Writing a Church AI Policy That Sounds Like Your Church: A Starter Kit

By Bobby Macintosh · Updated October 2026

Imagine three people in one small church, all in the same week. The prayer-chain coordinator pastes the week’s requests into a chatbot to tidy them up before emailing them round. A youth leader makes an AI picture of “our youth group on the beach” for the summer camp poster. The treasurer gets a voicemail in the minister’s voice asking for an urgent transfer before Sunday. None of them means harm. None has been told what the church expects.

Barna’s October survey found that 94% of US Protestant pastors think churches should have AI guidelines, but only 4% say their own church has a formal policy, 17% have informal guidance and 79% have neither (Barna, 6 October 2026). Our guide to using ChatGPT for sermon preparation already looks at those figures and offers a short policy for the pulpit. This piece is for the rest of church life: pastoral data, children and youth, deepfakes, Bible copyright, buying tools and being honest about how things are made.

Below you’ll find three convictions, eight clauses, a one-page template and a prayer. One thing first: this is practical guidance, not legal advice. For data protection, safeguarding and copyright, check with your denomination and a qualified adviser where you are.

Why your church needs this now

The worry isn’t new. In Lifeway’s September 2025 survey of 1,003 Protestant pastors, 84% picked “AI generated content must be edited assuming it contains errors” as a concern about using AI in ministry (Lifeway, 21 April 2026). What’s new is real material to work from, all published in the last few weeks:

• Lutherans. On 15 October the Michigan District of the LCMS published “Artificial Intelligence and the Church” by Rev. Dr. Paul Biber (Michigan District). It builds on Resolution 4-09, which the Synod’s July convention adopted by 96.21% (LCMS Reporter, 20 July 2026).
• An ecumenical toolkit. The free Church AI Toolkit, convened by the TryTank Research Institute at Virginia Theological Seminary, offers sample policies and templates (Canadian Mennonite, 9 October 2026).
• The Orthodox. At their 3–6 October meeting, the Assembly of Canonical Orthodox Bishops of the USA “discussed the growing effects of artificial intelligence”, with concerns “that call for greater discernment and discussion” (Assembly of Bishops, 6 October 2026). That isn’t a policy, but the conversation has begun.

A policy won’t make anyone wise. But it can protect the people whose stories your church holds, and spare a nervous volunteer from guessing.

Start with what you believe

“Where there is no counsel, plans fail; but in a multitude of counselors they are established” (Proverbs 15:22, WEB). Don’t let one enthusiast write this alone. Bring in the elders, vestry or parish council, and whoever actually handles the prayer list, the youth rota and the bank account.

The Toolkit’s “Naming Our Values” exercise is a good way in (Church AI Toolkit). In four steps you tell stories of the congregation at its best and name the values at work, cluster them down to five to seven, test them against real cases, and notice where your honest answers don’t trace back to any named value. Then build on three convictions that churches widely share.

People, not tools

LCMS Resolution 4-09 encourages “all who employ these tools to do so in honest, transparent, and Christ-like ways that focus on loving God and loving our neighbor as ourselves (Matt. 22:37–40)”. It also asks for a study rooted in “anthropology, the image of God, and Lutheran teaching on vocation” (resolution text, as reproduced by Ad Crucem News). Southern Baptists used almost the same words in their 2023 resolution (SBC).

Biber wrote his article with ChatGPT and marked his own comments as he went. He approves the machine’s rule of thumb: “Use AI to assist your vocation, not to abdicate your vocation.” His own note points to Acts 6, where the apostles handed the daily distribution to others: “It is not appropriate for us to forsake the word of God and serve tables” (Acts 6:2, WEB). Notice why: so that the widows were fed and the apostles could “continue steadfastly in prayer and in the ministry of the word” (6:4). A good AI policy asks the same question: what does this tool free us for?

Truthfulness

Pope Leo XIV’s encyclical says “Christian communities, too, are called to commit themselves to transparency in communication and to the honest pursuit of facts” (Magnifica Humanitas §138; see our reader’s guide to Magnifica Humanitas). The next sentences, on the Church’s failures over abuse, are sobering. For a church, transparency is a safeguarding matter as much as a communications one. Antiqua et Nova is simpler still: “the use of AI should always be transparent and never misrepresented” (§84).

Accountability

The Rome Call for AI Ethics (2020) lists accountability among its six principles: “there must always be someone who takes responsibility for what a machine does” (Rome Call). The ERLC’s evangelical statement says the same in its own words: “We deny that humans can or should cede our moral accountability or responsibilities to any form of AI” (Article 3, ERLC). So every clause below needs a named person, not just a rule.

The eight clauses every church policy needs

1. Permitted uses, and who approves new tools

Start with yes. Biber’s article calls editing newsletters and handouts “probably one of the safest and most practical uses”, and lists admin, captions, transcripts and plain-language versions. Then say who approves a new tool. The Toolkit’s Governance & Procurement Standard is blunt: “A church should never approve an AI product merely because it is popular, inexpensive, easy to use, described as ‘Christian,’ or recommended by another congregation” (Church AI Toolkit). Ask where your data goes, how long it’s kept, what it costs and how you’d leave.

2. Not permitted

Traditions differ here, and your policy should sound like yours. The LCMS Commission on Theology and Church Relations, in a study document that doesn’t claim to be an official Synod position, “strongly urges against Synod pastors preaching sermons generated using this technology” (CTCR, December 2024). The Catholic Diocese of Fort Worth’s AI Use Policy, effective 1 April 2026 for employees (including clergy) and volunteers, goes further: those “engaged in preaching or pastoral ministry must not use AI to write or generate homilies, sermons, spiritual reflections, sacramental guidance, pastoral counseling responses, or other pastoral content” (Diocese of Fort Worth). Almost everyone can agree on two things: no AI impersonation of a real person, and no machine-made message passed off as personal pastoral care.

3. Confidentiality and pastoral care data

This is the clause where I’ll be firmest. Prayer requests, counselling notes, safeguarding records, health details and anything said in confession don’t go into general AI tools. The Catechism of the Catholic Church is plain: “The sacramental seal is inviolable”, and “Professional secrets must be kept” (CCC 2511). The Diocese of Fort Worth’s policy is just as plain: “Never input information or content related to confession, pastoral counseling, or other spiritual direction into any AI tool.” The ERLC’s statement defends privacy “even in the pursuit of the common good” (Article 8). Biber goes further than the machine-drafted text he annotated: “just don’t put any of that in.” For training, use invented cases, as our piece on using AI to train for biblical counselling with fictional case studies explains.

UK churches should know that personal data “revealing religious or philosophical beliefs” is special category data under UK GDPR, which gets extra protection (ICO). Much of what a church knows about its members is likely to count.

4. Verification

Biber’s article puts it plainly: “Treat everything AI produces as untrusted until checked.” His own note adds, “Check everything.” Name who checks what: every Bible reference, quotation, date and figure. “Test all things, and hold firmly that which is good” (1 Thessalonians 5:21, WEB). The sermon guide’s section on hallucinated verses and citations covers the checking itself.

5. Disclosure

Model it, as Biber did by marking his own words, and agree some categories. The United Methodist Discipleship Ministries policy, offered as one of the Toolkit’s downloads, sorts content into “Human-Generated with AI Assistance”, “Collaboratively Generated Content” and “AI-Generated Content” (AI in Digital Ministry download, v2.1). A one-line label is usually enough: “Drafted with AI assistance and checked by the editor.”

6. Bible translations and copyright

Crossway allows up to 500 ESV verses without permission, provided they’re not more than half of any book, not 25% or more of the work, and not in a commentary or reference work (Crossway). Biblica’s NIV guidelines are similar, but on a media platform the limit applies “to that platform as a whole, not to individual posts”, which matters for a church website. Biblica now also says “Any use of Biblica content in connection with artificial intelligence, machine learning, large language models, chatbots, or similar technologies requires a valid license” (Biblica). That paragraph is addressed to “Publishers, developers, ministries, and other organizations”, and doesn’t say whether a volunteer pasting one verse into a chatbot is covered. The simple path: use public-domain text (BSB, WEB or KJV) with AI tools.

7. Children, youth and vulnerable adults

The Toolkit’s Children, Youth & Family AI Safeguarding Standard offers model clauses worth borrowing (Church AI Toolkit). Among them: staff and volunteers “may not create synthetic images, voices, avatars, or recordings of minors without written parental or guardian consent and organizational approval”; “Adults may not establish private AI-mediated relationships with minors outside approved ministry channels”; and “Internal reporting never substitutes for a legally required report to child-protection authorities or law enforcement.” That rules out the youth leader’s beach poster unless parents have agreed in writing.

A UK note. The Toolkit’s contributors come mostly from US churches, and its safeguarding standard leans on US agencies alongside UNICEF. In England the statutory guidance is Working Together to Safeguard Children 2026 (GOV.UK), which applies to England only, so churches elsewhere in the UK should check their own nation’s guidance. Church of England parishes work within the National Safeguarding Team’s standards and codes of practice (Church of England). Whatever your denomination, tie this clause to your existing safeguarding policy and lead rather than creating a second system.

8. Incidents, deepfakes and review dates

Antiqua et Nova says of deepfakes: “While the images or videos themselves may be artificial, the damage they cause is real” (§87). The Toolkit’s incident template is practical: name two contacts, preserve the evidence, and stop replying to a suspected scammer or impersonator (template PDF). Its fraud kit suggests that no emergency request for money or confidential information be acted on “until the requester has been independently contacted through a previously verified channel”, and that two unrelated people approve electronic transfers above a set threshold (Church AI Toolkit). Our imaginary treasurer should ring the minister back on the number already in the directory. And set a review date: the tools will change faster than your constitution.

A one-page starter template

Copy this, cut what doesn’t fit, and rewrite it in your own church’s voice.

[Church name]: Our use of artificial intelligence

Adopted by [elders / PCC / vestry / parish council] on [date]. Owner: [name, role]. Review by: [date].

Why. We use tools to serve God and our neighbour, never to replace the people and callings God has given his church. [Add one or two sentences from your values exercise.]

1. What we welcome. Staff and volunteers may use approved AI tools for [editing, admin, notices, captions, translation]. [Role] approves new tools after checking data handling, cost and how we’d leave.
2. What we don’t do. We don’t [preach AI-generated sermons]. We don’t use AI to impersonate anyone, or present machine-made messages as personal pastoral care.
3. Confidentiality. We never enter prayer requests, pastoral or counselling notes, confessions, safeguarding records, health information or members’ personal details into AI tools. Training uses invented cases.
4. Checking. Everything AI produces is untrusted until [role] has checked it, including every Bible reference, quotation and figure.
5. Openness. When AI has substantially shaped something we publish, we say so: “[Drafted with AI assistance and checked by ___].”
6. Bible text. We use public-domain translations [BSB, WEB, KJV] with AI tools, and keep within publishers’ permissions everywhere else.
7. Children, youth and vulnerable adults. We don’t create AI images, voices or videos of children without written parental consent and [safeguarding lead] approval. No adult uses AI for private contact with a young person outside approved channels. Concerns go to [safeguarding lead], and legally required reports are always made.
8. Incidents. We act on no request for money, passwords or confidential information until we’ve called back on a number we already hold. Two people approve every electronic payment over [amount]. If something goes wrong, keep the evidence, don’t reply, and tell [incident contact, phone].

Signed: [name], for [council], [date].

A suggested exercise: red-team your draft

Once you have a draft, ask a general chatbot to find the holes, pasting in the policy text only: no names, no church data. I haven’t run this prompt for this article, so treat what comes back as a starting point, not a verdict.

Prompt: Find the gaps

Here is our church’s draft AI policy: [paste policy text only]. List five realistic situations in a small church this policy doesn’t cover, and for each, say which clause is closest. Do not invent laws, statistics or quotations. If you’re unsure whether something applies in [country], say so.

Note the tool, version and date. The chatbot can spot a gap. Deciding what to do about it is your job.

Frequently Asked Questions

Does a small church really need an AI policy?
Probably, yes. Even one page helps volunteers know what’s expected, especially with prayer lists, children’s photos and money.

Can church volunteers put prayer requests into ChatGPT?
I’d say no. Prayer requests are confidences, and the Michigan District article’s advice, as Biber edited it, is not to put such material into AI at all.

Can we use the NIV or ESV with AI tools?
Biblica now says any use of its texts with AI or chatbots needs a licence, though it’s unclear how far that reaches individuals. Crossway’s permissions page doesn’t mention AI. Public-domain translations such as the BSB, WEB and KJV avoid the question.

Is the Church AI Toolkit free to use?
Yes. It’s licensed CC BY-NC-SA 4.0 for non-commercial use with attribution. TryTank’s executive director, Lorenzo Lebrija, asked churches to treat it as a beta and to “try to break it”. Parts of it are US-focused.

A prayer for the people who’ll write it

A policy is a small thing. It won’t make anyone holy, and it’ll need rewriting before long. But writing one together is an act of care for the flock, and worth praying over. You’re welcome to use or change this:

Lord God, you know each person in our care by name, and you have trusted us with their stories. Give those who write this policy patience with one another and honesty about what we don’t yet understand. Keep us from fear and from carelessness. Help us guard what is confided to us, tell the truth about how our work is done, and protect the young and the vulnerable among us. When the tools change, keep our hearts set on loving you and loving our neighbour. And when the work is finished, remind us that wisdom comes from you. Through Jesus Christ our Lord. Amen.

Sources and Further Reading

• Barna Group, “Pastors Say Churches Need AI Guidelines. Few Have Them.” (6 October 2026).
• Lifeway Research, “Pastors, Churchgoers See AI as Concerning and Confusing” (21 April 2026), with the survey report PDF.
• Paul Biber, “Artificial Intelligence and the Church“, Michigan District, LCMS (15 October 2026).
• LCMS Reporter, “2026 LCMS convention news: Monday, July 20” (20 July 2026); Res. 4-09 text via Ad Crucem News (independent, not affiliated with the LCMS).
• LCMS Commission on Theology and Church Relations, Online Technology in the Church: Study Materials (December 2024).
• Southern Baptist Convention, “On Artificial Intelligence and Emerging Technologies” (15 June 2023).
• ERLC, “Artificial Intelligence: An Evangelical Statement of Principles” (2019).
• Church AI Toolkit (TryTank Research Institute, Virginia Theological Seminary), Organizational Preparation resources, including the PCUSA sample policy (accessed October 2026).
• Canadian Mennonite, “The Church AI Toolkit” (9 October 2026).
• Assembly of Canonical Orthodox Bishops of the USA, press release (6 October 2026).
• Leo XIV, Magnifica Humanitas (15 May 2026), §138.
• Dicastery for the Doctrine of the Faith & Dicastery for Culture and Education, Antiqua et Nova (28 January 2025), §§84, 87.
• Catechism of the Catholic Church, 2511.
• Rome Call for AI Ethics (28 February 2020).
• Catholic Diocese of Fort Worth, AI Use Policy (effective 1 April 2026).
• Crossway permissions and Biblica permissions (both checked October 2026).
• Information Commissioner’s Office, “What is special category data?“; Department for Education, Working Together to Safeguard Children (2026 edition); Church of England, Safeguarding.

Scripture quotations are from the World English Bible (WEB), which is in the public domain.

Keep reading on AI Prompt Theory: Using ChatGPT for Sermon Preparation: A Complete, Theologically Careful Guide · Using AI to Train for Biblical Counselling: Fictional Case Studies, Real Care · Magnifica Humanitas: A Reader’s Guide to Pope Leo XIV’s Encyclical on AI