ChatGPT Prompts for Sermon Prep: 15 That Actually Work (by Stage)

A pastor types "write me a sermon on grace" into ChatGPT, gets back 800 words of theological oatmeal, and concludes the whole thing is useless for preaching.

The tool wasn't the problem. The prompt was.

A good sermon-prep prompt isn't a magic spell — it's a well-briefed request. You wouldn't hand a research assistant the words "grace, go" and expect anything usable. You'd give them the passage, your angle, your congregation, and the exact output you need. That's all a prompt is.

Below are 15 prompts that actually earn their keep, organized by the stage of prep where you'd use them. None of them write your sermon. Every one of them hands you back time you'd have spent on mechanical work — so you can spend it on the text, on prayer, and on your people.

The one rule that makes every prompt better

Before the list, the rule that matters more than any individual prompt: give the model everything you already have.

The difference between a generic answer and a genuinely useful one is context. Start most of these prompts by pasting a short brief like this:

Context for everything below:
- Passage: [book chapter:verse]
- Translation: NIV
- Central truth I'm landing on: [one sentence]
- Congregation: [e.g. small rural church, mostly 50+, going through a building campaign]
- Sermon length: [25 minutes]
- Style: [expository / topical / narrative]

Paste that once, then run the prompts underneath it. The model will carry the context forward in the same chat. Skip it, and you get oatmeal.

Stage 1 — Understand the text

1. The context brief.

Give me a 200-word background brief on [passage]: author, audience, occasion,
and where it sits in the book's flow. Flag anything in the historical or
cultural setting a modern congregation would miss. Cite the kind of source
each claim comes from so I can verify it.

Why it works: it front-loads the study you'd do across three commentaries, in a form you can scan in a minute. Watch out: verify names, dates, and "cultural background" claims against a real commentary — this is exactly where models invent confident-sounding detail.

2. The cross-reference map.

List 6-8 cross-references that genuinely illuminate [passage] — not just verses
that share a word. For each, give one line on how it connects. Group them by
theme.

Why it works: it surfaces connections you might not hold in memory. Watch out: open every reference in your own Bible. Models routinely cite verses that are thematically close but say something slightly different — or don't exist.

3. The word-study lookup.

Identify 2-3 key words in [passage] in the original language that carry weight
for the meaning. For each: the word, a transliteration, its semantic range, and
why it matters here. Keep it to what a lexicon would actually support.

Why it works: it points you to the words worth studying. Watch out: treat this as a pointer, not an authority. Confirm the lemma and gloss in a real lexicon before you say it from the pulpit — a made-up Greek word repeated with confidence is a credibility landmine.

Stage 2 — Build the structure

4. The outline draft.

Draft a 3-point expository outline for [passage] that serves my central truth
above. For each point: a memorable heading drawn from the text, the main idea,
the supporting verses within the passage, and one application question. Keep the
movement faithful to the passage's own structure.

Why it works: it gives you a scaffold to react to — often you'll know your real outline the moment you see one that's almost right. Watch out: the outline is a starting point, not the sermon. If you preach it as-is every week, you've stopped doing the wrestling that makes you the pastor of your church. (More on that line in Using AI for Sermon Prep Without Losing Your Voice.)

5. The big-idea sharpener.

Here's my central truth: "[your sentence]". Give me five sharper one-sentence
versions — each under 15 words, each preachable, each faithful to [passage].
Then tell me which one is clearest and why.

Why it works: your big idea is the single most important sentence in the sermon, and AI is genuinely good at compression. Watch out: pick the one that's true, not just the one that's catchy.

Stage 3 — Find illustrations

6. The illustration angles.

Suggest sermon illustration angles for the theme "[theme]" in [passage], across
these categories: everyday life, history, science/nature, sports, and current
culture. Give a one-line seed for each — not a full story. Avoid anything I'd
need to fact-check as a specific quote or statistic.

Why it works: it jolts you out of your default illustration well. Watch out: never preach a named quote, stat, or "true story" the model produced without verifying it independently. Invented quotations attributed to real people are the most common AI failure in sermons.

7. The modern analogy.

Give me three modern-day analogies for [the theological concept], each one an
ordinary situation my congregation ([context]) would recognize from this week.
No clichés (no "GPS," no "bank account").

Why it works: analogies are the bridge between an ancient text and a Tuesday, and this forces freshness. Watch out: test each one against your actual people — a great analogy for a suburban church can fall flat in a rural one.

Stage 4 — Drive the application

8. Application by life-stage.

For the central truth above, give me concrete applications for four groups in my
congregation: a teenager, a young parent, someone mid-career, and a retiree
facing loss. One specific, this-week action per group.

Why it works: it stops your application from defaulting to one generic "us." Watch out: keep the ones that fit your people; cut the rest. You know the room.

9. The "so what / now what."

Read my outline above. For each point, tell me honestly: is the application
specific enough to obey by Friday, or is it still an abstraction? Rewrite any
vague ones into a concrete next step.

Why it works: vague application is the quiet killer of good sermons, and AI is a blunt, useful critic here. Watch out: none — this is one of the safest uses on the list. It's editing your own work.

10. Anticipate the objections.

What are the three most likely objections or honest questions a thoughtful
skeptic in my congregation would raise against the claim of [passage]? For each,
how would I address it briefly and fairly within the sermon?

Why it works: preaching to the objection in the room builds trust. Watch out: you're looking for real questions your people carry, not debate-club points.

Stage 5 — Sharpen the delivery

11. The clarity editor.

Here's my sermon manuscript. Don't rewrite it. Tell me: where am I being too
academic, where am I burying the main point, and where does a paragraph run long
enough to lose people. Quote the specific lines.

Why it works: this is AI as coach, not co-author — you wrote it, it's pressure-testing it. Watch out: take the diagnosis, keep your voice. Don't let it flatten the lines that sound like you.

12. Three openings.

Give me three different ways to open this sermon in under 90 seconds: one with a
question, one with a scene, one with a tension. Each should land the listener on
my central truth. No throat-clearing.

Why it works: the first 90 seconds decide whether people lean in, and options beat a blank page. Watch out: the opening has to be true to you and, if it's a story, true period.

Stage 6 — Keep the sermon alive after Sunday

This is the safest and highest-leverage zone for AI: the sermon already exists, so the model is reformatting your work, not inventing content.

13. Small-group questions.

From my sermon (pasted below), write 6 discussion questions for a mixed-age
small group: two observation, two interpretation, two application. Open-ended,
no yes/no. End with one question that sends people out to act.

Why it works: it turns Sunday into Wednesday's discussion in minutes. Watch out: almost none — but skim for questions that assume knowledge your group doesn't have. (Full method: Small Group Discussion Questions From Your Sermon.)

14. Five devotional seeds.

From my sermon below, draft five short daily devotional seeds — one per weekday.
Each: a verse from or near the passage, a 120-word reflection in a warm pastoral
tone, and one reflection question. Keep them tied to Sunday's central truth.

Why it works: a weekly devotional dies for lack of a source; your sermon is the source. Watch out: read them for tone before you send. (Full template: Turn Sunday's Sermon Into a 5-Day Devotional.)

15. Social captions.

Pull three shareable quotes from my sermon below (the actual lines, not
paraphrases). For each, write a short Instagram caption that sets it up, plus 3-5
relevant hashtags. Keep my voice.

Why it works: it fills your church's feed from work you already did. Watch out: if it's posted as your voice, the words should be your words — edit, don't rubber-stamp. (More: Church Instagram in 90 Seconds.)

The catch every one of these shares

Notice how many "watch out" lines say the same thing: verify the facts.

AI models are fluent, confident, and occasionally wrong in ways that are hard to catch precisely because they sound right. A misquoted verse, a Greek word that doesn't mean what it claims, a "true story" that never happened — these don't announce themselves. They just show up in your sermon and cost you trust when someone in the third row notices.

So the rule that governs this whole list: the model can draft, suggest, and reformat. You verify, discern, and preach. Where to draw that line in detail is its own subject — we wrote a full framework in AI for Pastors: What's Helpful, What's Risky, and Where to Draw the Line.

The faster path (fewer prompts, same output)

Here's the honest downside of a prompt library: you're the one pasting your passage into fifteen different prompts, one at a time, chasing the context across a chat window — every single week.

That's the exact chore Sermoneer removes. You give it one sermon — a manuscript, a document, or a YouTube link — and it runs the "after Sunday" stages for you at once: the small group guide, the five-day devotional, the social cards and shorts, the summary, the outline. Same work these prompts do, without the copy-paste marathon, and with your sermon as the fixed source so it's reformatting your content rather than inventing new content.

The prompts above are the manual version, and they work. Sermoneer is the version for the week you don't have a spare hour to run them.

Either way, the principle holds: let AI carry the mechanical load so you can carry the text.