AI May Assist the Sermon, but It Must Not Become the Preacher

Pastors are no longer debating artificial intelligence from a safe distance. They are using it to search for ideas, summarise commentaries, structure sermons, create illustrations, translate material and reduce the time required for weekly preparation. Reporting published on 19 July describes pastors who welcome these efficiencies while wrestling with a question that reaches the heart of Christian ministry: where should assistance end and pastoral authorship begin?

The scale of adoption makes the question urgent. Barna’s 2026 State of the Church research reports that only thirteen per cent of surveyed United States pastors do not use AI at all. Half use it for brainstorming, thirty-six per cent for biblical or theological research and twenty-four per cent for writing or editing sermons. At the same time, seventy-one per cent describe themselves as cautious and forty per cent as conflicted. These findings are drawn from a United States sample and should not be treated as a global or African prevalence estimate. They nevertheless reveal a ministry practice advancing faster than its theology and governance.

A sermon is more than religious content

A language model can generate a recognisable sermon: a title, introduction, biblical references, three points, illustrations and a conclusion. Research evaluating AI-generated evangelistic sermons confirms that models can reproduce familiar gospel-preaching patterns, including testimony-like language, biblical claims and appeals to listeners. Structural competence, however, is not the whole reality of preaching.

A sermon is biblical truth interpreted, prayed, embodied and proclaimed by a person who stands within a particular community. The preacher has listened to Scripture, but also to the grief, hope, conflict and calling of real people. The words carry moral responsibility because the preacher can be questioned about the interpretation, challenged about the application and held accountable for the consequences.

AI has no congregation, calling, repentance, obedience or experience of the Spirit. It can imitate the language associated with these realities, but it cannot bear the pastoral responsibility of proclamation.

The decisive boundary is assistance versus authorship

Christian leaders have always used resources. Preachers consult translations, lexicons, commentaries, concordances, theological books, colleagues and previous sermons. A South African peer-reviewed study published in April argues that AI can likewise function as a sermon-making resource without necessarily destroying authenticity, provided its material is discerned and applied within a fundamentally spiritual process.

That distinction is constructive. The problem is not simply that a machine contributed information. The problem arises when the preacher transfers interpretive judgement, theological ownership or pastoral responsibility to the system. Assistance expands the preacher’s work; substitution allows the preacher to present generated work as personal spiritual labour.

A useful test is whether the pastor can explain every major interpretive decision, verify every source, defend every theological claim, connect the application to the actual congregation and truthfully say, “I take responsibility for this message.” If not, AI has moved from tool toward concealed author.

Appropriate uses remain accountable uses

AI may help a preacher compare possible outlines, identify questions a listener might ask, improve clarity, generate discussion prompts, locate themes for further research, create a first translation, shorten an announcement or test whether an illustration communicates as intended. It can also serve as a critical reader by identifying unsupported claims, confusing transitions or assumptions that require evidence.

Each use still requires verification. Generated references may be fabricated. A biblical explanation may ignore literary or historical context. A quotation may be inaccurate. A culturally attractive illustration may never have happened. A translation may change theological meaning. AI increases the number of possibilities available to the preacher; it does not reduce the preacher’s responsibility to judge them.

Some practices cross a pastoral red line

A pastor should not present an invented testimony as fact, attribute fabricated words to a scholar, copy another preacher without acknowledgement or preach generated exegesis that has not been personally examined. Confidential counselling information, prayer requests and identifiable congregational data should never be placed into an unapproved public AI system. A pastor’s voice or image should not be synthetically reproduced without clear consent and safeguards.

The deepest red line is concealed spiritual outsourcing: asking a system to produce the message, making minor changes and presenting the result as though it emerged from the preacher’s own study, prayer and pastoral discernment. The issue is not that every technological contribution must be narrated from the pulpit. It is that the congregation should not be deliberately led to believe that borrowed or generated spiritual labour is personal work.

Pentecostal preaching is Word, Spirit and witness

Pentecostal preaching is not anti-intellectual improvisation. It requires responsible exegesis, theological understanding, homiletical discipline and sensitivity to context. Yet it also expects the Holy Spirit to illuminate Scripture, convict the preacher, direct application and address the gathered community through a living witness.

The Spirit’s role should not be invoked as an excuse for weak preparation, just as AI efficiency should not become a substitute for prayerful dependence. The preacher studies because the text deserves care. The preacher prays because information alone cannot create obedience. The preacher listens because application concerns real people rather than an imaginary audience assembled from training data.

The fire of the Spirit and the discipline of serious interpretation belong together. AI may support the discipline; it cannot manufacture the fire, character or responsibility of the witness.

The preacher’s knowledge of the congregation is irreducible

A sermon on grief sounds different after the preacher has sat with a bereaved family. A message about work changes when the congregation includes young adults searching unsuccessfully for employment. Teaching on giving requires different care in a community facing economic pressure. A sermon about faith must not shame a person who has prayed without seeing the desired outcome.

This pastoral knowledge is not a database to be uploaded. It is a relationship of trust. AI can help a preacher organise language, but only the accountable pastor knows which detail must remain confidential, which illustration may reopen a wound, which correction requires gentleness and which silence has become avoidance. The congregation is not a target audience. It is a community the preacher is called to know and serve.

Homiletical education must assess process, not merely performance

Theological institutions should redesign preaching formation for the AI era. A final sermon manuscript can no longer reveal enough about the learner’s competence. Assessment should include an unaided interpretation of the biblical text, an exegetical record, source annotations, an AI-use declaration, oral defence, live preaching, audience or mentor feedback and reflective revision.

Within Competency-Based Theological Education, homiletical competence should integrate biblical interpretation, doctrinal faithfulness, communication, pastoral sensitivity, ethical source use, contextual application and responsible tool judgement. A student may use AI appropriately and still fail if the sermon misinterprets Scripture or harms listeners. Another may avoid AI entirely and still require substantial development. The assessed outcome is faithful preaching, not technological conformity.

Lecturers also need formation. They should demonstrate how to test generated interpretations, trace sources, disclose assistance, compare translations and protect pastoral data. Students need a model of disciplined use, not only a list of prohibitions.

Congregations deserve trustworthy transparency

Not every spelling correction or brainstorming prompt requires public announcement. Transparency should be proportionate to the significance of the AI contribution. If AI substantially shaped the sermon’s structure, wording, illustration or media, a simple acknowledgement may protect trust and teach responsible practice. Published sermons, courses and books should disclose material AI assistance according to their context and relevant standards.

Church leaders should agree on the principle before controversy arises. A short digital-ministry policy can distinguish administrative support, research assistance, substantial content generation, confidential information and prohibited uses. It should state who remains accountable, what requires review and how errors will be corrected.

Africa should develop a theology of resourceful integrity

African pastors often serve with limited staff, libraries, internet access and preparation time. AI can give a rural or bi-vocational preacher access to language support, outlines and research assistance that previously required resources beyond reach. That potential should not be dismissed from a position of privilege.

Resource constraints, however, do not remove the need for integrity. Imported systems may reproduce foreign assumptions, flatten African experience, misunderstand Pentecostal theology or create dependence on subscriptions and platforms controlled elsewhere. African churches and colleges should build approved source collections, Kiswahili and local-language resources, contextual evaluation rubrics and peer-review communities that place AI inside accountable relationships.

The Global South contribution is neither rejection nor imitation. It is resourceful integrity: using appropriate tools to widen capacity while keeping Scripture, community, local knowledge, human dignity and spiritual responsibility at the centre.

Adopt a seven-question sermon covenant

Before preaching an AI-assisted sermon, the pastor should ask: Have I studied the biblical text myself? Have I verified every source, quotation and factual claim? Can I explain and defend the interpretation? Does the application arise from truthful knowledge of this congregation? Have I protected confidential information and intellectual property? Is the level of AI assistance disclosed where it should be? Can I take personal responsibility before God and the Church for every word I will proclaim?

These questions do not make preaching mechanical. They create a boundary within which technology remains a servant. The aim is not to prove that the pastor worked without tools. It is to ensure that the tools did not displace the preacher’s calling, judgement or integrity.

The pulpit requires a person who has heard and will answer

AI will become more capable, personalised and integrated into ministry systems. It may learn a pastor’s archive, imitate a preferred style and generate material that sounds increasingly authentic. The Church must therefore define authenticity through more than stylistic resemblance.

The authentic sermon comes through a preacher who has encountered the text, submitted to its truth, listened to the people, tested the message and accepted responsibility for proclaiming it. AI may assist that labour, but it must not become the hidden preacher. The pulpit requires more than words that sound alive. It requires a living witness who has heard, discerned and is willing to answer for what is said.