The Knowledge Graph
How a search engine (and now an answer engine) turns your website into an entity it can name, trust, and cite. The map of what the machines actually read: string to entity to connected graph, wired by @id and disambiguated by sameAs.
How to read this, and the one idea underneath it
To a machine, your business starts life as a string: a run of characters on a page, indistinguishable from any other. Structured data is how a string becomes an entity: one node, with a stable identity, a type, properties, and links to the other entities it relates to. Do it well and a search engine can resolve "you" to a single thing it can name, trust, and describe. Do it badly, or not at all, and you stay a pile of disconnected strings.
This is not decoration and it is not a ranking trick. It is the difference between being read and being understood. The left column of each block draws the mechanism; the right column is the why, with the spec it rests on. Where a tactic is a trap, there's a flagged refusal.
A string is not an entity
Until a machine can resolve "you" to one node with an identity, you're just words on a page.
A graph, not a pile
Isolated snippets don't compound. Entities linked by @id and sameAs do.
Valid is the floor; recommended is the game
schema.org validity gets you parsed. Google's required and recommended properties get you the result.
You don't get to rate yourself
Self-applied review stars on your Organization are ignored, and a manual-action risk. Honesty is the only durable schema strategy.
The map is bigger than the rich result
Most schema is understood, not rewarded. In an answer-engine world, being understood is the point.
The shape of the thing
Before any property or rich result, there's one move: turning the words on your page into a node a machine can hold onto. @id is that identity: a stable, resolvable handle every other node can point at.
#org
An @id is the difference between text and a thing.
Give the entity a stable, page-resolvable identity so every other node (the website, the people, the services) can reference the same thing instead of repeating it.
A knowledge graph is nodes and edges. Without identity there are no edges, and isolated snippets never compound into a graph.
The backbone
Three nodes carry every site: the entity, its website, and the page being viewed, cross-referenced by @id. And the single highest-leverage move for a local business: resolve to the most specific type you are.
Be the subtype you actually are.
The page references the website; the website references the entity. One graph per page, value objects nested inline (PostalAddress, GeoCoordinates, Offer).
Google rewards the most-specific LocalBusiness subtype (Plumber, Dentist, CafeOrCoffeeShop), not a generic one. LocalBusiness requires name address.
The people and the proof
A Person node, wired worksFor ⇄ founder, gives the business a human entity. And the single most common manual-action trap: putting review stars where Google won't honour them.
Stars belong on the product, never the org.
Ratings are legitimate on Product, SoftwareApplication and Course, things customers rate. They are not legitimate self-applied to your Organization or LocalBusiness.
Self-applied aggregateRating/review on your Organization or LocalBusiness. Google ignores the stars and fabrication risks a manual action. The Entity Map strips this and tells you why.
The links out
sameAs is the disambiguation lever: it tells the knowledge graph that your entity and these external profiles are the same thing. A Wikidata QID is the highest-leverage target there is.
Point at the authorities that already know who you are.
Each sameAs link collapses ambiguity: the engine maps your node to an entity it already trusts. Wikidata, Wikipedia, LinkedIn, and official profiles are the strongest targets.
Consistent NAP (name, address, phone) across the web reinforces the same entity. Inconsistency splits one entity into several weak ones.
What earns a result vs what's merely understood
schema.org validity gets you parsed. Whether you earn a visible rich result depends on Google's per-type required and recommended properties, and on what they've quietly removed.
Required is the gate; recommended is the win.
Product needs name image offers and at least one of offers/review/aggregateRating. Organization has no required properties; logo and sameAs are the levers.
FAQ rich results were removed 2026-05-07; the Sitelinks Search Box / WebSite SearchAction in 2024-11-21. Valid schema, but never a rich-result win. Don't prescribe them as one.
The AI layer
The honest position on whether structured data helps answer engines cite you. Some of this is supportable; some is unproven. This section is labelled method / informed opinion, not specification.
Get the fundamentals right; treat citation as upside.
Correct, connected schema makes your business unambiguous to the crawlers and knowledge graphs that search and AI answer engines draw on, and makes your facts machine-readable.
It does not guarantee an LLM will cite you; there's no published standard for how LLMs choose sources. Experiments to date are early, small-sample, and partly vendor-sourced. Treat AI citation as upside, not a guarantee.
The Entity Map, running live
Everything above, made interactive. Pick your archetype, answer a few questions, and it prescribes the connected JSON-LD @graph your site should have, validated against schema.org 30.0, graded against Google's requirements, and refusing the traps. Then read your current schema and diff it.
Drop a URL (this page only)
…or drag See my schema to your bookmarks bar, then click it on any page. It captures even JS/GTM-injected schema.
Or paste your page's HTML (View source → select all → copy)
Physical goods sold direct. Product + Offer, brand entity, reviews on products (not the org).
Two lines of code. It stays in sync, costs nothing, asks nothing.
Paste the URL of the page you'll embed the tool on. We'll read its colours and fonts and show you what the calculator would look like in your brand.
Extracted colours
Click an area below, then tap colours above to paint it. The area stays selected, so try a few quickly. The green (money kept) and red (leaks) are generated automatically so the chart always reads right.
Fine-tune surfaces (cards, fields, buttons, active tab)
Pick a colour from the strip above
How the collaboration badge will appear:
× <div data-cwc-tool="entity-map" data-cwc-title="The Entity Map by Christopher White Consulting" data-cwc-reserve-narrow="2600" data-cwc-reserve-wide="1700" style="min-height:1700px"></div>
<script src="https://christopherwhite.com.au/embed/loader.js" data-cwc-tool="entity-map" async></script>Method & sources
- schema.org: the vocabulary (version 30.0, 2026-03-19). Type hierarchy, properties, and the data model the Entity Map validates against. Licensed CC BY-SA 3.0.
- Google Search Central: per-type rich-result requirements (required vs recommended), the self-serving-review policy, and the Search Gallery deprecation notices (FAQ 2026-05-07; Sitelinks Search Box 2024-11-21).
- Wikidata / Google Knowledge Graph: entity reconciliation and the role of
sameAsand the Wikidata QID in disambiguation. - AI-citation evidence (method): current platform statements (Bing/Copilot) and the 2025–26 citation experiments, cited as informed method rather than settled fact. Schema makes you machine-readable and unambiguous; it does not guarantee citation.
Reads alongside The Signal Chain (measurement), The Conversion Journey (the page the entity describes), and The Acquisition Map (the territory).
White, C. (2026). The Knowledge Graph. Christopher White Consulting. https://christopherwhite.com.au/playbooks/the-knowledge-graph/