How do I set up my questions?
- Enter your website on the Website step. Proofsource reads your site to find your category, the names you go by, your competitors and the topics buyers ask about.
- Review Check what we found. Correct the category, add names you go by, and remove or add competitors.
- Under Prompts we will ask, tick the questions you want, add your own, or swap any. Each prompt is one question put to every engine, grouped by topic.
- Under Questions that name your brand, keep or untick the questions buyers ask about you by name.
- Pick Answer language and the country under Asking AI as a buyer in, then click Start first scan.
If you change the category, the prompts built on it are rewritten, so the questions keep naming your market. More prompts you can add lists further questions read from your site; none is asked until you add it to a topic.
How are the default questions picked?
The default selection takes questions from each topic in turn, so one topic cannot crowd out the rest. Within that, a question backed by a measured demand signal ranks ahead of one without.
- Your own Search Console, once connected: non-branded queries real visitors typed into Google.
- Google's related questions with a monthly search volume.
- The same question asked in the open on forums, which proves it is asked but not how often, so it earns only a small weight.
A question with no signal is not treated as low demand. It keeps its place behind the measured ones, for you to keep or cut.
How many prompts and responses do I need for a reliable score?
On the trial you track up to 25 prompts, one cap that covers buyer questions and questions that name you alike; the default is 20 buyer questions and 5 that name you. On a paid plan the cap is the number of prompts per brand your plan includes.
Reliability comes from answers, not prompts alone. 20 buyer questions on 4 engines give 80 answers a scan. If half of them name you, the 95% Wilson interval runs from about 39% to 61%. Daily runs narrow that range as answers add up, and every rate in the app shows its interval. Below a minimum number of answers the first scan shows counts only, because a percentage would mean little.
Why are questions that name me counted apart?
A question with your name in it names you by construction, so it cannot show whether new buyers find you. By default 5 such questions are asked to see how AI describes you when someone asks by name. They count toward your prompt cap but are left out of the headline share of buyer answers that name you. The first scan needs at least one prompt that does not name you.
For a public example: for Tesla, a buyer question is Which electric SUV is best for a family road trip? A question that names the brand is Is the Tesla Model Y reliable? Only the first kind shows whether Tesla wins buyers who have not decided yet.
How do I change the questions later?
- Open AI Visibility, go to the Prompts tab and choose Review prompts.
- Edit a prompt's wording, or pause one to stop asking it. Resume a paused prompt to start asking it again.
- Use Add a prompt to type a question buyers ask and pick its topic, or accept one from Suggested prompts to add, ranked by opportunity.
Changes apply from the next daily run, within your plan's prompt cap. On the free trial, questions can be changed only before the first scan starts.
