The question behind the question
Most pastors asking about AI are not really asking whether the technology works. They are asking whether using it costs them something they should not spend. That is the right instinct. Preaching is not content production; it is the overflow of study, prayer, and pastoral knowledge of a particular people.
The useful distinction is between the work of the preacher and the work around the sermon. Wrestling with the text, discerning what this congregation needs to hear, and standing behind the claim — that is the preacher’s work, and no tool should touch it. Formatting, transcription, discussion questions, background research, and clarity edits are work around the sermon. AI is very good at the second category and quietly corrosive in the first.
A six-step workflow
01
Study first, prompt second
Read the passage, do your own exegesis, and write a one-sentence claim before opening any AI tool. Prompting before study lets the model set the direction of the sermon, and its instinct is toward the generic. Your claim is the anchor everything else serves.
02
Use AI to widen the research, not decide it
Ask for historical and cultural background on the passage, a summary of how major interpretive traditions have read a difficult verse, or a list of cross-references on a theme. Treat every result as a lead to verify, not a conclusion to quote.
03
Pressure-test your outline
Paste your own outline and ask where the logic jumps, which point is doing the least work, and what objection a skeptical listener would raise. This is where AI is genuinely strong — it is a tireless critic that never gets tired of your sermon at 11pm on Saturday.
04
Generate illustration candidates, then reject most
Ask for ten everyday analogies for your main idea, aimed at your actual congregation. Nine will be clichés. The tenth is often a usable seed you would not have reached alone. Never use an AI-supplied anecdote as if it happened to you.
05
Verify every reference
Open your Bible and check each citation. Search for each quotation attributed to a theologian. Confirm each statistic at its source. Assume fabrication until proven otherwise — this single discipline prevents nearly every public failure associated with AI use.
06
Let AI handle the downstream work
After the manuscript is finished, AI can produce small-group discussion questions, a bulletin summary, social captions, and a transcript-based follow-up email in minutes. This is the highest-return use for most pastors, and it carries the least theological risk.
Prompts worth keeping
Vague prompts return vague sermons. These five cover most of what a preacher actually needs during the week. Replace the bracketed parts with your passage and your people.
Historical context
Give me the historical and cultural background a first-century audience would have brought to [passage]. List each claim separately so I can verify it, and note where scholars disagree.
Outline critique
Here is my sermon outline on [passage]. My main claim is [claim]. Identify where the logic jumps, which point is weakest, and what a thoughtful skeptic in the room would object to.
Illustrations
Suggest 10 everyday analogies that illustrate [main idea] for a congregation of [describe your people]. Avoid sports and war metaphors. Mark any that are overused.
Application
Based on this main idea, draft 6 specific application questions for small groups — three for personal reflection, three for group discussion. Keep them concrete, not abstract.
Clarity pass
Read this section of my manuscript aloud in your head. Flag sentences that would be hard to follow when heard rather than read, and suggest simpler phrasings that keep my voice.
Where AI gets it wrong
The most common failure is fabrication. A language model predicts plausible text, so it will confidently produce a verse reference that does not say what it claims, a Spurgeon quote Spurgeon never wrote, or a statistic with no source. These errors are fluent, which is exactly what makes them dangerous from a pulpit.
The second failure is flattening. AI writes toward the average sermon on a passage, because that is what it learned from. Left unedited, it produces preaching that is competent, forgettable, and addressed to no one in particular. Your congregation does not need the average sermon on Philippians 4; they need yours.
The third is confidentiality. Anything typed into a consumer AI tool may be retained or reviewed. Counseling notes, congregant names, and pastoral situations do not belong in a prompt — not in a paraphrase either.
A short rule of life for AI use
- The sermon's claim, exegesis, and conviction originate with you — always.
- Every verse, quotation, and statistic is verified against a primary source.
- No AI-generated story is ever presented as personal experience.
- No pastoral care, confession, or congregant detail is pasted into an AI tool.
- Prayer and study time are protected, not replaced by the time AI saves.
Frequently asked questions
- Is it wrong for a pastor to use AI to write a sermon?
- Using AI to draft a sermon for you and preaching it unchanged bypasses the study and prayer that shape a preacher. Using AI as a research assistant — surfacing cross-references, summarizing commentary positions, tightening an outline you wrote — is closer to using a concordance or a study Bible. The dividing line is authorship: the conviction, exegesis, and application must be yours.
- Can AI be trusted with Scripture references?
- No, not without verification. Language models regularly fabricate verse numbers, misattribute quotes to theologians, and invent statistics that sound plausible. Every reference an AI produces must be checked against your Bible and primary sources before it reaches the pulpit.
- Should I tell my congregation I used AI?
- If AI functioned as a study tool, most pastors treat it like any other resource and do not announce it. If AI generated substantial content you delivered as your own reflection, disclosure is the honest path. A simple internal standard — 'would I be comfortable if this were known?' — resolves most cases.
- Which AI tool is best for sermon preparation?
- General assistants such as ChatGPT, Claude, and Gemini all handle outlining, summarizing, and rewriting well. Purpose-built sermon tools add denominational settings and commentary libraries but are not necessary to start. Begin with one general tool, learn to prompt it well, and add specialized tools only when a specific gap appears.
- Will AI make preaching less personal?
- It can, if it replaces reflection. Used to remove administrative friction — formatting, transcription, discussion questions, social copy — it usually returns hours to study and prayer, which makes preaching more personal, not less.