Google's Knowledge Graph: what it is, how to get into it and why it matters for AI

The Knowledge Graph is the database where Google keeps entities, meaning people, companies, places, products and concepts, along with the facts about them and the relationships between them. It's what lets Google work out that "Ferrari" in one search is the car maker and in another is someone's surname.

Google launched it on 16 May 2012 with a post by Amit Singhal called "things, not strings". At launch it held more than 500 million objects and 3.5 billion facts. In 2020 Danny Sullivan wrote that it had grown to more than 500 billion facts about 5 billion entities.

For a business the useful question is a different one: does Google know I exist as an entity, or does it only know me as a collection of pages?

Knowledge Panel and Business Profile: not the same thing

The most visible place the Knowledge Graph shows up is the box to the right of the results. But that box has two different origins, and mixing them up means working on the wrong thing.

• Entity Knowledge Panel: comes from the Knowledge Graph. It shows a description, founders, founding date, social profiles and sometimes an extract from Wikipedia. You can claim it, but you can't create it directly.
• Local business listing: comes from the Google Business Profile. It shows the address, opening hours, phone number and reviews. The business creates and manages it.

A real example from one of our projects in Italy. Searching for "macropix" on Google Italy on 27 September 2026, the panel on the right shows Macropix: its address in San Donato Milanese, phone number, opening hours, 19 reviews averaging 4.8, products and a carousel of profiles with YouTube, Instagram, X and LinkedIn. The SERP data shows that the panel comes from the Business Profile: the attributes are of the "local" type. The profile carousel, on the other hand, is a Knowledge Graph attribute ("social media presence"). It's a typical picture for a small or mid-sized business: a strong local identity, with the first signs of an entity.

For local businesses Google itself points to the Business Profile. For a company that sells across the UK or abroad, being a recognised entity matters more.

What entities are, in practice

An entity has a unique identifier. In Google's Knowledge Graph these are strings such as /m/0k8z (a format inherited from Freebase) or /g/11.... Two pages that talk about "Visilay" in different ways are traced back to the same entity if Google finds enough consistent signals: the same name, the same website, the same profiles, the same facts.

According to Google, the Knowledge Graph's sources are Wikipedia, "hundreds of sources across the web", licensed databases for areas such as music and sport, and the Business Profile. Until 2016 there was also Freebase, the database Google bought in 2010 and shut down on 2 May 2016, with its data migrated to Wikidata.

Opinions differ on how much weight Wikidata carries today. Ahrefs argues that a significant part of the Knowledge Graph comes from there and that, for a company, a Wikidata entry is more achievable than a Wikipedia page. That's Ahrefs' assessment, not a statement from Google, but it matches what we see in our projects.

Why it matters for AI

Language models reason in terms of entities even when they don't use Google's Knowledge Graph. If a company is described consistently across many sources (its website, Wikidata, LinkedIn, Companies House, trade publications), the model "knows" it better and is less likely to confuse it with others of the same name.

One point calls for caution. The only case where Google has said it uses a knowledge graph to ground a model's answers is DataGemma, which uses Data Commons, a statistical database with more than 240 billion data points. The claim that AI Overviews and AI Mode use the Knowledge Graph to verify entities comes from third-party analysis, not from Google.

What tests show is simpler: companies with a clear identity, repeated across external sources, get named more often in answers. An Ahrefs study of 75,000 brands found that brand mentions on the web are the factor most closely correlated with appearing in AI Overviews. We cover this in AI citations and mentions.

How to check whether your company is an entity

Three checks, from the quickest to the most precise:

1. Search for the company name on Google. If a panel with a description and social profiles appears on the right, you're probably an entity. If you only see an address and opening hours, that's the local listing.
2. Search for the name with a qualifier, for example "Visilay agency" or "Macropix LED wall". If the panel stays the same, Google is connecting the two.
3. Query the Knowledge Graph Search API. The endpoint is https://kgsearch.googleapis.com/v1/entities:search and you need a Google Cloud API key, which costs nothing for this API. It returns up to 10 entities with name, type, description and a relevance resultScore. Google describes it as "not suitable for use as a production-critical service", but for a manual check it's perfectly fine.

https://kgsearch.googleapis.com/v1/entities:search?query=COMPANY+NAME&key=YOUR_KEY&limit=5&languages=en

If the response includes your company with the type Organization or Corporation, you're an entity. The resultScore isn't a quality score: it shows how closely the entity matches the search.

How to get into the Knowledge Graph

There's no form to request inclusion. You work on the signals Google uses to recognise an entity.

1. Organization structured data on your website. Official name, logo, URL and, above all, sameAs, which according to Google's documentation is "the URL of a page on another website with additional information about your organization". Put your LinkedIn page, official social profiles and Wikidata entry there. The logo must be at least 112x112 pixels. The general guide is in schema markup.
2. A Wikidata entry. Wikidata accepts companies that have independent sources: press articles, public registers such as Companies House. Without sources the entry gets deleted. Enter the same details as on your site: name, official website, founding date, head office, profiles.
3. Consistency everywhere. The same name, the same description and the same details on LinkedIn, your Business Profile, Companies House and trade directories. Small differences ("Ltd" versus "Limited", an old address) cause confusion.
4. Mentions in independent sources. Articles, interviews and case studies published by others. It's the slowest signal to build and the one that counts most.
5. A complete "about us" page, with your history, your people and verifiable facts. It's also an E-E-A-T signal.

Claiming the Knowledge Panel

If the panel exists, it's worth claiming. The process is described in Google's guide:

1. search for the entity on Google;
2. click "Claim this knowledge panel" (it may sit under the "About" section);
3. verify yourself by signing in to a linked official account: YouTube, Search Console, X (still listed by Google as Twitter) or Facebook.

Once verified you can suggest changes. Not every panel can be claimed, and Google still reviews any changes.

How long it takes

It depends on external sources. A company with coverage in national newspapers and a well-built Wikidata entry can appear as an entity within a few weeks. A small business with no external mentions can wait months, even with perfect structured data. The work on your website is necessary; the work off it is what sets the timescale.

Further reading: SEO, AEO and GEO: the differences, the acronyms and what to actually do.

FAQs