For each question you monitor: the pages AI engines cite, whether those pages name you, your nearest page, and what to do next, with a map of all of it.
Last updated
A brand reaches an AI answer in two ways: one of its pages is cited, or the pages the engine reads name it. The Semantic Landscape shows both for every question you monitor: what the AI is reading, whether you are in it, and what to change.
Available on Ultra and Enterprise, in Brand Protection → Semantic Landscape.
Nothing on this tab updates by itself. It changes when someone on your team presses Recompute.
The table above the map has one row for each question you monitor, measured over the last 14 days of answers.
| Column | What it tells you |
|---|---|
| Question | The monitored prompt |
| You cited | The share of citations for this question that went to pages on your site |
| Named in sources | Of the cited pages Vizzybl could read, how many name your brand, and how much of the citations those pages account for |
| Your nearest page | Your page closest in meaning to the question, with a closeness score (higher is closer) |
| Verdict | One reading of the row (below) |
| Action | The one thing to do next |
Each row gets the first verdict that applies, in this order. The thresholds are first settings and may be tuned, so every row shows the numbers behind its verdict.
| Verdict | When | Action |
|---|---|---|
| No sources | Fewer than 20% of the answers cited any page: the engines answered from memory | None |
| You are a source | Your pages took 25% or more of the citations | Keep it fresh |
| Not enough read | Fewer than 10 of the cited pages could be read | None. Recompute to try again |
| Third-party territory | Other sites took 60% or more of the citations, and fewer than 30% of the pages read name you | Build outreach list |
| Close, not cited | Your nearest page is as close to the question as the cited pages, but is not being cited | Improve this page |
| Content gap | You have no page that close to the question | Plan a new page |
A page that could not be read counts as unknown. It never counts as a page that does not name you.
Viewers see the table without the action buttons.
Click a row. The map narrows to that question: its AI answers, the pages cited for it, and your five nearest pages, joined by lines that carry the real closeness. The right panel lists:
Show all returns to the whole map.
Each point is a passage of text. Points near each other cover similar topics.
Position is a 2D summary of meaning, good for seeing what sits together. For how close two things really are, use the closeness figures in the table and on the lines: they are measured in the full embedding space, not in the flattened picture.
Click a point for its details:
At each Recompute Vizzybl reads the most-cited pages for each question from the public web, as VizzyblBot, and obeys each site's robots.txt. The line under the title counts what happened, for example "643 pages cited by AI in the last 14 days: 476 read, 136 could not be read, 31 PDF or video."
The text of a cited page is kept for 7 days and then deleted.
Use the left panel to focus the map:
The map zooms to fit the points that are showing.
A map computed before the table was introduced shows the older Critical Gaps list and red regions instead, with Bridge This Gap drafting a brief for a gap. Press Recompute to get the table.
The map uses the same content embeddings as the rest of Vizzybl. Your own site must be indexed first, from the Content Index tab. Competitor sites you have indexed appear as orange points. AI answers come from your monitored prompts; answers to questions you no longer monitor are left off.
Press Compute map the first time, and Recompute to refresh. The new map replaces the old one for everyone on your team when it finishes.
