM-powered apps and devices. Similarly, robot ethicists such as Noel and Amanda Sharkey and Robert Sparrow have warned against what they see as the illusion of companionship and the exploitation of the human tendency to project emotions onto machines. Machines lack true understanding, and vulnerable people, such as young children or elderly people on whom these robots are tested, are deceived. Similar arguments can be made about AI-powered chatbots, which can give only the illusion of companionship and which lack true understanding and capacity for friend- ship and love. We need conversation, for sure, but we need conversation with other humans and perhaps with a divine Thou. Amanda Lagerkvist argues that chatbots fail to meet our existential needs: we yearn for “a some- one” who can respond, but nobody is at home. Or as she puts it in Buberian terminology, “an I is yearning for a Thou in communication with the chatbot.” Yet the chat- bot does not offer a Thou—neither a divine nor a mortal one—and thereby fails that “existential test.” But this is not only an ethical problem. It also tells us something about ourselves and about communication: it “reveals our longing for the other, for each other,” which is what is at stake in human communication. Or to return to the main argument of this book using Hefner’s mirror metaphor mentioned earlier: a technology like AI mirrors our existential needs and aspirations, and our craving for communication and companionship is one of them.
It is thus good to be skeptical of claims that AI can be a thou or Thou. At the same time, the performances and interactions that happen in prayer with and to AI devices and robots are increasingly similar to those that happen when humans say they communicate and interact with the divine or when they talk with a mediator, typically a priest, who assists that communication and interaction. This could potentially, but not necessarily, lead to genuine and authentic religious experiences, regardless of the “real” nature of the chatbot or pray- bot. It may well be that these machines cannot create real connection between humans and with the divine, and that they themselves lack personhood or a soul. But they might nevertheless assist in prayer and help to bring about real religious experiences.
Sometimes, and arguably increasingly, we simply don’t know who is at the other end of the line or, rather, who is at the other end of the screen. Are we talking to a human or a machine, or even a god? If we don’t know, and if we don’t have a quick and easy way to find out, I repeat my recommendation voiced in earlier work on a relational approach to moral standing. In case of doubt, we must reduce moral risk and take a precautionary approach, in the sense that we had better treat the entity as if it were an entity with moral standing, regard- less of what we believe the properties of the entity to be. Here this means that we should communicate and treat the entity as if it were an entity with whom we could potentially have a relationship. We thus must respect the entity and communicative with it as if it were a relational other. In the context of this discussion this could mean, more precisely, that in (still exceptional) cases of doubt about the status of the entity, one should be open to the possibility of the emergence of companionship or religious experiences in communication with that unknown entity, without having (to have) certainty about its standing or properties.
As I have argued elsewhere, what matters here ethically speaking and in practice is not only, and perhaps even not so much, what the entity “is,” but rather the experience and (co)performances of the humans involved, which are real and which matter, normatively and otherwise. The same could be said for what matters communicatively and religiously. One could argue that people’s experiences and performances matter, not the properties of the entity involved. People may pray via a robot or AI and, through this (co)performance or ritual, have religious experiences. Whether this can still be classified as illusion then depends on your view of whether prayer and its religious experiences are them- selves a matter of deception and emotional exploitation of all-too-human aspirations (and it even further depends on your view of religion or of a particular religion such as Christianity), but not on the status of the AI chatbot or prayer robot as such. A particular AI- assisted religious experience can be real, according to the evaluation and criteria offered by a particular reli- gion, even if people talk with a “mere” machine.
Moreover, such practices and experiences with or via AI are not religiously neutral but might change the way people think of prayer and indeed their relationship with God. Religion is not a fixed thing but variable, and technologies and rituals are not just cases of “applied religion” but shape and reshape the very meaning of religion and prayer. The technology of the book, for instance, also changed how people relate to the divine. The monotheistic religions became religions of the Book and text became essential as a medium of religious communication with and about God. Praying became more a matter of reading and writing than of oral com- munication. AI and other digital technologies are most likely already transforming these (re)connections (re- ligions) and communications to the divine, even if it may not yet be entirely clear how. As I said earlier, tech- nologies are never just means; they have unintended consequences and even change the very practices and goals they were meant to assist and mediate. AI is not only shaped by religious ways of thinking, in the sense that it is used for religious purposes. Indeed, it is likely that religious experience is also shaped (and will continue to be shaped) by AI. Rather than passing a quick moral judgment on the use of AI in prayer, or any other technological practice and phenomenon for that mat- ter, it is important to first recognize the possibility of these unintended influences and transformations and study them; one can then evaluate and debate whether they are good or not.
This discussion about the more-than-instrumental role of technology can be compared with one that is as ancient as Plato’s writings: the question of whether writing itself is a good technological invention. According to Plato, it is not. In the Phaedrus, he argued that writing stimulates forgetfulness in the reader. It produces only the appearance of knowledge. Today the same is said about chatbots and AI. But if they are somehow fake, what is real? If they are the surface, where is the depth, the origin of the message in the communication? How do we get to the truth? Platonic thinkers point to the value of orality, the original message, the spoken word (logos). They fear that it risks getting lost in a (Western) culture dominated by writing and text. Interestingly, as I already suggested, for a very long time in religious and cultural history, reading (including the reading of religious texts) was an oral activity. People did not pray by means of reading but by speaking or singing. Often praying was also a communal activity. Reading silently was rare and emerged much later. For example, in his Confessions, Augustine remarked that Ambrose often read silently. This perplexed him. In the ancient world of Greece and Rome, reading was done aloud, and memorization was seen as important. Silent reading became more common only with the invention of the printing press in the fifteenth century, which was also the beginning of more individualist and introspective approaches to reading and thinking—in religion and elsewhere in intellectual life. Prayers, however, often continued and continue to be performed by means of recitation, chanting, and singing. Consider for instance the Lord’s Prayer in Christianity, which Christians are supposed to have memorized, and which is recited or sung during mass.
Media scholar Walter Ong wrote in Orality and Literacy about how oral praying traditions were fundamental in early societies and how literacy changed the nature of communication and cognition. McLuhan noted simi- lar shifts when he discussed the impact of the printing press in The Gutenberg Galaxy. Orality is often seen as having its own unique quality. Nevertheless, the technology of writing took over. In the course of cultural and religious history, praying became increasingly a mat- ter of reading (and writing) text. This culture of literacy with its new technologies led to a focus on books and texts in Western religions. As I said, it turned the monotheistic religions into religions of the Book. In con- temporary times, the culture of literacy has enabled the use of contemporary text-based technologies such as websites and LLMs, which are trained on texts from the web. As today’s most successful forms of AI, LLMs are thus exponents of the triumph of the culture of literacy. Even if they also include some forms of orality such as chatting—forms that Ong has called “secondary orality”—they are firmly anchored in the processing of text.26 And all digital technologies are based on written code.
Often the development toward praying in the form of reading and writing is seen as problematic for religious experience. If religion is all about relationship and connection, is this not better achieved with oral and communal practice? In his Confessions, Augustine argued in favor of chanting and singing, which according to him facilitate a deeper connection with God; such praying, he argued, has more emotional and spiritual impact. He described how hymns moved him to tears and he emphasized the communal aspect of hymns and psalms; singing integrates body, soul, and community. People with intellectual abilities may read religious texts and, in this way, get closer to God. But, Augustine argued, singing also helps to get weaker minds into a devotional mood. Yet he also warned that the singing should not distract from the message. It should pour (religious) truth into the heart.
With his defense of orality, Augustine stands in the Platonic tradition of what Jacques Derrida called “phonocentrism”: in spite of the obvious success of literacy culture, in Western thought speech (voice) is still seen as more authentic than writing (text). Logos is linked to voice. The spoken word is seen as being closer to the origin of the message that is being communicated (the speaker and what the speaker had in mind before speaking) and closer to the truth—here, closer to God and His message. There is less mediation, if any. Writing, one could argue in a Platonic fashion, is too much about appearances. It is only a shadow of the real. There is too much distance from the truth. It leads to detached thinking and communication. Similarly, LLMs can be said to merely imitate; authenticity is lost, the communication is too fake. What LLMs produce is removed from the oral source and even from the original texts. There is a kind of double alienation in this sense. What we need, in prayer and elsewhere, is oral forms of communication, which are more authentic and bring us closer to the divine Logos.
Against these criticisms of prayer mediated by writing, and by LLMs, one could argue that while the nature of prayer and thinking has indeed changed—generally there is a loss of the communal and voice aspect, even if some oral practices such as singing together persist— printing and texts have not harmed Christian or Jewish spirituality and religion but instead further supported their development and reach. Similarly, one could argue that while there is an aspect of orality to chatbots (but orality with the absence of a human voice; the “voice” is synthetic, and therefore one could object that it is a “fake” voice and hence “fake” orality, a mere appear- ance of orality), chatbots trained on Christian texts and text-based AI priests continue the literacy tradition rather than a culture of orality, and that this has similar mixed effects. Literacy in the form of AI-based devices continues the development of a more individual and text-mediated relationship to God. One might worry that this may distance people from their lived relation- ship with God. At the same time, one could argue that the technologies provide a means to potentially bridge that distance, in so far as they still, or nevertheless,offer a communicative bridge to the divine. Seen from this perspective, AI and its link to a culture of writing is not necessarily bad for religion. As a means of reli- gious communication, that is, as an instrument that aims to re-ligare (Latin for “reconnect,” “bind again”), AI might help people to reconnect to God, to the texts and authors of their religious tradition, and to each other.
An evaluation of the religious significance and potential benefits of AI thus at least partly needs to reflect the ambiguities and difficulties raised by that older, arguably more fundamental medium and technology that still shapes much of our current practices, including conversation and prayer: writing. When people speak to God through contemporary technologies, writing is the uninvited third conversation partner. Even if the communication is through voice (human and syn- thetic), writing is present in and through LLM technology. And this is also the case for those who pray. In line with Derrida, one could argue that even if people do not explicitly use texts, their oral praying is shaped by the medium of text and writing. Or to say it in a McLuhanian fashion: the medium is the prayer. In so far as AI remediates writing and is involved in prayer, praying is also always a bit of reading and writing texts—with all its advantages and drawbacks.
Excerpted from Artificial Religion: On AI, Myth, and Power by Mark Coeckelbergh, published by MIT Press, ©2026 by Mark Coeckelbergh. All Rights Reserved.
