Manage your content knowledge graph

Martha says: “Start to think about entity SEO and managing your content knowledge graph.”

How would you define an entity?

“I always like to think about an entity as a thing. When you’re in a business, you can think about the most important strategic ‘things’ for the business. What makes an entity different from a keyword (or a word in general) is that it is multi-dimensional. You can describe that entity and how it relates to other things.

The language that we use to articulate what an entity is, is Schema.org. Schema.org is a machine-readable vocabulary that is really robust, and it has all the properties to describe what an entity is about. You can think about it as a keyword being one-dimensional – just words – while an entity is multi-dimensional – it has names, sizes, colours, prices, connections to other entities, and more.”

Do SEOs sometimes think something’s an entity when it’s not?

“Probably. Entities are unique and distinguishable things or ideas. They have defining characteristics or attributes – that can be described using the Schema.org properties. They also exist in relation to other entities.

It’s not just a person, but a person with a name, a background, an education, and a job within an organisation. Entities have all these parameters that allow us to ask questions, make inferences, or have a deeper understanding.

I always ask organisations what their important entities are. They are the things they sell, their locations, and the key executives in the business. It could also be an asset, or it could be a blog that talks about how all these things are connected. However, it’s likely not a sentence or just a word.”

Why should SEO and content teams care about entities?

“Another way to think about that is, why have we started talking about entities and why is this something we’ve defined over time?

This is most important in 2025 because of large language models and AI. AI-driven search engines have been around for over a year, and entities are important in this new world. With these large language models – in AI-driven search and tools like ChatGPT – and this new landscape of search, they’re getting smarter.

While we would historically trust Google to understand and figure it out, the questions people are asking are much more complicated now, so the risk of getting them wrong is higher.

These large language models have hallucinations and make incorrect inferences. Managing your entities, taking control of how you describe them, and standardising the languages you use to describe them (which is what schema markup does) reduces those hallucinations and the likelihood that the models will get them wrong.

On the other hand, it also increases the confidence in those inferences and therefore it should get you more clicks, and more qualified traffic. Historically, people have thought that they don’t need to read about entity SEO or understand what a knowledge graph is, but if you’re not doing that now, you’re starting to lag.”

What is a knowledge graph?

“A knowledge graph is a collection of the entities you’ve identified and described and their connection to other entities. The way I explain it is in how I introduce myself: My name is Martha van Berkel. I’m an expert in schema markup. I’m the founder of Schema App. I’m Canadian. I used to drive a 1965 Austin Healey Sprite, and my red sports car was in the movie Losing Chase which Kevin Bacon directed.

In North America, we have the Six Degrees of Kevin Bacon game, and you can now infer how to win that Kevin Bacon game because you understand those relationships.

Me describing who I am, and describing all these connections, is my knowledge graph. When you’re describing entities on your website, or things within your business, and you’re explaining how they’re connected – ‘This is the author of this blog, and this author works for this company, and this company sells this solution, and this blog mentions that solution.’ – all of a sudden, you can make more inferences.

In the context of an organisation’s website, that’s quite a lot of data. That said, the most recent article I read shows that you can boost the accuracy of large language models by 300% if the data is in a knowledge graph versus a relational database.

To bring that into marketing terms, if it’s just in a basic structure in the CMS, you’re not going to get the accuracy that you could otherwise get if you’re describing the entity in detail, describing the relationships, and building what I like to refer to as your ‘content knowledge graph’ across your website.

In this world of AI search, where we want to mitigate risk and optimize opportunity, you need to think differently about how you’re translating your data for machines and controlling how you’re defining those relationships.”

Is it a type of modern-day PR and brand management?

“It is, in a sense, and then it also helps with SEO. SEOs can be the hero here because, by doing really good entity optimization or schema markup optimization across your site, you’re also helping your IT team because of this reusable content knowledge graph you’re developing (if you’re doing it well).

Gartner recently put out a list of 10 emerging technologies that are going to be really important for business decisions, which all the CIOs and heads of IT understand to mean the technologies they should be investing in. Number one was generative AI, unsurprisingly, but the second technology that was most critical to invest in was a knowledge graph. I think it’s so cool that SEOs have this opportunity.”

Is entity management something you can just do once and then forget about?

“I wish. When you’re doing entity management, you’re thinking about how you are translating and defining the entity that’s contained on a web page, which is your entity home. It could be describing a product, for example.

Just like you’re managing the content on that site, you also want to manage how that entity is being defined using schema markup, which is the standard vocabulary that search engines and other machines are using. Part of defining the entity is also defining its relationships with other things on your website and the internet.

What’s important is that, if the relationships change, you must keep them up to date. You can’t just translate this once and forget about it. This is a living data source of how all the entities on your site relate to one another, so that graph is going to keep changing.”

Who should be responsible for entity management?

“It’s a collaborative effort. The translation of an entity to schema markup should sit with a technical SEO but in strong partnership with IT so that they understand what’s there.

However, it’s all derived from the content on the page. So it really is a collaborative effort between the content, SEO, and IT team.

The technical SEO team often leads it, but all teams have to work through this together for it to be successful.”

How should SEOs train other teams about this and articulate its value?

“Some of the research Gartner has published can be useful because they speak to different audiences. The Gartner Data & Analytics Summit 2024 in London talked about adding semantic integration of knowledge graph data, as well as its standardisation and vocabulary. The research from this conference is helpful for articulating the value of knowledge graphs to IT or technical teams, and they can get educated by reading some of those pieces.

With content teams, I love to explain that there’s this vocabulary or dictionary of what machines want to know about certain topics. I’ve shown content teams the Schema.org site and said, ‘I’m so glad you’re working on this physician profile for a healthcare organisation. Here are all the things that Google, Yahoo, and Bing want to know about a physician. Is that data available, and is that important information for your consumers? Where’s the happy medium between them?’

Your content knowledge graph is a data layer that’s not visible to the consumers, but a lot of it needs to be on the webpage because it is derived from the content on a page. They’ll often have a look at that and see that they have this information, so they should connect the dots to that.

This is why it’s an ongoing conversation and it’s not just for SEO’s sake. Yes, schema markup is going to get the rich result, but it’s also going to help with a much broader machine understanding, which didn’t exist before.”

How do you measure the impact of entity SEO and AI visibility, and how do you optimize to ensure that your entity is discovered within AI search?

“This is new, and a lot of measurement platforms are coming up with new ways to see how things are performing in this new world. I have less advice on this area because it’s still so new.

What we have been able to measure, though, is when you have a page that’s speaking about an entity, and it’s been optimized and connected well within the content knowledge, from query data that mentions specific entities, we can see an increase in click-through rate for non-branded versions of that entity.

Another part of entity SEO is about disambiguation. You’re describing things, but you’re also clarifying what they are. This means connecting them to definitions – whether that be your own definition from your own taxonomy or glossary or on Google’s Knowledge Graph or Wikidata.

We have seen that, in the local space, if you disambiguate locations very specifically in the context of the thing you’re talking about – a location, local business, a product offering, queries like ‘near me’, or queries that mention that specific location – we’ve seen an increase in click-through rate for non-branded queries after we’ve done that work.

I do know that other tools are starting to look at how well you’re clustering around that specific topic or entity. At Schema App, we’re working on enabling content teams to see how often they’re talking about entities and what the coverage is.

However, this is a whole new area of marketing – because it’s much broader than just SEO. As we get more data from the ChatGPTs and Google SGEs of the world, we can merge it into an understanding of how often we show up in AI, and we’ll know whether we’re being effective in the work that we’re doing. It’s pretty early days yet.”

If an SEO is struggling for time, what should they stop doing right now so they can spend more time doing what you suggest in 2025?

“Stop just hunting for keywords and start writing helpful content around topics that explore a multi-dimensional view. What is the thing, and how does it connect? Don’t just change words in and out. Keywords are the most base structure, and they then need to be grouped and put into more context to become a topic or an entity.”

Martha van Berkel is the CEO at Schema App, and you can learn more at SchemaApp.com.