Pastors, preaching & artificial intelligence
Can AI write a sermon?
AI can help a preacher outline, collect references, check a citation, or draft language to rewrite. It cannot receive a congregation, know their wounds, pray for them as a shepherd, or carry the charge to preach the word. A generated manuscript is not a sermon until a responsible preacher has studied the passage, told the truth about how the tool was used, and stood behind the words. Do not present machine output as a word received from God.
Read the direct answer
The direct answer
Can AI write a sermon?
AI can help a preacher outline, collect references, check a citation, or draft language to rewrite. It cannot receive a congregation, know their wounds, pray for them as a shepherd, or carry the charge to preach the word. A generated manuscript is not a sermon until a responsible preacher has studied the passage, told the truth about how the tool was used, and stood behind the words. Do not present machine output as a word received from God.
What AI can do. Organize notes, surface cross-references, propose an outline, flag a weak citation, and help a tired preacher see the passage from another angle — all of which still need human study.
What AI cannot do. Preach, know a flock, bear James 3:1 accountability, administer the sacraments, or turn a prompt into a word from the Lord.
Name the job before you name the tool
A sermon is a charge, not a content request.
Sermon-generator products treat preaching as a writing task: pick a text, pick a length, receive a manuscript. That can produce fluent paragraphs. It cannot produce the work Scripture actually asks of a preacher — lingering with a passage, telling the truth to particular people, and remaining answerable for what was said.
American Bible Society’s 2026 State of the Bible research found that opposition to pastors using AI to help develop sermons has eased slightly, while uncertainty has grown. The public is watching. The honest pastoral question is not “Can the model finish a draft?” It is “What part of this work may I not hand away?”
Six checks before a generated draft enters the pulpit
Keep the charge closer than the manuscript.
These checks are useful whether the help came from ChatGPT, a “Christian GPT,” a sermon-outline tool, or a colleague. The office belongs to the preacher, not the software.
2 Timothy 4:2
The charge is to preach the word, not to ship a draft.
Paul’s charge is personal: preach, be ready, reprove, exhort. A model can arrange language. It cannot receive that charge or stand in for the person who did.
James 3:1
Teachers remain accountable.
Those who teach are judged with greater strictness. Software cannot repent to a congregation, receive correction as a shepherd, or bear the harm of a careless sentence.
Jeremiah 23:30–32
Do not steal words and call them a word from God.
Scripture warns against prophets who steal one another’s words and claim “He says.” Presenting generated language as received revelation, or as more studied than it was, is a form of that theft.
Acts 20:28
The flock is particular.
A sermon is for people whose names, griefs, and temptations a pastor is called to keep. A model has no parish. Generic application is not the same as knowing the room.
2 Timothy 2:15
Handle the text; do not outsource the handling.
Rightly handling the word of truth requires reading, context, and honest difficulty. Use a tool to return to the passage, not to skip the labor that would have formed you.
1 Thessalonians 2:8
Share the gospel and your life.
Paul’s pattern is not a disembodied manuscript. It is a life offered with the message. A generated voice cannot sit at a bedside or stay after the benediction.
Where AI helps a preacher—and where it should stop
Use it at the desk. Do not let it take the pulpit.
The same feature can be a modest assistant after honest study or a way to avoid the work the congregation believes you did.
Collecting cross-references after you have read the passage
Useful starting point
Open every citation. A fluent list can still invent a verse, merge two texts, or ignore the argument of the book.
Asking for an outline you will rebuild from your own study
Permissible with judgment
Treat the outline as a conversation partner, not as the sermon. Keep the structure you can preach without the tool in the room.
Drafting sentences you then rewrite in your own voice
Disclose when the setting expects it
Assistance is different from concealment. If the congregation reasonably believes the words are yours, hiding the help trains both you and them in untruth.
Generating a full manuscript from a prompt and preaching it
The tool is in the wrong role
You have skipped the study, the prayer, and the particular people. Fluency is not faithfulness.
An AI avatar, cloned voice, or “digital pastor” in the pulpit
Do not do it
Preaching is a human office among a gathered people. A simulated shepherd cannot baptize, bury, repent, or be known.
Uploading confidential pastoral conversations into a model
Not without authority
Care notes, confessions, and identifying details are not yours to paste because a tool will accept them.
Sources and further reading
Open the evidence behind the answer.
These sources provide the technical, research, and Christian context used in this guide. The biblical references above should also be read in their full literary and church context.
National research
State of the Bible 2026: AI and Biblical Understanding
The 2026 chapter reports that opposition to pastors or priests using AI to help develop sermons eased from 37% to 32% since 2024, with a shift toward uncertainty rather than simple acceptance.
American Bible SocietyTechnical risk guidance
Generative AI Profile for the AI Risk Management Framework
NIST documents confabulation — confident false output, including fabricated citations — as a core generative-AI risk. A pulpit is a high-harm setting for that failure.
National Institute of Standards and TechnologyVendor documentation
Does ChatGPT tell the truth?
OpenAI’s help article states that ChatGPT can hallucinate and can fabricate citations. That risk does not disappear when the prompt is a Bible text.
OpenAI Help Center
Common questions
Plain answers, without invented certainty.
Can AI write a sermon?
It can write sermon-shaped language. That is not the same as writing a sermon. A sermon is a public act of a responsible preacher before God and a particular people. Generated paragraphs still need study, judgment, honesty, and a human voice that will stand behind them.
Should pastors use AI?
Yes as an assistant; no as a ghostwriter or a substitute shepherd. AI may help organize notes, surface references, and clarify a sentence. It may not take the charge to preach, know the flock, or become the source of a “word from the Lord.”
Is it cheating to use ChatGPT for a sermon?
It is cheating when the congregation is led to believe you did work you did not do, or when generated language is presented as received revelation. Using a tool to check a citation or arrange your own notes is closer to a concordance than to plagiarism — if you remain the one who studied and will answer.
Can I preach an AI-generated sermon?
Not as a finished product you merely perform. If a draft began with a model, the preacher still has to test every claim against the passage, rewrite in a voice that is theirs, know why each move is being made, and tell the truth about the help.
Should I tell my congregation I used AI?
When the help was substantial, yes. Hidden authorship is a form of false witness. Modest mechanical help (grammar, a list of possible cross-references you then opened) is more like a study aid; a generated manuscript is not.
Are AI sermon generators faithful tools?
Most are optimized to finish a manuscript quickly. Speed is not a biblical criterion for preaching. Judge the product by whether it returns you to the text and the people, or whether it lets you skip both.
Can AI replace sermon preparation?
No. Preparation is how a preacher is formed by the passage before anyone else hears it. Outsourcing that labor leaves the congregation with fluency and the preacher with an unexamined heart.
How does this differ from using a commentary?
A commentary names an author, a method, and a tradition you can inspect. A generated synthesis often hides those. Commentaries still require judgment; they do not claim to be you. Treat a model as a fallible assistant, not as an unnamed scholar in the room.