How to See What Your Website Tells AI About Your Brand

AI systems piece together their understanding of a brand from website content, third-party sources, and knowledge graph entries. The brand's own website should be the authoritative source — but it's difficult to know whether the site clearly expresses the right entities, claims, and relationships, or whether it leaves gaps that AI fills from elsewhere. Inspecting the structured knowledge extractable from a site reveals whether the content matches the brand's intent.

The Knowledge Graph in QueryBurst Site Intelligence displays every entity extracted from the site with statement counts, reinforcement rates, focus ratings, prominence scores, linking gaps, and peer ratios — showing exactly what the site contributes to AI's understanding and where it falls short.

How to Access

  1. Open a site in QueryBurst
  2. Click Site Intelligence in the sidebar
  3. The Knowledge Graph subtab is the default view

If no analysis has been run yet, click Run analysis to start the pipeline. Processing time depends on site size. See the Site Intelligence overview for details on the extraction pipeline.

Understanding the Interface

Summary Cards

Three stat cards appear at the top:

MetricDescription
Entity profilesTotal unique entities extracted across all pages
StatementsTotal subject–predicate–object triples extracted
ReinforcedEntities mentioned on more than one page (higher = stronger signal)

Quick Actions

  • Watch AI read your site — A step-by-step replay showing how the LLM builds its understanding page by page.
  • Explore 3D Knowledge Graph — An interactive 3D visualisation of all entities and their relationships.

Collapsible Panels

Three panels sit below the summary cards. Each can be collapsed to a pill to save space:

  • Entity Universe — A PCA scatter plot showing how entities cluster by semantic similarity. Entities of the same type share colours.
  • Dedup Groups — Automatically detected groups of entities that may refer to the same thing (e.g., "Google" and "Google LLC"). Shows severity and recommendations.
  • Top Statements — The most reinforced statements across the entire site, ranked by how many pages support each one.

Entity Table

The main table lists every entity profile. Each row shows:

ColumnDescription
EntityThe canonical name of the entity
AliasesOther names or spelling variants found for this entity
TypeEntity classification (Organization, Person, Product, Concept, etc.)
MergedBadge shown if this profile was created by merging duplicates — click to manage
StmtsNumber of unique statements extracted about this entity
Reinf.Number of statements that appear on multiple pages
PagesVisual bar showing what percentage of total pages mention this entity
FocusConcentration rating: Strong, Moderate, Contested, Diluted, or Absent
ScoreOverall entity prominence score (0–100%)
GapsNumber of missing internal links between pages that discuss this entity
PeersRatio of actual vs possible peer links between co-occurring entity pages

Click any row to open the full entity profile detail view.

Search & Filtering

Type in the search box to find entities by name. An autocomplete dropdown suggests matching entities as you type — click a suggestion to go directly to its profile, or press Enter to filter the table.

Active search terms appear as a removable pill below the toolbar.

Type Filter

Use the dropdown to filter by entity type (Organization, Person, Product, etc.). The count next to each type shows how many entities of that type exist.

Merged Filter

When merged entities exist, a Merged toggle button appears. Click it to show only profiles that were created by merging duplicates.

Merging Entities

When the same real-world entity appears under different names, you can merge them:

  1. Click Merge in the toolbar to enter merge mode
  2. Select two or more entity profiles by clicking their rows
  3. Click Merge selected and choose the canonical name
  4. The merged profile combines all statements and pages from the source profiles

To undo a merge, click the merged badge on any merged profile and select the entities to unmerge.

Subtabs

The Knowledge Graph tab contains seven subtabs accessible from the top navigation:

SubtabDescription
Knowledge GraphEntity profiles list (this view)
Graph ExplorerInteractive knowledge graph traversal with BFS and PageRank
Topics & FocusTopic clusters and topical focus analysis
Facts & ClaimsExtracted claims with filtering and semantic search
QuestionsQuestion-style headings found across your pages
Comparison PagesBrand and product comparison detection
ArchitectureURL path treemap with drill-down navigation

Interpreting Results

Healthy Signs

  • High reinforcement — Key entities appear across many pages, reinforcing their importance
  • Strong focus — Primary entities have concentrated, consistent coverage
  • Few gaps — Pages discussing the same entity link to each other
  • Clean dedup — Few or no duplicate entity names detected

Warning Signs

  • Diluted or contested focus — Multiple pages compete to define the same entity differently
  • High gap count — Pages about the same entity don't link to each other, weakening topical signals
  • Low peer ratio — Related pages are isolated from each other
  • Many dedup groups — Inconsistent entity naming may confuse AI models

Tips

  1. Fix high-gap entities first — Add internal links between pages that discuss the same entity
  2. Review dedup groups — Merge genuine duplicates and note any inconsistent naming to fix in content
  3. Check the Focus column — Entities rated "Contested" or "Diluted" may need clearer, more consistent definitions
  4. Use the 3D graph — It's the fastest way to spot isolated clusters or entities with weak connections
  5. Watch the replay — Understanding how the LLM reads your site page-by-page reveals sequencing issues

Technical Details

How Entities Are Extracted

The pipeline uses a two-stage process:

  1. Stage 1 — A lightweight pass over each page's title and meta description to identify the primary entity
  2. Stage 2 — A full extraction pass over page content that uses the primary entity as context to extract all entities, relationships (triples), summaries, and target queries

Entities are automatically deduplicated using embedding similarity and automated review. Only pages with changed content are re-processed on subsequent runs.

Enrichment Scores

The Focus, Score, Gaps, and Peers columns are computed by the entity enrichment service, which analyses each entity's coverage depth, internal linking patterns, and cross-page consistency. These load progressively after the main table appears.

  • Entity Profile — Deep dive into a single entity's statements, pages, and linking gaps
  • Graph Explorer — Walk the knowledge graph interactively
  • Flow Overview — Entity prominence and interlinking health across the site
  • Facts & Claims — All extracted claims with verification status