What Is a Knowledge Graph, and Does Your Business Have an Entry?

Someone told you that AI assistants trust businesses that are "in the knowledge graph". It sounded important, and nobody explained what it was, whether you are in it, or how you would ever find out.

The idea is simpler than the name. A knowledge graph is where a machine keeps what it has decided is true about a thing, including your business, after reading about it in many places.

Here is what a knowledge graph is in plain English, how to tell whether your business has an entry, what an entry is built from, whether it matters for AI answers, and what is worth doing about it.

What is a knowledge graph, in plain English?

A knowledge graph is a database of things and the facts that connect them. A business is a thing. So is its founder, its city and its website. The graph stores "this coaching practice is in Denver and was founded by this person" as a record, rather than as a sentence on a page.

Google introduced its own version in 2012 with a post called Introducing the Knowledge Graph: things, not strings. At launch it described more than 500 million objects and more than 3.5 billion facts about and relationships between them. "Things, not strings" is the whole idea: the search engine stops matching words and starts recognizing what the words refer to.

That matters for a small business for one reason. Two coaches with similar names are two different things, and a record is how a machine keeps them apart. The same problem from the other side, where two assistants describe you differently, is in why two AI assistants say opposite things about your business.

Does your business have a knowledge graph entry?

You cannot look it up in a public list, but you can look for the visible sign. Search your exact business name on Google. If a knowledge panel appears for your business, you have an entry. If nothing appears, you may have a thin entry, or none.

Google's help page About knowledge panels says panels appear when you search for people, places, organizations and things that are in the Knowledge Graph, that they are generated automatically, and that they are updated automatically as information changes on the web. It also says an official representative can get verified, claim a panel and suggest changes.

Developers can query the graph directly through the Knowledge Graph Search API. Google says it returns individual matching entities rather than graphs of connected ones, and that it is not suitable as a production-critical service. It is a curiosity for most business owners, not a tool to build a routine around.

What a knowledge graph entry is built from

An entry is built from the same facts appearing the same way in more than one place. Each fact is only as solid as the number of independent sources that agree on it, and a fact stated two different ways gives a machine two candidates instead of one.

ILLUSTRATIVE: ONE BUSINESS, SIX FACTSWHERE EACH FACT IS CONFIRMED

Fact Own siteBusiness ProfileLinkedInDirectory Record

1. Business name

   • Own site
   • Business Profile
   • LinkedIn
   • Directory

   Agreed in 4 places

2. What it does

   • Own site
   • Business Profile
   • LinkedIn
   • Directory

   Two versions

3. City

   • Own site
   • Business Profile
   • LinkedIn
   • Directory

   Agreed in 3 places

4. Founder

   • Own site
   • Business Profile
   • LinkedIn
   • Directory

   Agreed in 2 places

5. Website

   • Own site
   • Business Profile
   • LinkedIn
   • Directory

   Agreed in 3 places

6. Year founded

   • Own site
   • Business Profile
   • LinkedIn
   • Directory

   Two versions

States it the same way Says something different Says nothing

AnswerHalo answerhalo.com

CLICK TO ENLARGE

Read the illustrative record in the figure row by row. The business name is stated the same way on the business's own site, its Business Profile, LinkedIn and a directory, so it is the strongest fact in the record. What the business does is stated in three places and contradicted in one, because the Business Profile uses a different category. The city and the website are each confirmed in three places and the founder in two. The year founded appears only on the site, and LinkedIn gives a different year, which is worse than silence.

Two kinds of source do most of the work. The first is the business itself, through its site and the labels it adds. Google's organization structured data documentation says this markup can help Google understand an organization's details and disambiguate it in search results, and describes the sameAs property as the URL of a page on another website with more information about the organization, such as a social or review profile.

The second is everyone else: profiles, directories, press and open databases. Wikidata, the open database many knowledge graphs read from, accepts an item that refers to a clearly identifiable entity described using serious and publicly available references. Most small businesses fail that bar for the same reason they fail Wikipedia's, which is covered in does Wikipedia matter for AI visibility.

Does a knowledge graph matter for AI assistants?

Partly, and nobody outside these companies can say exactly how much. Search features are documented as using knowledge graphs. Chat assistants are documented as learning from text and fetching pages when they search, and none of them publishes how an entry in any graph changes who gets named for a buyer's question.

That uncertainty points to the same work either way. The facts that build a solid entry are the facts an assistant finds when it reads your pages or fetches them to answer a question. How the two halves of an answer work, the part learned in advance and the part fetched while you wait, is in what is retrieval-augmented generation, in plain English.

Be wary of anyone selling knowledge graph placement for AI answers specifically. There is no application form, and no assistant documents a way to buy or request an entry that changes its recommendations.

What to do about your knowledge graph entry this month

Make your basic facts agree everywhere you control them, add organization markup that points at those profiles, and stop there. An entry is a side effect of consistent facts in several places. It is not something you apply for.

Write the six facts down once. Name, what you do in one sentence, city or none, founder, website, and year founded. That note becomes the source every profile is corrected against.

Correct the profiles you already have. Start with the ones that carry the most facts. Which profiles are worth claiming at all, and in what order, is in which profiles a coach should claim first.

Add organization markup to your homepage. List only profiles that are really yours in sameAs. A link to a profile you abandoned says an out-of-date version of you is also you.

The damaging admission: most solo coaches and consultants will never get a knowledge panel, and doing all of the above does not change that on any timeline anyone can promise. It also does not need to. Plenty of businesses with no panel are named in AI answers, because an assistant can name a business it read about without keeping a formal record of it. Consistent facts help both ways. Chasing the panel itself is a poor use of a month.

To see whether you are named for your niche's questions today, start with the free check. It asks ChatGPT 25 real buyer questions from your niche at $0 and emails a plain-English readout: your score, who got named in your place, and the one fix to start with. It runs as soon as you ask for it, and the report lands within the hour between 7am and 5pm Central, or by 10am the next morning outside those hours.