Daniel Nealon’s January 2026 essay, “Why Pastors Should Question AI for the Sake of Their Souls” (published here), raises legitimate questions about the formative effects of generative artificial intelligence on Christian ministry. Technologies are not merely passive instruments: they influence habits, expectations and practices. Pastors should ask not only what AI enables them to achieve but how habitual use may change their priorities and practices. To be fair, Nealon does not advocate prohibition; he treats generative AI as a matter requiring Christian discernment and asks what kind of pastors technology may form us to become.
The difficulty lies in Nealon’s answer. He repeatedly claims more than his evidence establishes. Personal experiences of substitutive AI use become general claims about technological formation; questions about efficiency become evidence of defective ministerial values; rhetorical questions sometimes substitute for demonstration; and theological affirmations concerning the Holy Spirit and Incarnation are asked to resolve questions they cannot settle on their own. The article serves well as pastoral testimony and caution but is less successful as a theological and ethical analysis of generative AI.
Nealon is Lead Pastor of Deer Creek Church, a Presbyterian Church in America congregation in Littleton, Colorado, and stated that since 2017 he has participated in theological examination committees assessing candidates for ministerial licensure and ordination. These are relevant credentials for discussing pastoral practice and ministerial formation. They do not establish specialist expertise in artificial intelligence, human-computer interaction, cognitive science, or the philosophy and ethics of technology.
This is not an argument from credentials: pastors need not be computer scientists to evaluate technologies ethically or theologically. But claims about the cognitive and social effects of AI require evidence proportionate to their scope. Nealon’s evidence is largely experiential and theological rather than technical or empirical, and it weakens his argument.
Efficiency, formation and necessary difficulty
Nealon frames his argument through Neil Postman’s image of a burglar distracting a watchdog with food. Technology is implicitly the burglar, questions about usefulness are the distraction, and human beings risk failing to notice what is being taken from them. The metaphor makes the point that technological change involves losses as well as gains, but it also assumes much of what the argument needs to demonstrate. If technology has already been identified as the burglar, attention to its benefits becomes evidence that the user has been deceived. The framing is rhetorically powerful but question-begging: it presumes the spiritually corrosive character of the technology before demonstrating it.
A similar problem appears when Nealon interprets questions about efficiency, precision and mundane tasks as evidence that pastors value efficiency over demanding study, precision over human communication and expediency over patient ministry. The inference is unwarranted. Asking whether AI can reduce sermon-preparation time tells us little about a pastor’s hierarchy of values. A minister may legitimately want to save time on repetitive work to devote more time to prayer, theological study, pastoral care or family. Christian ministry has long benefited from technologies that reduce labour, from concordances and printing to word processors and biblical software. The appropriate ethical question is not whether technology makes something easier, but what has been made easier, what may consequently be lost, and what Christian good the resulting efficiency serves.
Nealon is correct in assuming that certain difficulties should not be eliminated simply because technology enables elimination. Sermon preparation is not merely a process for producing a weekly religious address. Wrestling with Scripture, confronting interpretive uncertainty, constructing a theological argument, searching for appropriate language and praying through difficult questions can form the preacher as well as the sermon. Writing itself may be formative, since we often discover what we think through the effort of articulating it.
Nealon’s own experience illustrates this well. Before using ChatGPT to shorten sermon manuscripts, difficult editorial decisions sometimes prompted him to walk and pray. Once AI accelerated the editing process, those prayerful walks disappeared. This is valuable pastoral testimony, but its evidential reach is limited. It establishes that a particular pattern of AI use altered Nealon’s practice in an unwelcome way; it does not establish that this outcome is intrinsic to generative AI. Another minister might use time saved through AI-assisted editing or administration for additional prayer, study or pastoral visitation.
The deeper issue is the distinction between formative friction and transactional friction. Formative friction consists of difficulties through which patience, knowledge, humility, judgment or prayerfulness develop. Transactional friction includes repetitive or mechanical labour whose difficulty contributes little to ministerial formation. Removing formative friction can impoverish ministry; removing transactional friction may create greater capacity for formative work. The preacher should therefore ask, “Would delegating this activity remove an unnecessary burden, or would it remove part of the formation intrinsic to performing the task well?”
What generative AI does well
Here a significant limitation in Nealon’s analysis becomes apparent. He conceptualises generative AI according to a substitution model: a human possesses a task, transfers it to AI and receives a completed product. His examples of sermon editing, responding to difficult emails and seeking advice largely conform to this pattern. Such practices certainly exist, and some deserve precisely the criticism he gives them. But they represent only part of the range of possible human-AI interaction.
More sophisticated use can incorporate AI into an iterative intellectual process involving human interpretation, AI-assisted interrogation, further reading, counterarguments, verification, theological judgment, human composition and subsequent criticism. A pastor might ask AI to identify assumptions in an interpretation, construct the strongest objection to a theological claim, expose weaknesses in an argument or suggest questions requiring further research. Such practices need not reduce intellectual labour; they may increase it.
Consider the difference between asking, “Write my sermon on Romans 8,” and asking, “Identify the five weakest points in the argument of this sermon on Romans 8.” Both involve generative AI, but they are not intellectually or ethically equivalent. The first may substitute generated material for the preacher’s exegetical and theological work. The second may generate additional work: reconsidering assumptions, consulting scholarship, rewriting an argument or returning to the biblical text.
This exposes a selection problem in Nealon’s examples. He largely chooses cases in which AI replaces something already recognisably valuable: prayerful editing, personal correspondence, consultation with others or difficult intellectual work. Unsurprisingly, those cases generate negative conclusions. But AI might instead replace repetitive formatting, identify an unsupported assertion, expose a weak argument, test an illustration against different audiences or suggest a counterargument requiring further research. Once such cases are considered, the moral analysis becomes more complex.
The category “using AI” is therefore too broad to carry significant ethical weight by itself. Generative AI can function as surrogate, assistant, editor, critic, interlocutor or research aid. The crucial distinction is between substitution and augmentation. The important question is not whether AI contributed to the process, but whether its contribution strengthened or displaced responsible human judgment.
There is also a positive formative possibility that Nealon largely overlooks. Preachers inevitably approach Scripture through particular theological, cultural and personal horizons. Used critically, generative AI can expose assumptions, generate counterarguments, present alternative interpretive perspectives and identify questions the preacher has overlooked. It cannot reliably adjudicate among these possibilities, and significant claims require independent verification. Yet by confronting the preacher with possibilities demanding further thought, AI may sometimes cultivate intellectual humility rather than diminish it.
The relevant comparison is not always between human thinking and machine thinking, but between unquestioned human thinking and human thinking subjected to additional critical interrogation.
Truth, judgment and intellectual discipline
A more adequate understanding of generative AI does not require technological optimism. On the contrary, it reveals distinctive epistemological dangers. Large language models can produce highly plausible responses containing false claims, false or unverifiable citations, conflated sources or misleading interpretations. Contemporary systems may supplement language generation with search, retrieval and other tools, but polished prose remains no guarantee of truth. In theological work, claims concerning biblical languages, historical circumstances, quotations and scholarly positions require verification. The danger is not simply that AI “thinks for us,” but that users may confuse linguistic plausibility with warranted knowledge.
Repeated reliance may also encourage forms of automation bias documented in other professional contexts. A pastor who initially checks generated claims carefully might gradually verify fewer of them as the technology repeatedly proves useful. Responsible practice therefore requires deliberate disciplines: consulting primary sources, verifying significant claims, seeking contrary arguments and retaining sufficient independence to reject generated material. Nealon’s appeal to temperance is valuable here. Temperance need not mean abstinence; it means refusing to allow convenience to determine practice.
This caution applies even to apparently exemplary augmentative uses. A pastor who routinely asks AI to identify weaknesses, generate counterarguments or test applications may gradually become less practised in performing those functions independently. The distinction between augmentation and substitution is therefore not static. Assistance can drift toward dependency unless ministers deliberately continue exercising the capacities that technology assists. Responsible use requires asking not only what AI is doing for us, but periodically what we remain capable of doing without it.
This creates an important symmetry. Generative AI may deepen intellectual discipline by confronting the preacher with objections, perspectives and questions that might otherwise remain unnoticed; yet habitual reliance upon the same assistance may weaken the capacity to generate those challenges independently. Neither outcome should be assumed. Both depend substantially upon patterns of use, professional discipline and the practices within which AI is embedded. Technologies are formative, but their moral consequences are not thereby predetermined.
These issues also show why theological reflection on AI benefits from interdisciplinary engagement. Questions about automation bias, cognitive offloading, epistemic dependence and human-AI interaction are empirical as well as theological questions. Nealon’s pastoral perspective contributes usefully to this discussion, but personal experience cannot establish general causal claims about what generative AI does to cognition or professional practice.
This limitation is particularly apparent when Nealon asks whether AI makes pastors more holy, courageous, prayerful, patient, compassionate and loving, before treating the negative answer as effectively self-evident. It is not. A technology that saves administrative time might enable greater pastoral presence; an AI-generated counterargument might encourage intellectual humility; a critique of insensitive language might encourage greater pastoral attentiveness. None of these outcomes is guaranteed, but neither are Nealon’s negative outcomes. Rhetorical confidence cannot substitute for empirical evidence.
Authorship and differentiated use
Authorship presents a related ethical question. In most preaching contexts, congregations reasonably assume a substantial relationship between the preacher and the words being preached. A sermon ordinarily represents personal engagement with Scripture, theological tradition, prayer, experience and the community addressed. There is therefore a significant difference between using AI to identify repetition or criticise an argument and asking it to generate substantial sections of a sermon subsequently presented as one’s own work. No numerical threshold determines exactly where assistance becomes substitution, but the preacher must remain genuinely responsible for the interpretation, theology, argument and application proclaimed.
For the same reason, speaking generically about “AI sermon preparation” is inadequate. Sermon preparation includes exegesis, research, theological reflection, planning, drafting, editing, illustration, application and evaluation. AI presents different possibilities and risks at each stage. Asking AI to interpret a passage before seriously encountering the text oneself differs fundamentally from asking it, after substantial exegetical work, to identify objections to one’s interpretation. What augments the judgment of an experienced scholar may circumvent the formation of a novice. Ethical evaluation should therefore attach to specific practices rather than to AI use in general.
Spirit-enabled, embodied and ecclesial ministry
Nealon describes himself as outsourcing to a machine the kind of work that God had equipped him to perform by the Holy Spirit. The concern is understandable, but the formulation establishes an unnecessarily competitive relationship between divine agency and instrumental assistance. A preacher gifted by the Holy Spirit may consult commentaries, ask colleagues to criticise a sermon, use a lexicon or search a biblical database. Christian vocation has always been mediated through communities, traditions, teachers, texts and tools. The Spirit’s agency and instrumental assistance need not be competitors. The theological question ought not to be, “Did I do this work all by myself?” but “Have I exercised faithful and accountable Christian judgment in my use of that assistance?”
AI cannot pray, believe, love, possess a pastoral vocation or assume responsibility before God and the church. But material generated by AI may well contribute instrumentally to the work of someone who does possess those capacities. The distinction between instrument and agent is crucial. AI may contribute to an intellectual process without assuming the moral or spiritual agency that properly belongs to the pastor.
Nealon is on stronger ground when he insists that pastoral ministry is embodied. His concern that AI-generated communication could become a means of avoiding difficult interpersonal encounters deserves attention. Listening to grief, navigating conflict, accompanying doubt and discerning the needs of particular people cannot be reduced to information exchange. AI should not replace such relationships.
Yet Christian ministry is not a choice between the solitary pastor and the machine. Christian interpretation is itself ecclesial. Ministers learn through Scripture, traditions, teachers, scholars, colleagues and congregations. AI should therefore be situated within this ecology of discernment. It may supplement human conversation, but it should not replace the relationships through which pastoral wisdom develops.
The limits of Nealon’s theological argument
Nealon’s appeal to Christ is less persuasive. He suggests that Jesus could have saved humanity more efficiently or communicated directly from heaven, yet instead chose embodied relationships, preaching, discipleship and imperfect human messengers. The argument contains an important theological intuition but does not establish the conclusion drawn from it.
Within classical Christian theology, the Incarnation is not simply God choosing an inefficient method of communication. Embodiment belongs to the substance of Christ’s identity and saving work. Nealon’s argument risks confusing the means by which salvation is accomplished with the means by which information about salvation is communicated. Nor does Christ’s commitment to embodied human agency imply opposition to technological mediation. Christianity has always depended upon technologies of communication, from writing and manuscript transmission to printing, broadcasting and digital media. A theological affirmation about the constitutive significance of the Incarnation cannot simply be converted into an ethical principle concerning communication technologies.
There is a broader methodological problem here too. Christology and pneumatology can illuminate questions of human agency, embodiment and vocation, but they cannot by themselves determine what particular technologies do to cognition, learning or professional judgment. Such questions require theological reflection to engage technical understanding and empirical evidence. Nealon sometimes asks theology to carry more explanatory weight than it is intended to bear.
From AI to transhumanism
A similar overreach appears when Nealon’s discussion moves from ordinary pastoral uses of generative AI to emotional attachment to chatbots, suicide-related litigation and transhumanist aspirations. These are ethical issues that deserve careful investigation, but they do not establish much about the responsible use of generative AI in sermon preparation.
The association with transhumanism is particularly weak as an argument against ordinary AI use. The fact that some commentators envisage radical forms of human enhancement or speak extravagantly about the future of artificial intelligence tells us nothing about whether a pastor should use a language model to critique a sermon manuscript. Nealon’s reasoning risks guilt by association: extreme applications and ideological aspirations cast suspicion upon substantially different practices without demonstrating the relevant causal or ethical connection. A Christian critique of transhumanism may be warranted on its own terms. But it should not function as a proxy argument against generative AI generally.
The limits of pastoral testimony
These considerations clarify how Nealon’s article should be assessed. As an experienced pastor, he is well placed to identify changes in prayer, sermon preparation, interpersonal courage and ministerial habits that technical specialists might overlook. His experience assessing candidates for ministry further supports his concern for pastoral formation.
But these credentials do not establish authority concerning generative AI, its technical operation or its empirically demonstrated cognitive and social effects. More importantly, the article itself does not provide the interdisciplinary research that might compensate for this limitation. Its strongest evidence remains personal experience interpreted through a Reformed theological and Postman-influenced framework.
The appropriate conclusion is not that Nealon lacks standing to speak about AI. It is that the scope of his conclusions should correspond to the scope of his evidence. His article persuasively demonstrates that particular uses of generative AI can disrupt valuable pastoral practices. It does not demonstrate that such disruption characterises generative AI generally. Technologies are not neutral, but neither are their moral consequences predetermined.
From suspicion to discernment
Nealon is therefore correct about the central question but insufficiently discriminating in answering it. Pastors should ask what technologies are doing to them, resist cognitive dependency, protect prayer and study, preserve difficult interpersonal relationships and refuse to outsource theological and moral responsibility. But these concerns do not justify treating generative AI as though it possesses a single formative logic.
More adequate distinctions are between use and misuse, augmentation and substitution, dependency and disciplined assistance, transactional friction and formative friction. These distinctions preserve an important symmetry: AI can extend human intellectual capacities while habitual reliance may weaken them. Responsible use therefore requires both openness to genuine augmentation and disciplines that preserve the capacities being augmented.
Nealon has served us by insisting that generative AI should be evaluated according to formation rather than usefulness alone. Yet rejecting technological neutrality does not require technological determinism. Moreover, rejecting efficiency as the supreme value does not make inefficiency virtuous; affirming the Holy Spirit’s agency does not require technological self-sufficiency; and affirming embodiment does not entail resistance to technological mediation.
Nealon’s pastoral testimony offers a useful warning about what can happen when AI becomes substitutive but should not be treated as authoritative evidence of what generative AI necessarily is or does. A more credible Christian response requires pastoral and theological wisdom together with technical understanding, empirical evidence and a differentiated account of human agency.
Rev Dr Rod Benson is General Secretary of the NSW Ecumenical Council and a minister of the Uniting Church in Australia serving at North Rocks Community Church in Sydney.
Image source: Christian Communicators Worldwide
All Told 