Webinar: Semantic Topic Planning with Dixon from InLinks

Still clinging to keyword volumes? That’s ancient history.

Your content deserves to be more than a numbers game.

Let’s break free from the old mold and build real authority—without endless repetition.

Peter Mead, Nik Ranger, and I catch up with Dixon Jones from InLinks to share how it’s done.

Full Transcript of the Semantic Topic Planning with Dixon from InLinks Webinar

Peter Mead:
Welcome to the Australian Duda webinar. Today is such an important topic: semantic topic planning. We’re here with two very special SEO industry people, Dixon Jones and Peter Mead. The co-host is Nik Ranger. Let’s get into the introductions. I’d like to introduce Nik Ranger, the co-host. Nik Ranger, SEO extraordinaire, is an award-winning senior technical SEO specialist at Dijon Marketing and the co-director of the SEO Collective. Welcome, Nik.

Nik Ranger:
Howdy! Lovely to be here, and it’s an honor and a pleasure to talk about semantic SEO with such incredible minds.

Dixon Jones:
I’m in the wrong place, obviously.

Peter Mead:
“Incredible” is a great word to use. Not on me. You guys are so humble.

Nik Ranger:
You guys are so humble.

Peter Mead:
So Dixon, I do need to introduce you. I consider you as SEO OG. You’re an award-winning, multi-award-winning SEO speaker and an entrepreneur who has helped to build the internet marketing industry for over 20 years. Dixon is currently an investor and CEO of in-links.net, which is the first entity SEO toolset that’s built from the ground up around its own knowledge graphs and NLP algorithms. That’s fantastic. Welcome, Dixon.

Dixon Jones:
So you didn’t read that at all, did you, Peter?

Peter Mead:
Oh yeah, it’s something I prepared a little earlier. Thanks very much for having me. It’s early in the morning here in Blighty, so I’ve got my coffee. I’m ready to go.

Peter Mead:
Alright, let’s see where we can go.

Terrific. Thank you so much for joining us in the morning for our Australian audience. Before we—I know you’ve prepared some slides which we’re really keen to find out and learn more—but before we do, Nik, did you want to perhaps have some comments about semantic and maybe warm us up on this topic before Dixon dives in?

Nik Ranger:
Yeah, sure. What can I say about entities? You’ve got your entities, you’ve got your attributes, we’ve got the semantic relationships. Entities are described by identifiers and particular characteristics, which are those attributes, those properties. We’ve got the identifier, the URI, which usually consists of a sequence of numbers and is used by machines to identify what that entity is and how humans can recognize that according to their characteristics—which is a really bunch of fancy words. But this is why I absolutely love the work that Dixon is doing, because he takes these really, really complex things and makes it really quite simple for us, and is able to scale this up really nicely.

Because when we look at this, this is really the difference of thinking. When I was writing this post, I was trying to think about the shift of thinking, and I kind of liken it to: if you look at keywords versus looking at the actual entity that it represents, it’s kind of like looking at a grain of sand and not even realizing that you’re on a beach. So I’m really, really excited about—

Dixon Jones:
Yeah, right. I thought it was a bit clever.

Nik Ranger:
It’s very philosophical, and if you think stealing it isn’t beneath me, you’re completely wrong there, Dick.

Dixon Jones:
Oh, you can take it. It’s all yours. You heard it, you heard it.

Peter Mead:
Alright, well, we love Ranger sand, and we love being at the beach. So, Dixon, I would love to—obviously, there’s no beach around us at the moment, so the next best thing is for us to dive into this highly technical topic. Dixon, I’ve heard you talk about this stuff a lot, and I’m just so happy to continue this conversation with myself and Nik. So please, take it away and let us know what’s happening.

Dixon Jones:
Alright, I’d like to, and I’m glad that you think I try and bring it down to something simple. I do, absolutely, and I have to because I can’t understand it when it’s talked about in academic terms. I would say at an academic conference—well, I was supposed to be at an upcoming conference in New York in March. I got COVID, so I couldn’t get on the plane, and so I did it virtually. But the way that they describe things is just too far away from reality to have practical use, and that’s the same in a lot of research. So I spent a lot of time trying to figure this out and translating it into words I understand.

By the way, when you’re in with a tech SEO, they start saying URI instead of URL. Okay, can we dive into my slides? We’re going to be talking about a different way of looking at content planning, or as most people call it, keyword research. I’m calling it content planning. As you’ll see by the time we get to the end of this, it’s not so appropriate to call it all keyword research.

The purpose of what I’ve got here today is to hopefully get people to understand exactly how Google sees content as a set of entities—related entities—and how you should see your own content in that way. Then also explain how entity SEO—planning from an entity-based basis instead of a keyword basis—is better.

Oh, we’ve got somebody from Italy in as well. Hello! Also, we’re going to learn how to optimize our content using knowledge graphs instead of keywords, using topics instead of keywords. So let’s dive on in.

My slides don’t show at the top, but I’ll just explain anything that seems to not appear quite on the top. Who is this and why should you care? Yes, I’m Dixon. The other thing that you haven’t mentioned is I wrote a book at cseo, moving search from strings to things. It’s not a very long book. If you can’t afford the two dollars on Kindle, then most of it’s free on the endings.net site as well. So it’s not out there for the money, it’s out there for me trying to literally get these ideas down into print so I could understand them.

These are a few graphs, not all from our company. You don’t have to use our technology to get this kind of stuff, but it doesn’t happen for every site. If you start moving from a keyword approach to a topic-based approach and an authority-based approach, then these are some of the results from three different case studies which are up on the site as well. They can have an impact.

The slides—if you send an email to talk at Dixon Jones with “planning” in the title if you want the slides, then I’ll send the slides through as well if that’s okay. Let’s get into the presentation. Let’s start here.

I’ve got a new car, so I had to put it into a presentation. Oscar Martin, that is a Mustang. It’s a Mustang Mach-E electric. The alarm just went off just now so it’s clearly not quite right. The problem with these two entities is they’re both called Mustangs, one being a wild horse and one being a car. The issue is if you’re a search engine, or if you’re a user, and you’re typing “Mustang” on its own without any context, it’s not clear which of these things you’re referring to. This happens quite a lot in the car industry.

This is my other car. No one knows what this is going to be, but this is actually called a bulldog. It’s actually called a Pilgrim Bulldog. If somebody types in “Pilgrim Bulldog” or “bulldog”, it’s highly unlikely that they’re going to be interested in this car. With Mustang, you might think that people are obviously going to choose a car and the brand; with a bulldog, they’re much more likely to choose the dog. Pilgrim Bulldog gets a little bit more confusing though because, what is the relationship between a pilgrim and a bulldog? Well, there’s only one relationship, which is this. When you start connecting the dots, it becomes obvious in our own human minds what’s happening. If you connect the dots with a picture of a car or a horse, it becomes obvious. If you start connecting a picture of a pilgrim and a bulldog, it looks really odd. It doesn’t seem to make sense. It’s not ideal.

The other thing about entities is an entity is a thing, but a word is a label. All of these things mean the Eiffel Tower. Frankly, I can’t even understand some of these. That one with really curly writing, that’s Sri Lankan. That’s a beautiful writing, I like that. But they all mean the Eiffel Tower, whether it’s the Tory fell or a big metal tower in Paris. They’re all exactly the same thing. One of the big problems with keywords is that they are labels for entities and objects. Sometimes that label, a Mustang or Bulldog, can mean one of several different ideas. Sometimes, and almost always, an entity just has loads and loads of different words. In fact, they always will because they have them in different languages.

It’s not always easy to see the differences between entities either. This is a bunch of different representations of New York, all in graphical form, from the tube map to the sewer map of 1735, which is covered up because I haven’t got any transitions on there. They all show roughly where things are and how far apart. They’re all maps, they’re all maps around things of New York, but none of them is New York. It’s not always easy to say what an entity is. This is not New York, this is a representation of different things within New York.

That means that Google is pretty confused at times. If I type in “bulldog”, then I’m going to get a French bulldog in here. If I type in “French”, for example, then I get French Bulldog, probably because I was looking for French Bulldogs or something beforehand. If I type in “Sicilian”, I get loads of things to do with pizza. If I type in “French”, I get all sorts of things like French Bulldogs and French onion soup. If I type in “French versus Sicilian”, then I get the difference between Italian and French style. I get all these suggestions that are coming in, but then, fourth one down there, I get “Sicilian versus French versus Caro-Kann.” That is me, because I’m a big chess player, and frankly any chess player that heard “French vs Sicilian” would absolutely know that they’re talking about chess at that particular point. When you hit the results, all of the results are there because there’s a lot of authority that has been written around the French opening or versus the Sicilian opening. It’s very obvious when somebody really starts going down that authority that this kind of content becomes much more important than the French style versus a Sicilian style.

So the question is, how is it that Google starts to disinvigorate all of these dirty signals from queries, from corpuses of texts on the internet, and how does Google know why we’re talking about a game of chess instead of style, or about a car instead of a horse? Why does knowing this give your website a business marketing edge? It gives your business a marketing edge. That’s really what I’m diving into today, and then how to exploit it, hopefully.

What’s a Knowledge Graph? You explained it in one way, Nik. I explain it this way: this is a poster that was in a pub just before COVID that I saw and I thought, now I understand what a knowledge graph is. It is really a collection of ideas, and they have a relationship between each other. Within lager, there’s European lagers, there’s German lagers, etc. This pretty much for me is as good a knowledge graph for beer as we’re ever going to find. Ultimately, I see knowledge graphs as just these kinds of representations on the internet. I don’t think you’d see it that much more. What I think is really interesting is that Google starts with a knowledge graph of ideas, which is an encyclopedia with a bit of a database kind of feel to it. But then when it starts to read content, it starts to see ideas in content and it extracts those ideas and says, “Ah, that idea is related to this entity in my knowledge graph.” Essentially, it can sit there and see a small corpus of text and it can map it into numbers because all those numbers are a representation of entities and ideas.

Clearly, if you’re going to talk about everything under the sun, fine, you’ve just created yourself a new Wikipedia all in one page. But if you’re going to start talking about ideas that are around Mustangs, that is going to have a very different pattern to it if you start talking about horses than if you start talking about cars.

Nik Ranger:
Because you’re already talking about the idea of relevant topics.

Dixon Jones:
Yeah, so this chart is also trying to show you clustering as well. There’s a bunch of ideas in green, a bunch of ideas that fit into the blue cluster, a bunch of ideas that fit into the red cluster. But this is not talking about the red—sorry, the green cluster is not talking about the red cluster or the pink cluster. These kinds of ideas also differentiate yourself from other things. It means that you’re definitely not talking about horses. Of course, there are some overlaps. If you talk about horsepower, that’s common to talk about horses or talking about cars—except when it comes to electric cars, we like to talk about torque. Anyway, that’s a whole different story.

So then, it obviously starts to become much easier as humans and as machines to see the differentiation clearly, those different topics. It now becomes really clear, hopefully, if what I’ve said has made sense so far. It’s really obvious that if you’re talking about stables and the wild and animals, that’s a completely different thing to talking about torque, electric, and Ford, even though you could be talking about the Mustang as a wild animal. It does say that if you start talking about describing the Mustang as a wild animal, then you may be confusing the graph and you may actually be reducing your chances of reducing the confidence that a search engine has that you’re talking about the car instead of the animal.

The point is, using this is important, but yes, getting those words like “wild animal” out there in and describing a Mustang is interesting because if you put in Mustang and then look at all the words that are being talked about on all the pages that are ranking for Mustang, probably “wild animal” is going in there. If you were looking at it and analyzing it from a keyword base, you might say, “Yeah, I do want to put wild animal in there,” but using keywords just because they’re in a list of words around the content of the word Mustang is actually a huge mistake because you’re going to see words that are out of context to your situation.

So if you take all the words around the phrase Mustang, “wild animal” will be in there whether you’re talking about the car or the animal. But “horse” will be in there, and “hooves” would probably be in there because some of those pages that are authoritative around the concept of Mustang are talking about horses. To put them into your own content just because they’re in the list of content that’s ranking for the phrase Mustang is a big mistake if you’re trying to talk about cars, because you’re just polluting your own pool of authority. This is why you have to group these ideas into some kind of context. Does that make any sense so far?

Nik Ranger:
Yeah, just on that, using keywords in your content simply because you’ve done some keyword research and there’s volume for a certain keyword that happens to be related to what you think is your primary keyword.

Dixon Jones:
Yeah, and that seems, from the history we’ve had as SEOs, that kind of seems like not such a bad idea. From where we came from, originally, here’s your primary keyword, here’s your secondary keywords, just pepper them throughout the page. But it’s obvious really when you think about it, because Mustang is two very, very different things, and clearly not everybody that types in Mustang is thinking about the version of Mustang that you’re thinking about. So right at its start, it’s like garbage in, garbage out, isn’t it? There’s a million people this week who type in Mustang. What percentage of them are typing in Mustang because they’re talking about cars? What percentage are talking about horses? Whatever percentage that is, the one that’s not talking about what you want is absolutely diluting your audience. They’re not going to buy a car if they’re interested in a horse. We could be fairly certain.

This happens all the time. We take these words, and sometimes it goes the other way. Another example I use is house and home. Do you want house insurance or home insurance? If you go on to Google and type in house insurance, it comes back in the UK, or did when I was looking at it the other day, with home insurance in the results. So it’s already worked out that a house and a home are exactly the same thing. What the author writes doesn’t make a difference. It’s entirely the concept that’s important rather than the label that’s used to say it. Just because keywords are used in your competitors’ content doesn’t mean to say you should use it in your content, and they need to be grouped together.

There’s a lot of tools to do it. This is an Ahrefs export where they’re grouping keywords into parent keywords. I’ve moved away from cars now because it’s not everybody’s cup of tea, but I stayed with the wild and gone to the Azores, which is a beautiful set of islands in the Atlantic, probably not as beautiful as some of yours on the Barrier Reef, but anyway. So you can get keywords grouped. They’ll try and group the things. SEMrush also do it. At the bottom of their export, they’ve got spreadsheets, so they’re grouping them by Madeira and weather and travel and visits and flights to the Azores and stuff like that. So they’ve got those kinds of things grouped. The tools that are out there do group keywords and they have keyword volume, but the context and the intent, to my mind, are the missing links in the chain.

• Webinar with Tim Soulo from AHREFS

I think we’ve kind of tried to make some improvements on understanding why people are typing something in and trying to get that context, not just based around ideas, but also ideas that are relevant to your own website and your own business. That’s where we’re going to dive into next. I’m going to start with intent because I’ve got a huge problem with how SEOs understand intent. The truth is that most SEOs say that intent is one of four or five things. It’s either informational, navigational, transactional, commercial. Even Wikipedia has these types of intent out there. That’s really odd because those types, that little red line there, but this blowout in their first sentence says that intent is about the identification and categorization of what a user online intended or wanted to find. That’s nothing to do with transactional, five things. The idea that SEOs go up in the morning and say, “Right, I’m going to make sure that my website is a really good informational intent page or navigational intent page,” doesn’t really help to sell a Mustang, whether it’s a car or a horse.

Intent is much more than those five ideas. I’ve got no idea why. I think it was in some research paper, a Google research paper somewhere, that these five ideas came out, but frankly, it’s all—well, I don’t know if we’re outside on a podcast, but it’s all bollocks, really. If you start with the Azores and you start putting that into Google, Google says, “Okay, right, there’s the Azores. That is a region in Portugal,” and then it comes out with a whole load of suggestions based around the islands, the holidays, weather, whatever.

If I started with horseshoes, then it’s even more rich in concepts and entities. Are you talking about horseshoe crab, Horseshoe Lake, Horseshoe Park, Horseshoe Falls, or Horseshoe in Wallingham, which is local to me? It’s not that local, but it was when I did the screenshot. You can see this blend of keywords, results, and entities just in the Google suggestions up here. Google is trying to understand the underlying intent of the user at this particular point, and they’re not trying to work out at this stage whether they’re trying to buy something or learn about something. I think this is something that SEOs totally misunderstand. They just don’t understand that intent is so much more than one of those four or five ideas, and the results have come back really reflect what Google knows. When you’ve just got a word like “horseshoes”, you haven’t given very much information to Google about the intent, so it comes out with a bunch of results that are about local results because that was local to where I was at the time in North Wales, and it’s got a bunch of core results coming back. In this knowledge panel here, two ideas about horseshoe: one is called horseshoe, and one is probably the more traditional thing that goes on the foot of a horse.

Those are the ideas, and I think that understanding intent is where SEOs should be really starting because your business only caters to specific types of intent. You aren’t trying to sell that horse, don’t sell them a horse, don’t write about the horse, don’t confuse Google, certainly don’t get a horsey person to your website if you’re trying to sell a car. Building your knowledge graph means you’re building a knowledge graph that is personal to you, personal to your business. You’re creating a fingerprint of concepts and ideas that hopefully will suit a specific type of buyer or a specific type of person that is going to be interested very much in your world.

The next question is, well, how do you build your own Knowledge Graph? That’s right out there, right? The only people that build knowledge graphs are Google and IBM, surely. It’s actually not that difficult to build a Knowledge Graph. We’re going to go through one, we’re going to build one in three steps with a little bit of help from Excel maybe.

The first thing is, find authoritative content around a subject. If you wanted to build a Knowledge Graph around Ford Mustangs, then take the 10 best pages for Mustang Mach-E thoughts, the specific model of car that you want, find the best reviews around those. Don’t take the pages that are about wild horses. If you want to build a Knowledge Graph around Mustangs or wild horses, fine, take the 10 best web pages for that. Then you’ve got to go through that and look at the underlying entities. If we think of entities as anything that has an article in the Encyclopedia Britannica, then you’re in a pretty good place. Think about an entity as something that’s got its own page in a dictionary or in an encyclopedia, and then tabulate the results so you can see how often you talk about things. That then comes into a list of entities.

Now, it takes a long time to go and read 10 pages of content and break it down, and there’s going to be human error, a lot of human error. Fortunately, there’s plenty of tools out there and they don’t all cost money. Google has its own NLP API. It’s not very good, or rather the results that come back don’t show as many entities as we see as humans. I don’t know whether that’s on purpose or not, but we test it in Links, our own NLP, against Google’s, and we reckon that Google only reports around about 15 to 20 percent of the entities on a page that we see as in Links or you would say interesting.

Nik Ranger:
That’s very interesting because that’s definitely the tool that people talk about when they first start talking about semantic and figuring out the entities they’re using.

Dixon Jones:
I think it’s interesting as well, and if you go to inlinks.net’s homepage, there’s a resources tab. We’re going to change the website next week because we’ve got the .com now. Go to resources, industry reports. We’ve been tracking for years and years and years, well, for a couple of years, around different phrases. We’ve been taking the results and breaking it down into underlying entities and then running our entity analysis algorithm and Google’s entity analysis algorithm, and on average we’ve been seeing that they’ve only been getting 14 and a half percent, 15 percent. In real estate, it’s much better, and that’s because they’re really good at seeing brands and places. Real estate involves lots and lots of cities and towns, which are quite good at getting, and watches and jewelry are very brands-driven, so those work quite well. When you get to things like tools and techniques, they’re much worse at picking out entities because those are much more talking about concepts.

You can use Google’s NLP API, you can also use IBM. This screenshot is IBM’s Watson, and they break it down. We also beat them. The other one that’s well known is TextRazor, a French company, and they’re pretty good. We do beat them, but we try and beat them, and I’m sure they would argue that they beat us. The point is, there are plenty of tools that you can go to that will make that job of breaking down a page into its underlying topics a lot quicker than having to do it manually, and it’s going to save you some time.

The idea here is that you can go through and have a look at a bunch of different web pages, whether the ones in the SERPs or you could swap out ones in the SERPs for pages that you think are authoritative regardless of the SERPs, but the SERPs is kind of fun, the ones that Google thinks are good. Certainly get rid of the ones that talk about horses if you want to have a car website, and then you merge it into one set of ideas. You can see that some phrases are discussed, some topics are discussed much more than another. This is the output from doing a topic analysis for the Azores, and so we can see that if you want to talk about the Azores, then things like islands and archipelago and it’s owned by Portugal, so Portugal and the Atlantic Ocean, these are things that are mentioned by all of the competitors. In this representation, we’ve taken the top 10 websites for the Azores in the UK, and the concept of island is talked about on average five times by the competition. One of the competitors talks about islands 15 times on the page, one talks about it three times, so we’ve got a kind of high, medium, and low numbers here. We can sit there and say, right, we probably need to talk about the islands and archipelagos more than we need to talk about climate or the earth, if you see what I mean, or lakes. Very quickly, we build up a concept of ideas, and you don’t need anything to do that, but obviously it can speed it up if you use some tools to get those entities quicker.

What we’re trying to do is create some very actionable recommendations. This is the bit where we go to the next level, or we think we go to the next level. I invite the panelists to say whether we do or not. What we do is we throw all those ideas back into Google Suggest. You’ve got to sit on your computer whilst this is all happening, but we’ll take these underlying entities, we put them back into Google Suggest. Remember that Google Suggest where we saw horseshoe and you started seeing Horseshoe Falls and all these different things about horseshoes? We can then take those and then we can run an NLP algorithm into those and compare what those ideas are, and because we now have our own Knowledge Graph, we can see where they’re close to what the business is already talking about and discard the ones that are irrelevant. Now we can go, right, you’re talking about horseshoes, or we’re going to go through all your content, you’re talking about horses, not about the pub down the road, so therefore we need things that are related to horses, etc. We’ll throw out things about the local results of Wallingham or whatever it may be.

Now we’re coming back with a set of topic clusters and within that another set of keywords. We’ve got some volumes here, and we use the big boys for these overall volumes, but really I’m not so interested in these volumes. I’m really interested in the fact that the Azores tourism statistics is something that people are interested in, funnily enough, but not as interested in as tourism guides. I’ve now got a section on my website and got the underlying concepts I need to talk about, I’ve got the questions, and I got all this, and ultimately from Google’s data, so I know that these are things that Google thinks are important as well, and they’re all related specifically to my own website and my own business.

Nik Ranger:
Can I throw a spanner in or be a hypothetical devil’s advocate? If we all follow this process, whether it’s using your tools or any of the other tools, aren’t we all going to end up with the same kind of content?

Dixon Jones:
Absolutely not, unless you have no uniqueness to your business idea. Let’s take a hotel chain versus an airline. There are two very big differences between what those two businesses are trying to do with people going to the Azores, for example, or if you visit Azores tourism website. I built this knowledge graph. In order to build this knowledge graph, I had to start with a website where I started this with a keyword, but what we can also do is start with a business. We can take the whole website, and I’ll come on to that in a second and show you how I build a very different Knowledge Graph based on two different businesses and what your businesses do. If there is no differentiation between two businesses, then ultimately you’re competing for exactly the same share of voice, and all you do is the more businesses that are exactly the same, the more you’re diluting the market for it. Most businesses, most business research, at least when I did my MBA, which I did recently because I couldn’t afford to do it when I was a kid.

Nik Ranger:
Congratulations.

Dixon Jones:
Thank you. Product differentiation, regardless of SEO, is the mantra. If you’re going to be trying to develop your business, you’ve got to have products of differentiation, you’ve got to have some kind of uniqueness to your business that is either hard to copy or so unique that no one else wants to copy it. All of these are fine, but you’ve got to create a barrier to entry of some sort. It could be a knowledge barrier to entry or it can be a product barrier to entry, but you have to have a differentiator in your business. Where you start, if you start with just the phrase “the Azores” and do this, then yes, we’re all going to come out with the same information. So you make a very good point. How do you differentiate yourself? I’ll come on to that in a second, but before I do that, I want to also bring you back to the user intent layer.

The other thing that we did very differently now, and it’s only been out for a few weeks, is instead of thinking that user intent is all about navigation, transaction, local, whatever those five things were, it’s much better to understand user intent using verbs. When we started throwing all those ideas back into Google Suggest, we started picking out the verbs. There are people that are trying to travel to the Azores, people trying to go to the Azores, make, I don’t know what they’re trying to make, look at the Azores, but I’d have to click on the “make” to be able to see the things within there, or pronounce the Azores. People are trying to do different things around the Azores. Knowing the verbs and grouping things by verbs is going to be very interesting because if you take something like home or house, are you trying to have you got a business that decorates a house, mortgages a house, buys a house, sells a house, shows your house, whatever, or is it about house in a full house, game of cards, or something like that? I think breaking down user intent by verbs is a very interesting alternative way of looking at things because you can cut out the things that are not important to you and not important to your business. Pronouncing the Azores may actually help you rank better for the Azores, but it won’t necessarily help you sell tickets to the Azores, and Google will have worked out that when somebody is trying to pronounce the Azores, they’re in a very different place to whether they’re trying to go to the Azores, and hopefully Google won’t return the same results even though the keyword is the Azores.

That means you can also automate schema, probably not part of this presentation, so I’ll just move on to that. Going on to the question that you had, why doesn’t everybody do it and end up with exactly the same stuff? They do if they just go with that, but now you can also internally link the entities. This is where things start to differentiate because your corpus of authority is building up as you start writing this content and you start interlinking stuff, and you start to show a search engine where your authority lies in what particular aspect—pronunciation or travel—and Google knows whether your site is about pronunciation or travel, and it’s going to start slanting your results based on the internal linking, on the understanding of the website as a whole.

The best bit of it is this bit at the site level because if we extrapolate this whole idea and instead of building a Knowledge Graph around a keyword, we start building a Knowledge Graph around your business, then we suddenly start to have that differentiation that is really important to this whole process. If we take a business in real life, this is Azores Choice, which I have no relationship to at all, I’ve always wanted to go to the Azores, my wife doesn’t, she thinks there’s a barren set of islands in the Atlantic, but I think that looks pretty nice anyway. What I’ve done now is I’ve taken a Knowledge Graph not of the keyword “the Azores”, but a knowledge graph of all the pages on the Azores site because these were the keyword research reports from Ahrefs, SEMrush, and InLinks for the Azores, and none of these are useful at the site level particularly. We try to make this useful at the site level, realized we needed to improve the product, and so we started again. The reason is because starting with a keyword, but if you start with your own web pages and start crawling all of those and make a Knowledge Graph over those, then you build a knowledge graph that is personal to your business. Take all your pages, go through that same process of getting out those entities, and then create a knowledge graph that is very specific to your business.

Now, if you’re going to be talking about French versus Sicilian, then you’re a chess website. It’s going to be really markedly different to French and Sicilian compared to a style website. We can then do that, and then we can throw those entities back into Google Suggest, and that will expand that Knowledge Graph and come out with suggestions that Google is suggesting based on your business. Does that make that difference succinct and answer the question you had earlier, Peter?

Peter Mead:
Oh yeah, it does. I love this idea of building your own Knowledge Graph with your website. It’s a thing that I’ve thought about for quite a while, since really since Bill was talking about this. Then taking it on to the next level of getting the suggestions tells us where to go next, right?

Dixon Jones:
That’s the cream on top there, isn’t it? I’m hoping so. Instead of this—this is the Azores with the keyword research level, what do we got here—resource, travel, islands, adventure, and a bunch of keywords. What I get now is this, which is, okay, Azores Choice, right, we want to start talking about things around the Azores, and then a bunch of keywords around the Azores and a thing around holidays and stuff like that, and then within that we’ve got school holidays and stuff. Within here we’ve got Portuguese holidays, and these are all the Portuguese holidays because the Azores has Portuguese holidays as the public holidays. We start building a keyword research that’s based around suggestions that extrapolate from the existing business. Now, you’re building your site structure around your business, not somebody else’s business. You’re extending what Google thinks is important as the next steps for your site. If you’re going to be talking about beer, maybe you need to talk about beer styles or beer mugs or whatever is going to be there.

I’m pretty much near the end of the presentation. That’s kind of, as you say, the icing on the cake, going on to the next level, and it really does start to help give your content planning some structure based around entities. If you go back to the title of the talk, which was along the lines of content planning the semantic way or the topic way, hopefully that demonstrates how we’re trying to help people develop suggestions. We think that SEO is about matching topics and entities to real user intent, and we think that you can create your own ad hoc Knowledge Graph without any particular tools, but scaling it needs at least an NLP algorithm because humans will make mistakes, they’ll be blind to words and texts they’re very familiar with, and they’ll miss things out and they’ll read one page and pick it up and another page and not pick it up. At least an NLP algorithm gives some conformity to building that up, but there’s plenty out there.

Once you’ve got knowledge graphs, NLP, natural language programming, and Google Suggest together can create better content around clustering ideas. It can also create schema fast—we didn’t talk about it—but again, it can internally link things very quickly as well. I listed InLinks, hence the name. Then we can also plan content using this gap analysis of entities as recommended by Google Suggest based on your existing content, which is where I think the future of content research and keyword research should be going, and it should pretty much throw how many people typed Mustang into the search engine out of the window.

That’s the slides, and you’re the first people to see them, so hopefully it made sense.

Nik Ranger:
Heard it here first, guys.

Peter Mead:
That’s it, right, we can go off the slides now. I think that’s enough of me. Talk at dixonjones.com with “planning” in the title if you need them out there.

Nik Ranger:
I’m emailing you right now.

Peter Mead:
You’ve probably got a recording. I’m sure Duda will give you the recording so it’s going to be okay.

Nik Ranger:
I’ve got FOMO already. Did it make sense?

Peter Mead:
I think it definitely makes sense. Well, it made sense to me. I’m just very mindful that there are so many people out there that are like, “Great, I’ve listened to this, I’ve loved this, I’ve got a handle of the concepts,” but then they’re looking at their pages and the site’s maybe got a ton of boilerplate content. Whoever has been building all this stuff has just kind of been blind at the wheel, being like, “Cool, I’m just going to internally link this or I’m just going to create the content,” and they’re kind of having to go back to the same people and then start to talk to them about this shift of thinking and how they’re able to do it. What do you sort of say to, you know, right SEO or to someone else who’s like, “You’ve got plenty of places to go to cull that content and you don’t need to just take it off your website,” because that’s going to have political—you could have some evidence to take it off, but what you can do is instead of analyzing your whole website, you can either go to Google Search Console, look at the best pages according to Google Search Console or your analytics or just your knowledge of the business and sit there and say, “Frankly, we sell widgets, blue widgets and green widgets. We don’t sell red widgets, we don’t want a choice of widgets, we don’t want all these how to wipe your bottom with widgets,” so all that stuff, we just take the important pages, if you like the money pages, and build the knowledge graph around that. You don’t have to build it around your whole website.

Dixon Jones:
It’s quite possible that you should delete all of that other boilerplate stuff off your website, but that’s not the topic of today’s conversation. Another way of dividing out—for those that know me, I’m also a brand ambassador at Majestic and used to be a marketing director at Majestic—so I would use flow metrics to work out which pages are already important in the old sort of PageRank model and use those as my seed set. Essentially, you’re creating a seed set of content, usually out of your own business. If you haven’t got a business yet, maybe you need to take pages of website, the pages of your competitors that you aspire to, and sit there and say, “I want to sell this and make this and advise on this, and these are the three pages out there that are most important,” and take those to build your knowledge graph, and then you’re starting with the content from scratch. At that stage, you might as well consider taking your keywords, building a keyword research based on the top ranking pages for a particular phrase. Just don’t put Mustang, say at least Mustang horse or Mustang car so you get the right pages. Make sure you check your input before you go there. To your point then, Nick, just don’t put every page into the knowledge graph that you’re building. Put in the pages that are relevant for your business, not the ones that are relevant just because some SEO 12 years ago, when things were different, decided to try a keyword boilerplate for everything.

Nik Ranger:
Because I think a lot of people are just like, “We’re just going to tackle this with volume. We’re going to write everything we possibly can write about this topic.” You’re just going to punch it all in, but what they’re not doing is thinking about the content hierarchy, and they’re certainly not thinking from an internal linking perspective.

Dixon Jones:
No, and they’re missing context to that point, but they are right in that if they have decided that they want to be an expert in a field, they want to, as our knowledge graph manager says, sort of flood the service with authority on that topic. But don’t flood your authority yourself with authority on another topic. I always come up with Bill Gates, but Bill Gates is really an expert on AIDS research, funnily enough, and operating systems, not so much on Beyoncé. If he starts talking about Beyoncé, that doesn’t make him an expert on Beyoncé, unless he kept talking about it. Ultimately, he’s a fan of that particular beer, absolutely.

Peter Mead:
So Dixon and also Nik, how far down this track are we now? How far has the whole world of semantic SEO gone? Do we just completely, whenever we do keyword research and we look at the volume column, do we just completely ignore that, or can we still look at topics and find that there are actually some other keywords and we need to know the volume and we have to have some, or are we just going to go, “No, forget that, it’s just all about topics, entities, ontologies”? Is that what we’re doing? How far are we?

Nik Ranger:
I’m more interested in what motivates a click rather than just an arbitrary search volume, which to be honest, when you look at Search Console, I have some thoughts about search volume. Things get so much time and attention and impressions, and then you look at the volume, it’s like, this is like 10 searches per month, something just doesn’t add up here. I look more from what is actually motivating people to click through to these things, how are they interacting with the site, how much time on page, all those other little tiny things that build up an idea about the target audience that I’m more interested in.

Dixon Jones:
Of course, it’s a very quickly dropping graph of how many people type in any given query in the world. I remember Google saying, some years ago now, but I wouldn’t be surprised if it’s still the same today, that half the queries typed in every day are unique. Certainly, the long tail volume-wise ends up as the more important and it’s the high converting, so SEOs have known that for years. They want to rank for insurance or whatever their business is, they’d love to do that, that’s great, but ultimately if somebody just types in “insurance”, they’ve not given themselves enough context to understand what they actually want. Somebody doesn’t want insurance—imagine typing in “insurance” and saying, “I want insurance today.” Insurance and what? Insurance that your wife isn’t speaking with somebody else? Insurance—it’s just a meaningless concept on its own, and the volume is crazy. On the other end of it, getting prenuptial divorce insurance is probably highly specific—unfortunately, I haven’t been there—but highly specific, and very few people type it in, but if that’s what you provide, then that’s the only way you’re going to get money. It’s only got 10 searches, but if you’re the only person that does a particular insurance for ensuring your prenup, then that’s a very specific thing, it’s a product differentiator in the insurance market, and you don’t care there’s only 10 searches a month. Well, you might care from your business planning, but you don’t care from—you do definitely want that search phrase as something that you own, not just that phrase, but “how do I get prenup insurance” and all phrases around that. You want to be an authority in prenuptial insurance—I don’t know if it exists.

So at that point, keyword volume doesn’t matter. Intent is what matters, isn’t it?

Peter Mead:
So you mentioned that you don’t really need to use all the tools, you can manually build your own Knowledge Graph. What about just go out and hire a knowledgeable writer who’s an expert in the topic and just get them to start creating all this content and build your knowledge graph that way? Don’t worry about schema, don’t worry about any other thing. How far down the road will that kind of an approach go, do you think?

Dixon Jones:
It has already got to the point where you just type the Azores into GPT-3 algorithms or GPT-4 algorithms and it will create the whole content for you, hasn’t it? So it’s already gone in a way to that extreme. The problem is that you haven’t given it the context of your own business at that point, and you’ll end up creating a lot of different boilerplate content to what you had before. We’re throwing out what the SEO did 12 years ago, and Nik’s SEO will go along and say to their boss, “All that boilerplate that you’ve thrown, throw all that out, we’re going to put all this boilerplate stuff in now from YouTube.” It doesn’t substitute real knowledge, and I think if you’ve got a good content writer, a good content writer will be using these concepts. They might be using them still with a sort of a keyword mentality instead of an entity mentality, but ultimately, if they’re going to be experts in the field, they’re going to write hopefully good content. They still need to be a good writer.

I did a webinar yesterday on how the machines, how to train the machines in AI, and Marco Giordano, who’s in Switzerland, came up with the very interesting point that SEOs don’t spend enough time concentrating on sentence structure. These machines are trying to extract these entities, but if you’re an intelligent person on a particular subject, you have a tendency to write complicated sentences, and complicated sentence structures are difficult to unpick and so they’re not going to be as good at undoing those concepts. I was delighted, Nik, when you said at the start, “Dixon reduces this to something simple or tries to,” because I think I have to for my own little brain, and I know that Google’s search engines have probably got the intelligence—well, they’ve got no intelligence—but they probably can mimic a two-year-old, so they’re not at the level of an adult as we understand, so we need to give them concepts in very simple sentence structures, which a good writer might know. A good content writer may know the topic, but even so, they will have blind spots. We all have blind spots in our—we all have bias, we absolutely know that we all have bias, whether we like it or not, and we need an external force to get through that bias. The fact that we’re talking about a particular topic and we don’t talk about something that other people that don’t have that bias realize needs to be talked about, if you don’t pay attention to that, we won’t pick that up even if we disagree with it. Let’s say Trump versus Hillary Clinton, and you’re not addressing the other person’s point of view, doesn’t necessarily make you an authority in the subject if you’re just going to put out your point of view. Then you haven’t necessarily covered all of the issues.

Peter Mead:
Similarly, what about if you have a writer who’s not necessarily an expert, thinking of that here, but they’re not necessarily an expert in the subject and so the words and phrases, the sentence structure don’t come across like an expert who knows the subject matter. How does that—obviously that—well, my feeling is that still feeds back into your semantic SEO as well, but we’re sort of getting down here to two minutes left, actually only one minute left. I think that perhaps if we can get, what’s your 30-second final thoughts, Nik Ranger?

Nik Ranger:
My 30-second final thoughts? We’ve talked a lot about content. I want to see this methodology and this thinking be applied to products. I think that being able to understand what is an entity and the relationships between entities and apply that to your tags that you’re applying to products, because again, a lot of the time when we’re seeing these shifts in the focus, a lot of the time we’re getting either too much drill down or we’re trying to cluster too much into these product pages, and that’s why we get all these kinds of other pages added to a site architecture because no one has an idea how to actually order and structure this. I feel like we’ve talked a lot about content. What are people that I talk to care about? Revenue. Content seems like a little bit further down the road for them, so think about that. There are so many wonderful ways you can use this in your methodology to basically transform the way that you work, so also taking it from that perspective too.

Peter Mead:
That’s another perspective on that, isn’t it? Dixon, what do you think? Sorry, I think my 30-second takeaway is, well, firstly, a keyword is just a label for a thing, and you’ve got to remember that a word can mean many things, there could be many words that describe the same thing, certainly in many languages, and it’s the things that are important, not necessarily the way that you say them. I think that’s the main takeaway in thinking about it. The second part of this is we’re not saying we’ve got the ultimate way of doing search engine optimization. What we’re saying is we’re trying to come up with tools and methodology which plays better to a Knowledge Graph-based approach. We know that Google builds its life around Knowledge Graph these days. When we started inventing SEO 20, 25 years ago, 1999 for me, we didn’t work in that way, and so we have to come up with these new methodologies and these new approaches that are in harmony with the way in which search engines are working these days. We can provide the tools, but just like any other tool, you can use this tool and completely screw yourself up. You’ve got to understand the reasons for it before you start using these tools.

Nik Ranger:
Some people are just like they over-engineer it.

Dixon Jones:
Yeah, here’s a new shiny thing, let’s go and smash it up.

Peter Mead:
That’s the end. Don’t forget your LSI keywords too.

Dixon Jones:
Absolutely, that’ll help.

Peter Mead:
Amazing. Thank you so much, Nik, for being co-host and really helping to bring this whole topic alive. Dixon, why do we say this is blowing my mind again? I’ve already sent my email to get those slides.

Dixon Jones:
Thanks so much. Thank you very much for having me, and thanks for letting me go through all the slides as well. I just got them right in my own mind because you’re the first people to see those slides, so very privileged.

Peter Mead:
Thank you so much for watching here too.

Nik Ranger:
Absolutely, thank you to all the viewers.

Peter Mead:
We can’t wait to have another conversation soon. Thanks so much and we’ll talk to you all later. Send us a tweet.