A reflection on prompts, research maps, interpretive debates, testing, and transfer as ways AI can help preachers read, think, and prepare sermons more deeply.
By Muchangke
By Muchangke (Paul Zhang)
A few days ago, one of my students asked me:
“Teacher, AI has become so powerful. How should we use AI to write sermons?”
I told her, “When using AI to prepare a sermon, the most important thing is not to let AI write it for you. The most important thing is what kind of prompts you give it.”
She immediately asked a second question:
“Then how can I learn to write good prompts?”
That question is actually more important than the first.
Because a “prompt,” when you get down to it, is not some mysterious AI technique. A good prompt is essentially a good question. And behind a good question are a person’s genuine observations, thoughts, uncertainties, and pursuit of understanding.
If I have not carefully read a passage myself, have not observed it, prayed over it, meditated on it, or even discovered the questions within it, and I simply tell AI:
“Please write me a thirty-minute sermon based on this passage.”
AI can certainly do that.
It may even produce a complete structure—with an introduction, three main points, illustrations, applications, and a polished conclusion.
But it may only be a “correct sermon.” It may not truly be my sermon.
It has not passed through my struggles. It has not passed through my questions about the text. It has not collided with the realities of pastoral ministry. Nor has it passed through my prayer and reflection before God.
AI can draw on enormous amounts of public knowledge. But what a preacher truly has to face is something different:
What is God’s Word actually saying in this passage? What do I genuinely not understand? How does this text first judge me, comfort me, and change me? And what do the people God has entrusted to my care truly need to hear right now?
So I increasingly believe that, in the age of AI, what separates one person from another may no longer be simply “who knows more,” but rather:
Who can discover the real questions, ask the real questions, and keep pursuing those questions until they have studied them deeply?
Recently, I came across an interesting approach to learning with AI.
The first question is not:
“Please summarize the knowledge in this field for me.”
Instead, it is:
“Based on these materials, what are the five most important core mental models in this field?”
I think this is an excellent question to learn from.
Traditional learning can easily become:
definition → information → notes → memorization → exam.
But someone who truly enters a field does more than accumulate facts. Gradually, that person begins to understand:
How do people in this field actually think?
For example, when studying a biblical passage, we should not only ask:
“What does this passage mean?”
We can go further:
What are the five most important exegetical questions that must be answered to interpret this passage?
What historical, literary, theological, and contextual frameworks do I need in order to understand it properly?
What role does this passage play within the book as a whole?
How does it relate to the redemptive history of the whole Bible?
Which words or sentences actually determine the direction of interpretation?
Now AI is no longer merely giving me a pile of answers. It is giving me a map for research.
Once I have the map, I can return to the biblical text, examine the original languages, consult commentaries, and read scholarly articles with a completely different level of efficiency.
So the first principle of learning with AI is this:
Let AI help me see the structure of knowledge before asking it to produce conclusions for me.
A second highly valuable question is:
“What are the three most important debates in this field, and what are the strongest arguments for each position?”
This is a powerful question.
A beginner usually sees only the “answers.” Someone who studies more deeply begins to see:
the arguments behind the answers.
For example, when preparing a biblical passage that has several competing interpretations, I can ask AI:
What are the major interpretations of this passage?
But I should not stop there.
I should continue asking:
What is the strongest biblical evidence for each interpretation?
How does each position handle the context?
What is the hardest problem each interpretation must explain?
How has the church understood this passage throughout history?
Are there significant differences among evangelical, Reformed, Wesleyan, or other major Christian traditions?
Which issues belong to the core consensus of the Christian faith, and which are secondary exegetical disagreements?
At that point, AI becomes a tool that helps me compare, distinguish, and investigate.
This is especially important for preachers.
One of the most dangerous habits we can develop is searching only for material that supports what we already believe.
I decide in advance what a passage means, and then ask AI to find ten pieces of evidence proving that I am right.
That is not research.
Real research must be willing to ask:
“What is the strongest objection to my interpretation?”
I can even tell AI:
“Do not flatter me. Assume you are a biblical scholar who disagrees with my interpretation. Challenge it from four angles: context, original language, historical background, and whole-Bible theology.”
That is where AI becomes genuinely useful.
Its job is not to help me win.
Its job is to help me discover:
Where are the weaknesses in my thinking?
The third question may be even more important than the first two:
“Design ten questions that can determine whether someone truly understands this subject or has merely memorized some information.”
Notice what changes here.
This time, do not let AI immediately give you the answers.
Let it question you.
For example, after studying a passage, I can tell AI:
“Do not give me any more information right now. Based on what I have just studied, design ten questions to test whether I truly understand this passage. Ask me only one question at a time. After I answer, do not immediately give me the model answer. First tell me where my reasoning is strong, where it is weak, and what I have overlooked. Then let me answer again.”
That changes everything.
Before, I was questioning AI.
Now AI is questioning me.
Before, AI was an answer machine.
Now it becomes my sparring partner, examiner, and Socratic tutor.
There is one principle in this method that I especially appreciate:
When I answer incorrectly, do not give me the answer immediately. Tell me where I need to go back and study again.
For example:
“Your answer overlooks the logical relationship between verses 8–10 and verse 11. Go back and reread the context.”
Or:
“Your interpretation explains this word, but it does not explain why the author uses this transitional term here. Observe the text again.”
This is how knowledge slowly becomes our own.
After the first three questions, I believe we should add a fourth:
“Give me a new situation that the original material did not directly discuss, and let me use the principles I have just learned to address it.”
Why?
Because being able to answer a question does not necessarily mean we have truly mastered the subject.
One of the clearest marks of genuine understanding is:
transfer.
After learning a principle, can I still use it when the context changes?
For example, suppose I have been studying how Jesus responded to marginalized people.
At the end, AI should not simply ask:
“How did Jesus respond to this person?”
Instead, it might ask:
“Suppose someone comes to your church whose cultural background, economic situation, and political views are completely different from yours. Based on this passage, how should your pastoral approach be challenged?”
Now the biblical text moves from being merely an “object of interpretation” into the realm of “life and pastoral ministry.”
That is where a sermon ultimately needs to go.
For ordinary learning, the first four steps are already very helpful.
But for sermon preparation, I believe one more step is essential.
The Bible is not an ordinary textbook, and a preacher is not merely an organizer of information.
So I need to ask:
“Before I preach this passage to others, how is this passage first challenging, correcting, comforting, or calling me?”
AI cannot answer that question for me.
It can help me formulate questions, but it cannot replace the work of the Holy Spirit. It can help me find resources, but it cannot pray for me. It can analyze problems that a congregation may be facing, but it cannot replace the real relationship between a shepherd and the flock.
So throughout sermon preparation, I believe we should preserve this order:
Read Scripture first, then ask questions.
Think first, then use AI.
Let the text speak to me first, then consider how to speak to the congregation.
The more powerful AI becomes, the more important this order becomes.
Now I can answer my student’s question.
She asked:
“Teacher, how do I learn to write prompts?”
My answer is:
First learn how to discover questions.
Prompts are not magic formulas.
Truly good prompts often grow out of five things:
What do I see?
What do I not understand?
Why do different viewpoints conflict?
Where are the weaknesses in my interpretation?
How does this truth enter a new situation?
If I do not have any real questions of my own, but I collect a hundred “ultimate AI prompts” from the internet, I may simply end up copying other people’s questions.
But when I genuinely read a biblical passage and suddenly stop at a certain point:
“Why does Jesus answer this way here?”
“Why does the author place these two stories together?”
“What does this word actually mean in context?”
“Why does Paul speak about grace and then immediately turn to obedience?”
“Between my interpretation and another interpretation, which one actually fits the context better?”
Then good prompts begin to emerge naturally.
Because the prompt is no longer merely a technique.
It has become part of the thinking itself.
If I were to summarize my own approach to learning with AI, I would reduce it to five steps:
First, build the map: What are the most important mental models in this field?
Second, find the debates: Where are the real disagreements, and what is the strongest evidence on each side?
Third, be tested: What questions would prove that I truly understand it?
Fourth, practice transfer: Can I still use what I learned in an unfamiliar situation?
Fifth, bring it back to myself: How does this truth change my understanding, my life, and my practice?
These five steps are useful for more than sermon preparation.
We can use them to study theology.
We can use them in doctoral studies.
We can use them to explore an unfamiliar field.
We can use them to learn English.
We can even use them when trying to understand a social phenomenon.
I increasingly resist thinking of AI as a machine that does my work for me.
I would rather see it as an extraordinarily capable learning partner.
I read; it helps me notice what I missed.
I think; it challenges my logic.
I ask; it helps me build a map.
I answer; it points out the weaknesses in my reasoning.
I write; it helps me check my work.
But in the end:
I am still the one who reads.
I am still the one who thinks.
I am still the one who makes judgments.
And I am still the one who bears responsibility.
This is especially important for preachers.
We must never hand the sermon over to AI simply because AI can generate one in a minute.
On the contrary, the more powerful AI becomes, the more deeply preachers need to enter the biblical text, the more genuinely they need to shepherd people, the more theological discernment they need, and the more they need to become real people before God.
Because in the future, what is truly scarce may no longer be “information.”
Information is becoming cheaper.
Answers are becoming easier to obtain.
What is becoming increasingly valuable is this:
Can a person ask real questions?
Does that person have the patience to pursue real answers?
Does that person have the wisdom to discern among different voices?
Does that person have the courage to let truth change him or her first?
So I increasingly believe that the best learning in the age of AI is not to let AI know more for me, but to use AI to push myself to think more deeply.
It is not to let AI write for me, but to let AI help me become someone who reads better, asks better questions, thinks more carefully, and communicates more clearly.
And for preachers, I would add one more sentence:
The best AI prompts are not copied from a collection of prompts. They grow out of the “why” that begins to trouble you after you have seriously read the biblical text—the question that keeps pressing you to search further.
Perhaps this is the kind of “prompt engineering” we truly need to learn in the age of AI.
Allow Google Analytics to help us understand anonymous website usage? Prayer text, form content, and other sensitive text are not sent to analytics.
