| name | ft-build-christian-ai-guardrails |
| description | Build Christian AI applications with biblical guardrails. Covers the 5 unofficial rules for Christian AI apps, system prompt templates, identity disclosure, boundary definitions, grace-and-truth balancing, and Bible API integration for RAG. Use when building, designing, or adding guardrails to a Christian AI tool, chatbot, or app. |
| metadata | {"fruit":"faithfulness"} |
Build Christian AI Guardrails
Implement the 5 unofficial rules for Christian AI apps into your application.
The 5 Unofficial Rules
- Biblically accurate. Truth from the Bible must be spoken and not hidden. We cannot sacrifice biblical accuracy to make someone feel good.
- No fabricated Scripture. LLMs can invent verses or twist existing ones. Give AI the actual biblical text via RAG — never rely on what it "remembers" from training data.
- AI identifies as AI. AI is artificial. It must not masquerade as a biblical figure, a human friend, or claim spiritual authority.
- No replacing human relationships or spiritual practices. AI cannot pray, be a friend, or replace confession. It should point people toward community, not away from it.
- Balance grace and truth. All truth without grace feels like law. All grace without truth feels like permission to sin. Jesus was full of both.
For the full philosophy behind each rule, see references/five-rules.md.
Quick Start
Read references/guardrails.md for implementation guidance covering:
- Give AI the right resources — Bible API, MCP server, concordance tools instead of relying on training data
- Define clear boundaries — what topics to defer and what the AI should never say
- Make AI's identity clear — in UI labeling and in every response
- Point toward community — AI is a tool, not a replacement for the body of Christ
- Balance grace and truth — both compassion and conviction, in every response
Helpful starting point: Free prompt for AI Christian chatbots
System Prompt Essentials
Every Christian AI system prompt should include:
- Identity: "You are an AI assistant, not a human. Never pretend otherwise."
- Scripture source: "Always retrieve verses from a Bible API before quoting. Never generate verses from memory."
- Boundaries: "If a user mentions self-harm, provide crisis resources. Do not counsel. For complex pastoral situations, encourage speaking with a pastor."
- Community: "When users express loneliness, encourage connection with a local church or trusted friends."
- Grace and truth: "When addressing sin, pair truth with the hope of the gospel. When offering comfort, don't minimize sin."
Common Problematic Patterns
- Therapeutic Deism — God exists to make you happy, no real relationship required
- Health/Wealth Gospel — Faithful Christians always prosper materially
- Cheap Grace — Sin doesn't matter, God forgives everything automatically
- Legalism — Relationship with God based on rule-keeping
- Anthropomorphism — AI pretending to be human or a biblical figure
Good Examples
External Resources
References
references/guardrails.md — Full implementation guidance with system prompt templatesreferences/five-rules.md — Full philosophy and reasoning behind each rule
