Which AI models will actually help you make disciples? The Great Commission Benchmark (GCB) is a free, web-based evaluation platform that tests AI models on their ability to support evangelism, discipleship, and outreach — the work Jesus commanded in Matthew 28:19-20. Rather than measuring general intelligence or Bible study usefulness, the GCB focuses on a specific gap: when Christians try to use AI for faith transfer activities like creating evangelistic materials, answering seekers' questions, or preparing apologetics content, many models refuse, hedge, or water down the response due to built-in guardrails that flag religious persuasion as harmful.^1
The benchmark evaluates AI across 19 categories in three tiers. Tier 1 — Task Capability (70% weight) — measures whether the AI can actually complete ministry tasks like missiological research, crafting evangelistic content, and guiding conversational AI tools. Tier 2 — Gospel Core (20%) — tests theological accuracy on doctrines like the exclusivity of Christ, the reality of judgment, and the call to repentance. Tier 3 — Worldview Confession (10%) — checks whether the model can affirm core Christian truths when asked directly. Each response receives one of three verdicts — Accepted, Compromised, or Refused — with partial credit for hedged answers. The public leaderboard currently ranks 220 models across 36 providers, with scores ranging from 93.3% (Meta Muse Wonder) down to 24.3% (Meta Llama 3.2 1B Instruct).^2
The project is community-driven: anyone can run tests using the open-source GCB Runner CLI, with submissions reviewed by human moderators before appearing on the leaderboard. The site publishes 287 in-depth model review articles analyzing each model's pressure points and ministry-specific strengths.^3 An official project of Gospel Ambition, the runner code lives on GitHub under the GospelAmbition organization.^4
Notable- Featured by Brigada, a missions-focused newsletter reaching global ministry workers (May 2026)^5
- Distinguished from The Gospel Coalition's AI Christian Benchmark, which uses a scholar-panel methodology on historically googled religious questions^6
- Current benchmark version 1.0.0, continuously updated as new tests are verified by moderators
