The problem
A website drifts. A form stops sending, a page gets slow, a link breaks, a certificate lapses, and nobody notices until a client mentions it. Separately, a customer might ask an AI assistant who to call. Unless someone checks, you do not know whether you come up, or how you are described.
What it does
Every week, the site is checked the way a careful person would check it: pages load, forms send, links resolve, the important pages are still indexed, nothing has slowed down. Every month, a set of questions a customer might ask an AI assistant is put to the main assistants and the answers are scored for whether you appear, and how you are described. A short list of competitors is watched the same way. A summary lands on Monday, and anything broken is flagged the day it breaks.
What it needs from you
The questions a customer would actually ask, in their words, because a score against the wrong questions is worth nothing. The competitors you care about, by name. Who should get the alert when something breaks, who reads the Monday summary, and what you would want to know first if the site went down on a Friday night.
What it does not do
It does not fix what it finds, and it does not change the website. It does not promise a ranking or a mention. The scores are a measure taken the same way each month, nothing more, and a person decides what to do about them.