Knowledge Graph

What the Knowledge Graph is

Google's Knowledge Graph is a giant structured database of entities — people, organizations, places, works, concepts — and the relationships between them, launched in May 2012 under the tagline "things, not strings." It matters to you because it's the layer search and AI features consult before your pages: panels, entity carousels, disambiguation, and increasingly the grounding behind AI Overviews all draw on it, so a wrong or missing record silently misroutes demand away from you.

Every record in the graph has a machine ID (the older /m/ Freebase-style ids and newer /g/ ids you'll see in the API and in Google Trends URLs), a set of attributes, and typed relationships to other entities. Your brand being "in the Knowledge Graph" literally means one of these records exists and points at your website as its official home.

A real lookup you can run right now

Google exposes a public read API. Request:

curl "https://kgsearch.googleapis.com/v1/entities:search?query=wikipedia&key=API_KEY&limit=1&indent=true"

Typical response shape:

{
  "@type": "EntitySearchResult",
  "result": {
    "@id": "kg:/m/0d07ph",
    "name": "Wikipedia",
    "@type": ["Thing", "Organization"],
    "url": "https://www.wikipedia.org/"
  },
  "resultScore": 8536.7
}

Three things to read off any response for your own brand: the machine ID (your permanent handle in the graph), the @type array (is Google classifying you correctly?), and resultScore (relative confidence — useful for tracking direction over time, meaningless as an absolute number).

Machine IDs: the handle you'll keep meeting

Once you know your entity's machine ID, you'll start seeing it everywhere. Google Trends encodes it in topic URLs, which is how "topic" searches track a thing across every language and misspelling while "search term" tracks one string. Some SEO tools accept kgmids directly for entity rank tracking. And when you file panel feedback or corroborate a Wikidata item, you're effectively lobbying the record behind that ID. Write yours down — it's more permanent than your domain. Freebase-era /m/ ids from 2012 still resolve today; entities that predate their current owners' websites carry their whole history in one identifier.

Where the graph gets its data — and how to influence each source

How to check your Knowledge Graph presence

1. Run the API query above with your brand name. No result at all means you're working on graph entry; a result with wrong types or a stale URL means you're working on graph correction — different playbooks.
2. Note the machine ID and re-query it monthly. Track resultScore direction alongside your corroboration work.
3. Search wikidata.org for your brand. Missing item? That's usually the highest-leverage single fix. Existing item? Audit every statement and reference.
4. Compare the graph's facts (via the panel or API) against your site's schema. Any disagreement — founding date, HQ, logo — resolve in favor of one canonical version everywhere.
5. Check Google Trends: if your brand offers a "topic" option (not just "search term"), the graph knows you.

Common mistakes

• Trying to schema-markup your way in with zero external corroboration. Your own JSON-LD is a claim, not proof. The graph is built on agreement between independent sources.
• Editing Wikipedia about yourself. Conflict-of-interest edits get reverted and can poison the well for legitimate future coverage.
• Letting an old Freebase-era record rot. Acquired companies and rebrands often have stale graph entries pointing at dead domains. File feedback on the panel and fix Wikidata; don't just launch new markup and hope.
• Confusing the Knowledge Graph with knowledge panels. The graph is the database; the panel is one rendering of it. API presence without a panel is common and still valuable.
• Inconsistent facts across your own properties. If your About page, LinkedIn, and GBP disagree on your founding year, you've given the graph three answers and it may pick none.

FAQ

Is the Knowledge Graph the same as schema markup?

No. Schema is a vocabulary you publish on your pages; the Knowledge Graph is Google's internal database. Your markup is one input among many — see the entity entry for the full signal stack.

Can I pay or apply to get into the Knowledge Graph?

No application exists. Entry is algorithmic, driven by corroborated evidence that you're a distinct, notable thing. Anyone selling guaranteed graph inclusion is selling weather.

Does Knowledge Graph presence help with AI search?

Yes, and increasingly so. Entity-grounded systems disambiguate brands against graph records, and our Knowledge Graph optimization for AI guide covers that angle specifically.

Why does my brand show a resultScore of 12 while competitors show 900?

Score reflects Google's confidence and prominence for that query string, driven mostly by external corroboration volume. Low score with a correct record is fine — it grows with coverage, not with markup tweaks.

What happened to Freebase?

Google acquired it in 2010, used it to seed the graph, then shut it down in 2016 and pointed the community to Wikidata — which is why Wikidata is now the most actionable public input you have.