Most of these tools are simply told to behave. This one is built so it can't say something that isn't yours — and it won't invent doctrine, take positions you haven't published, or speak for your pastor. Here's the honest mechanism, not a marketing promise.
Built from your website, your sermons, your ministries — and nothing else. Ask it something outside your content and it tells you, instead of guessing.
After it drafts a reply, a second pass reads that reply back against your content. If any part isn't supported by your own material, the answer is held and the person is handed to a human. A step that runs on every answer — not a rule it's asked to follow.
Answers show where they came from — your page, your sermon — so a member (or your staff) can check the receipt. Trust you can verify, not trust you're asked for.
It's your content and your assistant. See something off? One correction, in plain English, and it's fixed — no ticket, no wait.
If someone types something that signals real distress — suicide, abuse, self-harm — in English or Spanish, they are routed to the 988 Suicide & Crisis Lifeline, the Crisis Text Line, and your own pastoral care team before the AI ever replies. It's hard-wired in code to run first, and it can't be talked out of it.
A deterministic check reads every incoming message for acute-risk signals before a single token of AI runs. Match → the person gets a real crisis response and your care contact, and the moment is logged for your team. No model in the loop for the moments that matter most.
General questions are untracked and private — no names, no accounts. Someone is only identified when they choose to share their info to get prayer follow-up, serve, or plan a visit — and that goes to your team to act on, not to us. The everyday questions stay anonymous; we keep them only to improve answers.