Skip to content

Article 10 min read

Google's Knowledge Graph: What It Is, Where It Gets Its Facts, and How to Check

by Daniel Sócrates published

Someone at your agency said your company needs to get into Google’s Knowledge Graph. You wrote the phrase down and looked it up that night.

We ran that search on September 14, 2026, on Google in Brazil, where we are based. The results were IBM, SAP, Neo4j, and artificial intelligence tutorials. All of them explain how to build a graph inside a company. None of them talks about your company inside Google. 1

The search got nothing wrong. The term is bigger than Google, and the first page covers the general sense. This page draws a line and stays inside it. Here, the knowledge graph is the one Google keeps about the world, where your company may or may not be.

I will go through five questions. What this graph is, where it gets its facts, and how to check whether an entity is in it. Then, what it is not, and what is in your hands.

A corkboard and red string

Remember the evidence board in crime movies? Photos held up with pushpins, and red string running from one to another. The suspect connects to the witness, who connects to the car, which connects to the address.

Each photo is a thing. Each string is a relationship between two things, with a name written on it. Lives in. Works for. Founded. This is called a knowledge graph.

Any company can build its own, and that is what the IBM and SAP results are about. Google built one about the whole world. It gave it a name on May 16, 2012: the Knowledge Graph. 2

The announcement, signed by Amit Singhal, explains the word with a wink. In the post’s own geek-speak, a model that understands real-world things and how they relate is a “graph.” 2 The title of the post carried the whole idea: things, not strings.

What that shift changed in search is in the piece on the difference between traditional and semantic SEO. How a name becomes a registered thing, with the launch numbers, is in how Google learns your company is a thing.

One detail of spelling helps with everything that follows. With capital letters, the Knowledge Graph is Google’s. In lowercase, a knowledge graph can belong to anyone.

Where the graph gets its facts

Google has described the graph’s sources twice, eight years apart.

In 2012, the announcement cited public sources such as Freebase, Wikipedia, and the CIA World Factbook. It also said the graph was tuned by what people search for and what Google finds on the web. 2

On May 20, 2020, Danny Sullivan of Google came back to the subject. The graph draws on material shared across the web and on open and licensed databases. 5 The panels for music, sports, and TV use licensed data. The health panels use content vetted with medical providers. 5

One source on that list concerns you directly. The post mentions special coding that content owners can use, for example to flag an upcoming event. 5 That is the markup you write on your own page.

The knowledge panel help fills in the picture. On some topics, Google works with data partners. And some of what appears may come from verified entities that suggested edits to their own facts. 4

What the Freebase shutdown notice teaches

Freebase came to Google with the purchase of Metaweb, in 2010, and became the foundation of the graph. 6 The purchase and the shutdown dates are in the piece on entities linked above. What matters here is the text of the notice, which few people read.

On December 16, 2014, the Knowledge Graph team published the notice. Google would help move the data in Freebase to Wikidata, the open database run by the Wikimedia community. 7 And the service would wind down in mid-2015 as a standalone project. 7

Two sentences in the notice say more than the dates. The first promises, before the shutdown, a new API for entity search, powered by the Knowledge Graph. 7 That is the window you will use in a moment.

The second explains why the data could not simply be poured into Wikidata. Many of Freebase’s facts had come from Wikipedia. And they were attributed to Wikipedia, with no original source behind them. 7 Loading them as they were would not meet the Wikidata community’s rules on citation and sourcing. 7 The plan was a tool to match each fact with a possible citation. 7

Read that again with your company in mind. Not even Google could load a fact into an open database without showing where it came from.

Today, the Freebase page on Google’s developer site sums up the story. It was a data-sharing project that ran from 2007 to 2015, and its API has been shut down. 8

How to check whether an entity is in the graph

Two customers named John Smith walk into the same bank branch on the same day. The teller never mixes them up. Each one is pulled up by Social Security number, and the name is just what shows on the screen.

The graph does the same with every thing it records. Each one gets a unique code, generated and used by the graph itself. 9 This is called a machine identifier, or MID.

The official window: the Knowledge Graph Search API

Google keeps an API for finding entities in the graph, the Knowledge Graph Search API. 10 You send a name, and it returns the entities that match. Each one comes with a name, a description, an official website, an @id, and a resultScore. 11

In the documentation’s own example, a search for Taylor Swift returns the identifier kg:/m/0dl567. 10

Every request needs an API key, created in the Google Cloud console. 12 It is a job for a developer, even though the whole query fits on a single line. 10

The documentation carries three warnings, and they matter. The API is read-only. It is not suitable as a production-critical service. And it is being migrated to Cloud Enterprise Knowledge Graph. 10

A fourth warning changes what you should expect to see. The API returns entities one by one, without the web of relationships between them. For anyone who needs that web, the documentation itself points to data dumps from Wikidata. 10

The /m/ and /g/ formats

Identifiers come in two formats. The Google Cloud documentation mentions /m and /g MIDs, and says both still work for lookups. 9 The /m/ format comes from Freebase, whose page shows identifiers like /m/012rkqx. 8

I found no Google document that says when the /g/ format began. Wikidata keeps the two in separate fields. Property P646 holds the Freebase ID. Property P2671 holds the Knowledge Graph ID that starts with /g/. 13

Those fields work as a bridge for checking. The Wikidata item for the Taj Mahal records the identifier /m/0l8cb. 13

The address with kgmid

There is a shortcut with no Google document behind it, and that is why it comes flagged. On September 14, 2026, I opened google.com/search?kgmid=%2Fm%2F0l8cb in the browser. That is /m/0l8cb with the slashes encoded. Google opened the search for the Taj Mahal, the monument. 14

I looked for the kgmid parameter in the official documentation and found no explanation. Treat it as my own observation, which can change without notice. 14

How to read what comes back

If the API returns an entity, look at three fields. Is the name spelled the way you spell it? Does the description say what the company does? Does the website field point to your domain? 11

The resultScore calls for care. The documentation defines it as an indicator of how well the entity matched the request constraints. 11 It describes the match with the query, and says nothing about the strength of your brand.

If nothing comes back, you have an empty result on a specific day. My book SEO de Entidade, published in Portuguese, says to write it down with the date and treat it as a clue. 6 If something else with a similar name comes back, you found a namesake, and that is information too. 6

Four mix-ups about the graph

Panel and graph. Google’s help defines knowledge panels as information boxes that appear when you search for entities that are in the Knowledge Graph. 3 4 The panel is one of the graph’s outputs. On mobile, the 2020 post says, information from the graph often shows up spread across the results. 5 The piece on the knowledge panel uses the image of a museum display case, which shows what the collection has catalogued.

A guaranteed panel. Knowledge panels are generated automatically. 4 In the Google documents read for this page, I found no rule that promises a panel to every entity in the graph. 5 4 Anyone who promises a panel by a set date is promising a decision that belongs to the system.

Signing up. There is no form for adding an entity. The reason is in the piece on entities, along with the correction channel for panels already live.

Wikidata. Wikidata is the open database that received Freebase’s data, with Google’s help. 7 The graph belongs to Google, and Google itself says it draws from hundreds of sources. 5

What is in your hands

You do not edit the graph. What is in your hands is the material it finds about you, on three fronts.

Official home. One page, at an address that does not change, saying who the company is. It is the entity home, and everything that comes after it points to it.

The sameAs declaration. Google’s documentation on organization structured data describes this field. It holds the URL of a page on another site with more information about the company, such as a social media profile. 15 The same page says the markup helps Google tell your organization apart from others in search results. 15 The book gives the practical rule. The domain that receives outside corroboration is the one that declares sameAs for everything else. 16

Identity corroboration. Sources you do not control stating the same facts about you, in the same words. 6 An industry association, a professional licensing board, a news outlet that does its own reporting. You do not write those sources. You make sure the fact that reaches them matches your official home.

The Wikidata item comes last, and the order is in the piece on Wikidata and notability. The full map of the work is in the Entity SEO section. None of these fronts has an arrival date.

The lookup to ask for this week

From now on, when someone says your company needs to get into the graph, answer with a question. Under which identifier is it there today?

Ask whoever runs your site for an API lookup with the company’s exact name. On a sheet of paper, write down the @id, the description, and the website that come back, with the date. If it comes back empty, write down the empty result.

Then open your entity home next to the sheet. Every difference between the two is work on sources, and the first line on that list is your next step.

Frequently asked questions

Is the Knowledge Graph the same thing as a knowledge graph?

It is one knowledge graph among many, and Google named its own on May 16, 2012. 2 Any company can build one. That general sense filled the first page of the search we ran on Google in Brazil. 1

Does being in the graph guarantee a knowledge panel?

The official documents read for this page make no such promise. The help describes panels as generated automatically for entities in the graph. 4

Does the resultScore measure the strength of my brand?

No. The documentation defines the field as an indicator of how well the entity matched the request constraints. 11 It also warns that the API is not suitable as a production-critical service. 10

Is Google’s graph the same as Wikidata?

No. Wikidata is the Wikimedia community’s open database, the one that received Freebase’s data. 7 The graph belongs to Google, and it reads hundreds of sources. 5

Do I need a developer to check?

For the API, yes, because every request needs an API key. 12 Searching the exact name in a private window shows whether there is a panel. A missing panel does not prove the entity is missing from the graph, and that reading is mine, based on the help’s definition. 4

Notes and sources

  1. Search for “grafo de conhecimento” (knowledge graph, in Portuguese) on Google Brazil, in the author’s own browser tab, https://www.google.com.br/search?q=grafo+de+conhecimento&hl=pt-BR&gl=br&pws=0. The search ran on Google in Brazil, where we are based, with the interface in Portuguese. Accessed September 14, 2026. First page: IBM, Neo4j, SAP, English Wikipedia (entry “Knowledge graph”), the Alan Turing Institute, Graphwise, the repository of UFSC (a Brazilian federal university), Splunk, and DataCamp, plus an ad from Neo4j. No result about Google’s graph. The Semrush snapshot (Brazil database, same day) has the same profile: the first address about Google’s graph appears in 19th organic position.
  2. Amit Singhal, “Introducing the Knowledge Graph: things, not strings,” official Google blog, May 16, 2012, https://blog.google/products-and-platforms/products/search/introducing-knowledge-graph-things-not/. Accessed September 14, 2026. Byline: Amit Singhal, with the title “SVP, Engineering”. Passages: the model described, “in geek-speak, a ‘graph’”, which “understands real-world entities and their relationships to one another: things, not strings”; “Google’s Knowledge Graph isn’t just rooted in public sources such as Freebase, Wikipedia and the CIA World Factbook”; and “it’s tuned based on what people search for, and what we find out on the web”.
  3. Google, Knowledge Panel Help, Portuguese version, “Sobre ‘Painéis de informações’” (About knowledge panels), https://support.google.com/knowledgepanel/answer/9163198?hl=pt-BR, page with no date. Accessed September 14, 2026. The page gives the same definition as the English original in note 4: the panels “são caixas informativas que aparecem no Google quando você pesquisa entidades (pessoas, lugares, organizações, coisas) que estão no ‘Mapa de informações’” (are information boxes that appear on Google when you search for entities, people, places, organizations, things, that are in the “information map”). It carries the warning “Esta página pode ter conteúdo que foi traduzido com tecnologia de IA” (This page may contain content translated with AI technology). That machine translation renders the Knowledge Graph as “Mapa de informações” and knowledge panels as “Painéis de informações.” This piece quotes the English original.
  4. Google, Knowledge Panel Help, “About knowledge panels,” https://support.google.com/knowledgepanel/answer/9163198?hl=en, page with no date. Accessed September 14, 2026. Passages: “Knowledge panels are information boxes that appear on Google when you search for entities (people, places, organizations, things) that are in the Knowledge Graph”; “Knowledge panels are automatically generated”; “In some cases, we may work with data partners who provide authoritative data on specific topics like movies or music”; and “some of the information displayed may also come from verified entities who have suggested edits to facts on their own knowledge panels”. The page contains no rule that guarantees a panel to every entity in the graph. The reading that a missing panel does not prove absence from the graph is the author’s.
  5. Danny Sullivan, “A reintroduction to our Knowledge Graph and knowledge panels,” official Google blog, May 20, 2020, https://blog.google/products-and-platforms/products/search/about-knowledge-graph-and-knowledge-panels/. Accessed September 14, 2026. Byline: Danny Sullivan, with the title “Public Liaison for Search”. Passages: the graph understands facts about entities from “materials shared across the web, as well as from open source and licensed databases”; “We draw from hundreds of sources from across the web, including licensing data that appears in knowledge panels for music, sports and TV”; “We work with medical providers to create carefully vetted content for knowledge panels for health issues”; “We also draw from special coding that content owners can use, such as to indicate upcoming events”; and, on mobile, “information from the Knowledge Graph is often not presented through a single knowledge panel on mobile”. The post has no display rule for every entity in the graph.
  6. Daniel Sócrates, SEO de Entidade (published in Portuguese), 2026, chapter 4, “O cartório do Google: como strings viram things” (The Google Registry: How Strings Become Things): the purchase of Metaweb, owner of Freebase, in July 2010 (section “A herança que foi doada,” The Inheritance That Was Given Away); the section “A janela que ficou aberta” (The Window Left Open), where a lookup returns nothing, or something else with a similar name, and the empty result is written down with the date and treated as a clue; and the definition of identity corroboration, independent sources stating the same facts about the same thing.
  7. The Knowledge Graph team at Google, Freebase shutdown announcement, Freebase page on Google+, December 16, 2014, Internet Archive copy captured July 29, 2017, https://web.archive.org/web/20170729170416/https://plus.google.com/109936836907132434202/posts/bu3z2wVqcQc?cfem=1. Accessed September 14, 2026. This is the address that Google’s own Freebase page gives as the announcement (note 8). Passages: “we’ve decided to help transfer the data in Freebase to Wikidata, and in mid-2015 we’ll wind down the Freebase service as a standalone project”; “before we retire it, we’ll launch a new API for entity search powered by Google’s Knowledge Graph”; “Loading Freebase into Wikidata as-is wouldn’t meet the Wikidata community’s guidelines for citation and sourcing of facts”; “those facts were attributed to Wikipedia and not the actual original non-Wikipedia sources”; and “we’ll be launching a tool for Wikidata community members to match Freebase assertions to potential citations”. The notice set March 31, 2015 for the database to become read-only and June 30, 2015 for retiring the site and the APIs.
  8. Google for Developers, “Data Dumps,” Freebase API (Deprecated), https://developers.google.com/freebase, page last updated November 20, 2023. Accessed September 14, 2026. Passages: “The Freebase API has been shut down”; “Freebase was a data-sharing project that ran from 2007 to 2015”; and “In Freebase, objects have MIDs that look like /m/012rkqx”.
  9. Google Cloud, “Understand Machine ID (MID),” Enterprise Knowledge Graph, https://docs.cloud.google.com/enterprise-knowledge-graph/docs/mid, page last updated September 3, 2026. Accessed September 14, 2026. Passages: “A machine ID, or MID, is a globally unique identifier generated and used by a knowledge graph”; and “If you are already using /m or /g MIDs, you can continue to look up the entities”. The page does not date the /g/ format.
  10. Google for Developers, “Google Knowledge Graph Search API,” https://developers.google.com/knowledge-graph, page last updated April 26, 2024. Accessed September 14, 2026. Passages: “The Knowledge Graph Search API lets you find entities in the Google Knowledge Graph”; “we are migrating this API to Cloud Enterprise Knowledge Graph”; “The Knowledge Graph Search API is a read-only API”; “This API is not suitable for use as a production-critical service”; and “The Knowledge Graph Search API returns only individual matching entities, rather than graphs of interconnected entities. If you need the latter, we recommend using data dumps from Wikidata instead”. The page’s example, a one-line query to kgsearch.googleapis.com for the name Taylor Swift, returns "@id": "kg:/m/0dl567".
  11. Google for Developers, “Method entities.search,” Knowledge Graph Search API, https://developers.google.com/knowledge-graph/reference/rest/v1, page last updated June 26, 2024. Accessed September 14, 2026. Field definitions: @id, “The canonical URI for the entity”; url, “The official website URL of the entity, if available”; and resultScore, “An indicator of how well the entity matched the request constraints”. The three-field reading guide is the author’s.
  12. Google for Developers, “Authorize Requests,” Knowledge Graph Search API, https://developers.google.com/knowledge-graph/how-tos/authorizing, page last updated August 28, 2025. Accessed September 14, 2026. Passage: “Your application needs to identify itself every time it sends a request to the Google Knowledge Graph Search API, by including an API key with each request”. The key is created on the credentials page of the console.
  13. Wikidata, item Q9141 (Taj Mahal), https://www.wikidata.org/wiki/Q9141, and property P2671, https://www.wikidata.org/wiki/Property:P2671, read through the public Wikidata API. Accessed September 14, 2026. The item records the value /m/0l8cb in P646 (Freebase ID). Description of P2671 in English: “identifier for Google Knowledge Graph API, starting with “/g/”. For IDs starting with “/m/”, use Freebase ID (P646)”. Wikidata is edited by volunteers and is not a Google document.
  14. Author’s observation, address https://www.google.com/search?kgmid=%2Fm%2F0l8cb&hl=pt-BR&gl=br, opened in the author’s own tab, from Brazil, with the interface in Portuguese. Accessed September 14, 2026. The tab opened with the title Taj Mahal, followed by “Pesquisa Google” (Google Search), and results about the monument (Portuguese Wikipedia, UNESCO, and the monument’s official site). A web search for kgmid, restricted to google.com, found no Google page that explains the parameter; only the MID pages in notes 9 and 10 came back.
  15. Google Search Central, “Organization (Organization) structured data,” https://developers.google.com/search/docs/appearance/structured-data/organization, page last updated September 8, 2026. Accessed September 14, 2026. Passages: the markup “can help Google better understand your organization’s administrative details and disambiguate your organization in search results”; and, in the definition of sameAs, “The URL of a page on another website with additional information about your organization, if applicable. For example, a URL to your organization’s profile page on a social media or review site”.
  16. Daniel Sócrates, SEO de Entidade (published in Portuguese), 2026, chapter 12, “Dê à sua entidade uma casa oficial” (Give Your Entity an Official Home): the domain that receives outside corroboration is the same one that declares sameAs for everything else, and the declaration serves to disambiguate, never as a shortcut to ranking.

Want to see SEO Genome more often on Google? Add it as a preferred source.

Add to Preferred Sources

Daniel Sócrates

Founder and CEO of SEO Genome

Author of AI Search Optimization. Writes about what the company applies in client work.

danielsocrates.com.br

Silo Entity SEO

Keep reading

PreviousThe Difference Between Traditional and Semantic SEO: Three Google Shifts, With Dates and Documents Parent pageEntity SEO: your company exists in the public registry. Does it exist in Google's catalog?

Let us look at what the machine understands about your company

A diagnostic conversation, with no slide deck. You bring your website address. We bring what Google and the AI assistants already know about it today.

Talk to us on WhatsApp