Why not me?
For each question you lost, this screen names what happened to your pages: retrieved but not cited, read and then passed over, or never retrieved at all. Each needs a different fix, and knowing which one saves you rewriting the wrong pages.
The diagnoses
| Finding | What it means |
|---|---|
| Retrieved, not cited | The model fetched your page and chose another source. The page did not answer the question well enough. |
| Read, then passed over | Your site was read while the answer was being written, and the answer still cited someone else. This needs the Cloudflare or Fastly connector on your site, and is matched by day. |
| Never retrieved | You were not in the running at all. With semantic scoring on, this splits further: a content gap (no page of yours matches the search) or a visibility gap (a matching page exists but was not found). |
Semantic scoring runs on Visibian’s side, so there is nothing for you to set up. The screen also shows the searches the model ran and which page was cited instead, so you can read the winner and see what it did better. The Retrievability Audit goes deeper on each state.
Where the answers come from
The Source Domains and Source Pages screens show the domains and exact pages the models treat as authoritative in your market: review sites, trade press, forums, competitors. If the same 3 domains keep appearing, those are the places worth getting yourself onto.
Ground Truth
Everything else measures what AI says. Ground Truth measures what it sends you: real people arriving on your website from an AI answer. Connect the analytics you already use, and visibility stops being a proxy. “We appear in 40% of answers” becomes “AI sent us 300 visitors last month”.
Next: Experiments →Privacy policy · Terms of service · Legal notice · Cookie choices