Trust and safety
AI disclosure
The assistant says it is an AI in the first sentence of every call, and that cannot be switched off.
Every call opens with the assistant saying it is an AI. Not in the small print, not if asked — in the first sentence, on every call, to every caller.
This is not a setting. There is no toggle, no per-community exception and no "quiet mode".
What a caller hears#
The greeting is always assembled in the same order:
- The AI disclosure line — who is speaking and that it is an AI assistant.
- The recording notice.
- The opening question.
For example:
"Thanks for calling Maple Grove Senior Living, I'm Anna, the community's AI assistant. This call may be recorded. How can I help today?"
You write the words. You cannot write words that leave the AI out.
How it is enforced#
As you type
A chip beside the disclosure field reads Says it is an AI or Must say it is an AI. It checks for an explicit mention: "AI", "A.I.", "artificial intelligence", "automated assistant" or "virtual assistant".
Before you can publish
Says it is an AI in the first sentence is the first item on the publish checklist. If the disclosure does not mention an AI, publishing is blocked with: The first line must tell the caller they are speaking with an AI assistant.
Locked against communities
The disclosure line and the recording notice are locked by default, so no individual community can change or remove them. Only an account owner or admin can unlock them, and doing so still cannot bypass the check above.
On the call itself
The disclosure is spoken word for word as the first thing the assistant says, before anything else happens.
You will also get a warning — not a blocker — if the disclosure runs to more than one sentence: Keep the AI disclosure to one sentence so it is the first thing callers hear.
If a caller asks outright#
"Am I talking to a real person?"
The assistant answers truthfully: it is an AI assistant for your company. It is instructed never to claim otherwise, never to pretend, and never to deflect. If that answer changes how the caller feels, they can ask for a person, and that always works.
What we do not do#
| We do not | Why |
|---|---|
| Use a cloned or custom voice | Only stock voices from the voice company are offered. The assistant should not sound like a specific real person. |
| Call people | Inbound only. The assistant never dials a prospect. |
| Hide the assistant in the middle of the call | The disclosure is first, before the reason for the call is even asked. |
| Let a community turn the disclosure off | It is locked out of the box, and the publish check applies either way. |
Why it is built this way#
Being honest about it works better. Families who know they are talking to an assistant ask clearer questions, and the ones who would rather talk to a person say so straight away instead of ten minutes in. It also keeps you on the right side of a fast-moving set of rules about disclosing AI in conversations with consumers.
Info
This is how the product behaves, not legal advice. Disclosure requirements vary by state and change often. Have your counsel confirm what applies to your organisation.
Checking it yourself#
- Read the Callers hear preview at the bottom of How every call opens.
- Open any community's What callers hear tab to see that community's assembled greeting.
- Call the draft, or your test number, and listen.
- Open any past call and read the first line of the transcript.