Glad to be back with a new episode of The Search Session. I’m Gianluca Fiorelli, and today we’re joined by Dixon Jones, a veteran in SEO and digital marketing.
We talk about semantic context, entities, and topical authority—from Majestic’s Topical Trust Flow to tools like InLinks and Waikay—and how schema, internal linking, and topic hubs help brands improve their visibility in AI-driven search.
Some of the key topics we explore include:
Semantic context and topical relevance in SEO: how Dixon’s early work on topics and entities behind links led from Majestic’s Topical Trust Flow to the Entity SEO book and InLinks.
A business’s entity “DNA”: how clear entities, connected through internal links and structured data, create a knowledge graph that machines can understand.
Schema and internal linking: structuring entities and topics so one clear authority page exists for each concept, reducing cannibalization and helping machines understand site architecture.
Schema as a shared entity language: using widely recognized sources (like Wikipedia or other databases) to describe entities so machines can connect a business’s internal ecosystem with the broader web.
AI visibility and brand understanding: comparing what LLMs know about a brand with what a website actually covers to find visibility gaps.
Using AI visibility data strategically: combining prompt tracking, entity analysis, and visibility reports to guide content strategy and monitor brand perception in LLM responses.
The missing hub problem: how even large brands scatter their content instead of building central topic hubs that clearly establish authority around key subjects.
Building credibility before tools: why SEOs should establish expertise and industry trust—or partner with someone who has it—before launching a product.
Enjoy the conversation and get ready to learn from one of the best in the industry!
Dixon Jones is a long-time figure in the SEO and search marketing with more than 25 years of experience in the industry.
After co-founding the SEO agency Receptional in 1999, he served as Marketing Director (CMO) at Majestic from 2009 to 2018 and continues to work with the company as Global Brand Ambassador.
In 2019, he co-founded InLinks, an entity-based SEO platform, and launched Waikay in 2025, a platform designed to analyze how AI systems understand and represent brands.
Dixon is also the author of the Entity SEO book and a well-known international speaker in the search industry.
Transcript
Gianluca Fiorelli: Hi, I am Gianluca Fiorelli, and welcome back to The Search Session. Today we are going to talk to—and don’t take it wrong—one of the oldest SEOs still working in the industry and one of the most relevant SEOs.
I will tell you just three names: Majestic, InLinks, and Waikay. You probably already understood who is going to be our guest. And our guest is none other than Dixon Jones. How are you?
Dixon Jones: Gianluca, I’m very well. I’m very well. And yourself, sir?
Gianluca Fiorelli: So, how is SEO treating you?
Dixon Jones: Mostly great. Being tool-side gives you a different perspective than being agency-side. Back in the day, when I was running an agency, or what we called a consultancy, I found it really hard working with clients. Well done. Working with technology, I can do that and live with that. As long as people keep coming and liking our technology, I can sleep well at night.
Gianluca Fiorelli: I agree. I couldn’t be on your side because I surely don’t have the skills to create such complex tools to be sold to the public. I can try, experiment, and create my own little tools for my own use, but nothing that could be expanded to the public.
But I mean, yes, I, as a consultant, and all the consultants looking, watching, and listening to us, surely we have to face clients. On the other hand, your client is not just playing with tools, yours, because you actually have to satisfy the needs of a client and explain why your tool is so good.
Dixon Jones: And that’s sometimes a very difficult thing. Trying to differentiate Waikay, or InLinks, or even Majestic from the other tools in the field. I know why it’s different. Trying to tell somebody, especially with this new AI world now, where millions of people are building AI tools just by going to Claude and saying, "Build me a prompt tracker." Trying to show why you’re different, breaking through the noise, is just as hard as building a brand.
Gianluca Fiorelli: Yes, we will talk about Waikay and how it is different from the others. Personally, I see Waikay not as in opposition with other tools, for instance, what our host, Advanced Web Ranking, is doing, but as a needed complement to this kind of AI brand visibility tool in order to have a better, more nuanced perspective about visibility.
Advanced Web Ranking is a professional SEO tracking platform, built for experienced SEOs who need precision data in the current era of redefined "search visibility".
On the traditional side: live rank tracking (not cached data) across Google, Bing, Baidu, YouTube, Amazon, and more, with localization down to zip code and device-level splits for desktop and mobile.
On the AI side: dedicated tracking for Google AI Overview citations, Google AI Mode, LLM rankings inside ChatGPT and Perplexity, and AI Brand Visibility - how models like Gemini, Claude, and ChatGPT actually represent your brand when no one is asking about you by name.
The full search spectrum, in a single platform.
If you want one place to track where you stand across all of it, try AWR free for 7 days.
The Genesis of Entity SEO: Being Scared of Bill Slawski
Gianluca Fiorelli: But before that, I wanted to ask you something. You have always been obsessed with entities and semantics. I remember one of the biggest differentiations of Majestic was not only how many backlinks from domains, how many backlinks from IP ranges, and how many backlinks to specific pages there were, but that you were already, back in the day, giving us the topical, the thematic range of a domain, maybe when not too many people were understanding why Majestic was giving us this data.
Dixon Jones: Yes.
Gianluca Fiorelli: It was the time when we started, when people, especially after Panda, then after Penguin, started to say, "Okay, we need backlinks from thematically consistent external domains." Not from, let’s say, I am a pharma website receiving a link from a totally different type of topic.
So what made you put the focus so early on entity?
Dixon Jones: I’ll tell you what it was. It was an interesting journey, really. It wasn’t so much Panda, although clearly that was a big thing. What happened was when Google bought Metaweb, or Freebase as some of us know it, back in 2014, or something like that, or less. Well, whenever it was.
Gianluca Fiorelli: It was before the Knowledge Graph.
Dixon Jones: Yes. So, Freebase was the basis of the knowledge.
Gianluca Fiorelli: 2011-2012.
Dixon Jones: Yes. So that year, I was going to Pubcon, and that was going to be my topic. I knew that Bill Slawski was going to be either on the stage with me or in the audience. We used to talk, and I didn’t want to make an idiot of myself in front of Bill Slawski, because he was the brightest guy I knew.
So I was really trying to get my head around why Google would buy this massive database of ideas and things and trying to work it all through. They spent so much money on it. It was an undisclosed amount, but it was a huge amount of money, and Metaweb hadn’t even gone to market. They didn’t have a payable product.
That’s how I started to think about things and look at things. I remember doing this presentation at Pubcon about a little village in the desert of Vegas and seeing how the KGIDs, the little sort of KGID parameters, were in the Google URLs and things. And then that really started to get me thinking that, of course, a link from an accountancy website to a farming website is no help to man or beast when it comes to analyzing it from a search engine’s perspective. Context was king.
And so we created all the topical Trust Flow with Majestic. And then, as my time at Majestic drew to a natural conclusion, I felt I’d done what I had to do there, although I’m still doing a thing with Majestic tonight up in Leeds. But I took a bit of time off and wrote the Entity SEO book.
Then a guy from France, Fred Laurent, pretty much phoned me up and said, “Hey, I’ve got this internal linking tool, and it matches your book really.” So we kind of moved into that field with InLinks. So that’s kind of how it worked.
I blame Bill Slawski and the fact that I was scared of him.
Gianluca Fiorelli: We all miss Bill Slawski very much. He was a good, grumpy guy sometimes, but he surely was bringing us clarity. There are a few errors in his field, but still nobody has been equal to him in the clarity with which he presented things.
So let’s say you already started with these things with the topical flow.
Dixon Jones: Yes.
Gianluca Fiorelli: Topical authority flow. And then somehow, by case, you decided, you wrote this entity. Serendipity and, let’s say, this case of fate brought you to work on InLinks.
Dixon Jones: Yes.
Gianluca Fiorelli: InLinks was already very strongly related to entities.
Dixon Jones: Yes. But it hadn’t launched. So when Fred and I got together, Fred had been trying to work on his idea for some time, but he didn’t have a product to put to market. He had completely run out of money, actually. He had no money; he had no investment.
So I took some investment from my Majestic days, I suppose, and put it into what became InLinks. We launched it together.
Digital DNA: Internal Linking as a Knowledge Graph
Gianluca Fiorelli: Yes, but I meant that the concept of using internal links not only for the classic better, stronger page equity, link equity, redistribution, and everything, but a correct link equity redistribution between pages that are substantially about the same topics or subtopics. So, underlying entities and related entities.
So this is substantially what, at the end of the day, for instance, in my case, when I talk a lot about how we should create topical content hubs, is very related to what a tool like InLinks is facilitating.
Dixon Jones: Yes.
Gianluca Fiorelli: But this also means that, obviously, you were making the example of your talk in this little village close to…
Dixon Jones: Vegas.
Gianluca Fiorelli: …with a Knowledge Graph ID in the URL. So, substantially, links, even if not so many were paying attention at the time, internal linking is also the best way, along with other things like structured data and so on, to create an internal knowledge graph.
Dixon Jones: Yes. It turns out to be really effective. And it makes sense when you sit back and look at any business. It has a fingerprint. If you’ve got two barbers on both sides of the street, one’s a Turkish barber, and one’s a traditional hairdresser, and they have different DNA. Every person, every business has a different DNA.
If you think about your business as something with its own DNA, and then you think about your website representing that DNA, you start to see your business as unique to every other business. That’s where the whole entity idea becomes really valuable, because you can say, "This website is about these things, and this webpage is about these things."
You can define that. We define it with Wikipedia URLs, with a very specific reason why we do that. You can say, "This page is about hairdressing and Barton-le-Clay—that's my next village along—and Turkey or something.” That pretty much defines that business as a Turkish barber in Barton-le-Clay. And that’s done it.
So you’ve taken thousands of words and reduced them to three concepts that all appear in Wikipedia, which are database items. Then you’ve given a fingerprint for that content.
Now you don’t need a thousand words to look something up. You just need to go through as a search tool and say, "This person has looked for a hairdresser near Bedfordshire. And it’s like, “Okay, here are all the towns in Bedfordshire. There’s Barton-le-Clay there.”
It can do it, and it’s doing it with math. It becomes very easy to see that whole vector idea, where concepts are related to each other, both within your website and then, for a user, in the wider context of what they’re searching for. So the math of connecting these ideas becomes logical to a human, and we hope logical to a machine as well.
What SEOs Should Own in the Brand Conversation
Gianluca Fiorelli: Yes, because many times, for instance, before coming to Waikay, but it is somehow related. I don’t know if you agree with me.
Many times, when it comes to SEO, for instance, and the increased importance of brand and branding as a signal, many times what I see is—and I can understand it for digital PR—but in the more restricted landmarks of SEO, sometimes I see SEO trying to mimic or to do what others are already doing instead of concentrating on what the real contribution of SEO is. Because of its technical knowledge, it can bring to the relationship between the entities presented on the website. and eventually also outside the website and the brand.
So connecting, translating the signals that are present in the content, for instance, into machine-readable signals.
Dixon Jones: Yes.
Gianluca Fiorelli: A classic example is how well the schema organization is implemented. Many times, people forget and put Schema Organization on the homepage and then forget to link the idea attribution to the product, to the collection pages or to the person working with the organization.
So don’t you think that SEO should relearn how to do everything like this, as I just described, instead of trying to steal the job of brand marketers, for instance?
Dixon Jones: Maybe I’m playing devil’s advocate a little bit, just for the conversation. But I think there’s an interesting thing. Certainly, you should consider your brand in context with your business. Just increasing brand awareness out there is not something SEOs should be doing. It should be honing in, as you say, on the messages, on the topics, and on the entities that are important to the business.
Now I have an interesting double-take with schema. Of course, InLinks uses schema to anchor an entity to a page. So, we sit there, and a human in the loop says, "Yes, this page is about Barton-le-Clay and hairdressing. That’s what these things are about.” So the machine can then go in and find all the times you talk about Barton-le-Clay and hairdressing on the website and link through to the page in context. It uses schema as a way to define the internal links.
Interestingly, though, I have no proof that the schema we use is used in a search engine algorithm. What I do have proof of is that you don’t want ten pages about Barton-le-Clay or hairdressing. You want one, and you want that one. And any other mentions of things need to decide what to defer to the authority page for that topic or concept.
So internal linking is the end product of making sure that any way you look at the system, whether there’s a human or a machine, you will find the right authority page for your business on that site. Even if you have a hundred pages all talking about the same idea, it’s obvious where that authority is.
So it reduces cannibalization, and it’s like “all roads will lead to Rome.” I sometimes think of links as roads, really. And so, I use schema because I need it to build the internal linking system that we do. And it forces us as humans, when we look at that schema, to be able to see things. I can see it, and I can understand it and say, "Right, this is a page on this subject; this is the page on this subject in the context of this subject,” and this kind of idea.
So the two map. Whether the algorithm actually uses the schema turns out to be secondary to whether you’ve thought through the schema as a human. Schema preceded computers. It came from the idea of categorizing the web. You know, the libraries when we were kids, with all those little numbers and things. That was schema. That was the categorization of all the books in the library into logical topic areas.
That’s exactly what kind of schema is trying to do. And you’re absolutely right, people put schema on the homepage of a website and say, this is about this brand and this kind of webpage, and this is the address of the company. And then they completely forget about the fact that they’ve got authority in this subject over here and this subject over here. So it’s a bit of both, I think. We need to understand it as humans, and the fact that schema lets the machines understand it as well makes our job really unique.
Schema as a Thinking Tool, Not Just a Technical Signal
Gianluca Fiorelli: Yes, and I don’t want to enter into the polemics of whether LLMs are using schema or reading schema. But what I really like about schema and that’s why I insist a lot on implementing it. Apart from the fact that it is a way to create something bigger than the schema itself, combining with OWL, so with the triples you can really create a knowledge graph starting from the schema and so on.
But especially because it obliges us, whenever we need to create content or a specific type of content, let’s say a product page, if we look into the schema for products and all related schema like offers, aggregate rating, and so on, it gives us a sort of script draft of how the page should be structured.
Dixon Jones: Yes.
Gianluca Fiorelli: So it’s a way to understand what elements on the page should be present so that probably, the machine, search engines, or even LLMs after, need to understand that a product is a product. That’s why for instance, Amazon doesn’t use schema for the product pages. But they structure them in a way that substantially replicates what the schema is clearly suggesting.
Dixon Jones: Yes, I think that’s absolutely fair. Andrea, your countryman Andrea Volpini, commented on a thread with me the other day, and also Mark van Berkle, Martha van Berkle’s brother, or son, or dad—no, a brother, I think. Anyway, he was also on the thread, and a couple of interesting things came out.
Andrea mentioned the idea of—we were talking about LLMs, but the same idea applies—an introverted machine and an extroverted machine.
Check out the conversation between Gianluca Fiorelli and Andrea Volpini on the Search Session!
They discuss semantic search, structured data, and the rise of AI agents. The conversation explores how these changes are reshaping machine understanding of content and the future of SEO
Dixon Jones: And the idea of an introverted machine was trying to find all the entities within an ecosystem, within your business, within Amazon, within a Turkish barber, or whatever. And the extrovert one was trying to talk to the outside world or find things that are outside of your own ecosystem.
And Mark van Berkle came in and said that the interesting thing is when you can use open-source ideas in schema, things that everybody can understand, then those two things will work.
Inside a business, maybe your products are blue screws, green screws, and red screws. These are all entities. Red is an entity, blue is an entity, green is an entity, and screws are entities, and then you can make up the idea. But when you tell that to the outside world, in the shop, you’ve got SKU codes and stuff to describe all the products, but that’s just for yourself.
But when you can translate that to the outside world, so that when somebody says, “I’d like some screws to put into my red door,” then they can talk to each other. That’s when you need to use a common language. Schema gives that common language, or it can, but it does depend a little on how you use it.
That’s why we use Wikipedia. Because most SEOs, when they want to make their brand well known, the first thing they try to do is create a Wikipedia page for their brand and then get banned from Wikipedia. And then they learn that and move on.
But the thing is, the fact that Wikipedia is so hard to get a page on shows that the articles on Wikipedia have independent data sources.
You could use other ones. You could use Crunchbase, or IMDb for people or personalities, or whatever. But using an independent data source is much better than an SEO creating an entity for their client or creating things out of the blue for their client, because then they are shouting about things that nobody else understands.
The trick is to say, "All the things I do in this business—maybe the business is unique—but I need to describe these things in ways that the rest of the world can understand." Using something like Wikipedia, addresses, or postcodes, depending on what kind of schema you are using, allows that connection between machines.
Gianluca Fiorelli: Yes, and I think that, and I don't know if you agree with me, apart from the classic Wikipedia, apart from all these other first-tier and second-tier sources for building the Knowledge Graph, there is another one. People always say Wikipedia, but Wikidata is sometimes even more important, because that is where all the connections are put and from where Google immediately sees a trusted map of connections about an entity.
Measuring What AI Knows About You
Gianluca Fiorelli: From there came this revolution; let’s call it a revolution. It was already coming, but maybe nobody was really thinking about the post-COVID situation and how e-commerce was boosting, and then came OpenAI with ChatGPT and all the others.
Google did it first with Bard, which was horrible, and now with Gemini, which is better than ChatGPT, but also with Claude, DeepSeek, and so on. And you came out with this idea of building visibility, substantially a visibility tool, Waikay. For the people who still wonder what Waikay means, it is...
Dixon Jones: … what AI knows about you. You can’t pronounce it, you can’t spell it, and we don’t own the .com. It turns out that uniqueness becomes really useful when trying to be visible in AI.
Gianluca Fiorelli: Well, I didn't know that you didn't have the .com domain but I think that waikay.io sounds cool. And I also think it’s very "parlante," as we say in Italian, which means a very “speaking” name, because it clearly defines what it wants to do.
For the people listening and watching us who haven’t used it, the first thing you’re asked when creating a project is, "Tell me what your website is, tell me a couple of your competitors, and I’ll tell you what the LLMs know about you," because you will be asked, "What do you know about this brand?"
Dixon Jones: Yes, exactly.
Gianluca Fiorelli: And I like that, because it gives you a first screenshot of visibility. I like the idea of visibility, also because of all the experience you have with entity search.
You substantially scrape the website, analyze the website individually, and maybe also retrieve data that you already have from InLinks. So you see what the website is about. You can map the topics, and then people can add more topics. And then you can start with brand plus topic, asking the LLMs.
Dixon Jones: Exactly.
Gianluca Fiorelli: And doing so, people can see… for instance, I’m really passionate about mini painting, let’s say Warhammer, and contrast paint. And then you can see how much Warhammer is visible for contrast paint against, I don’t know, Army Painter or AK, which are two other brands producing paints.
And this is very interesting because, for me, it’s better than getting crazy and trying to figure things out using thousands of query fan-outs and made-up conversational queries to substantially do the same thing.
Dixon Jones: Yes.
Gianluca Fiorelli: Because I don’t think people do that just for tracking and seeing, “Okay, I’m visible for this specific conversational query.”
Dixon Jones: Well, I think they do, but I don’t think they should.
Gianluca Fiorelli: Yes, I think the most advanced users of these prompt tracking tools need a lot of information and a lot of queries to do substantially what Waikay does, which is to understand an overall indicative metric about visibility.
Dixon Jones: Thank you.
For teams who also want a more continuous, reportable layer of AI brand data sitting alongside their traditional rank tracking, AWR's AI Brand Visibility feature covers some adjacent ground.
It queries ChatGPT, Gemini, Claude, and Perplexity directly to surface three specific signals:
which topics LLMs associate with your brand,
the sentiment colouring those conversations,
and which other brands are being mentioned in the same breath as yours.
That last signal - co-mention mapping - can be particularly useful for understanding the competitive frame AI is placing you in, even when no one has asked a direct comparison question.
Think of it less as a replacement for the kind of strategic visibility analysis Gianluca and Dixon are describing, and more as an always-on measurement instrument for the same territory. Try it free with an Advanced Web Ranking trial.
Visibility Gap Analysis and Prompt Tracking
Gianluca Fiorelli: Then yes, there are a few things that can be done with prompt tracking. For instance, some specific prompt that you can use as your canary in the mine. If, for that prompt you were always visible and all of a sudden you disappear from being cited or mentioned, you can say, “Okay, let’s go and check what’s happening.”
Dixon Jones: We do prompt tracking. When we launched the product, we didn’t include prompt tracking because we knew that the variation of different ideas… If you ask something like “What’s the best hotel chain?”, it will come back with different answers every single time. And it will say, depending on what you want.
But now we do prompt tracking. It’s kind of like reporting, like accounting numbers after the fact. It’s like your management accounts a month later. It’s not telling you what to do. It’s not telling you how to be proactive to make your brand more visible in AI.
What we want to do is say, "Does the AI properly understand the business in context?" Because if it doesn’t, then everything else becomes impossible. It’s like the old days of getting into Google. First you get into the index, and then you can rank. If you don’t get into the index, you’re not going to rank.
So context becomes king. It turns out that this is really hard for other tools to do because they can’t build this knowledge graph. What we’ve taken from InLinks is the same foundational technology of reading a website, or the main pages of a website—you don’t have to read them all—getting a fingerprint of what that business does, and then getting a fingerprint of what the AI says that business does.
Then we compare the differences mathematically. That’s how we create those action plans that say, "These are the things this specific business needs to do to increase their visibility in the context of this topic." And it’s working well.
Gianluca Fiorelli: Now, in fact, I think this is maybe what I like most as a user of Waikay. You ask for two competitors immediately. It’s not just for competitive analysis; it’s more for gap analysis.
Dixon Jones: Yes, it is. We get criticized for not being able to show ten. People want more, but actually, you’re right. Sorry, carry on.
Gianluca Fiorelli: What I would say, and maybe if I have to give a tip to people using Waikay, is, “Remember, guys, that for each topic, maybe your competitors are not the same ones that you put the first time.”
Dixon Jones: And you can change them now. And as of a few months ago—well, last month, I think—you can now change the competitors for every topic. We’ve just put that in there.
Gianluca Fiorelli: Yes, visibility gap analysis on LLMs, I think it’s very good, because it tells you where you have to improve, what you have to defend, and how much you have to improve. I also think you did a very honest thing, because the action plan can be assessed.
Dixon Jones: Yes.
Gianluca Fiorelli: I always do that, because for certain clients I know them very well, and I can see that maybe I misunderstood something. But sometimes you are working with new clients and you don’t have this kind of strong knowledge. So it’s always very good to do it.
Dixon Jones: For the audience, what we do is that at the end of these reports, there is a button you can press. You can copy the report with a short prompt and then put it into Gemini, Claude, ChatGPT, or whatever, and it will rate the report out of 10.
By the way, the second thing you can do after that, Gianluca, which isn’t documented, is you can then ask how much this report is worth to a client. Then it comes out with a magic number.
Gianluca Fiorelli: Okay. But actually, what I do is, for instance, when I’m working on creating topic content hubs, one of the source materials I give to my super specific multitask prompt on Claude or Gemini is both the action plan of Waikay with the assessment.
This is a way for me to give the AI a sort of first draft of what could be done and how to orient it. Then I use many other signals, like scraping AI answers for some specific conversations, essentially more entity-focused conversations. For instance, to see what entities AI is already using. Because they are the paramount we have to consider for the consensus.
Then we can use these entity expressions from the AI answers to understand: this is the consensus, and this is where we can obtain the information gain about this topic.
Dixon Jones: Yes.
Gianluca Fiorelli: One question that I have regarding Waikay. For instance, you now offer the possibility to do some specific prompt tracking, but how can you eventually use, in combination, the more strategic vision, the visibility report, and the action plan, along with prompt tracking? How can you use them together or how can you eventually use them together?
Dixon Jones: Where we are working towards is this: the topic plans and the action plans push your content forward. They help you understand what content you need to write or how to connect the dots on your website, create hubs, and that sort of thing.
Then you use the prompt tracking to check your progress. One of the things we are doing in prompt tracking is using query fan-out. In the individual prompts you can see the Gemini query fan-out in the prompt tracking. We analyze those, and we analyze the responses.
When you ask something like “best hotel chain,” these are the kinds of prompts you should use. You don’t say, “Is Marriott the best hotel chain?” That’s a silly question to ask an LLM, because of course, Marriott will appear in the response.
Instead, for marketing purposes, you ask something like, “What’s the best hotel chain to take my kids on holiday in New York?” Then you want your brand to appear in the answer. You don’t ask for your brand; you see which brands come back.
So we conceal all the brands automatically. It doesn’t track just one brand; it tracks any brand that appears in the results. Then, in those responses, we also build a knowledge graph of the topics or entities that are mentioned.
So now, as the prompt tracking comes back, we start seeing that Marriott comes back with the concept of luxury, whereas Travelodge comes back with the concept of budget. So you start to see all the underlying entities in the context of these brands.
Now you begin to see that if Marriott wants to compete with Travelodge, maybe they need to start talking about budget more. That sort of thing.
By taking those entities, we build a way to analyze not just the prompts, not just where the brand appears in the prompts—which is what you give to the marketing director or the C-suite.
Gianluca Fiorelli: But the entity attribution.
Dixon Jones: Exactly. It gives you clues about what you need to do next in a longer strategic process, which then goes back into new topic reports.
Gianluca Fiorelli: So it is substantially like a deep dive into what the vector plan is.
Dixon Jones: Yes, exactly. We are trying to say: at the start, you know you need to be seen for luxury hotels. So you need to be seen around the context of luxury, for example. And it’s only after you’ve done that and you’ve kind of gone and then see, “How come this other brand is still ahead of us? Well, they’re focusing on these areas." And you can then feed that back into new topics. So it does become a loop.
But the work we haven’t talked about in between all of that, of course, is that the machines get it wrong. So fact-checking becomes important. Not all the time, but the ability for the tool to display everything in one line tells you the facts that are coming back.
When AI Gets It Wrong: The Hidden Risk of Bad Training Data
Gianluca Fiorelli: It’s present when it’s relatively easy to find out the factual part in Waikay. It should not be overlooked, because it happened to me with a client. All of a sudden, not all LLMs, but Gemini was giving me a fact in the training data, not even grounded data, about this brand, but totally unrelated.
This led me to ask the right question to my client, asking what happened back in the day. “Did you receive some sort of malware attack?” Because this kind of information is still there, even if you had the ability to hide it in Google Classic Search, but we know that LLMs surface everything.
Dixon Jones: Yes. The thing is, if you are misrepresented in the training data, then it has the real possibility of ballooning out into a falsehood in the real world at the far end, doesn’t it? There are a lot of cases like that. Once the AI gets it right, it’s great. But if it gets something wrong in the training data and you don’t fix it, it’s just going to fester. It’s going to be like a sore that you don’t treat.
Then all of a sudden, in five years’ time, there’s going to be a story about your brand, a narrative that you never wanted and never asked for. It’s just born out of mathematics, and it happens in real life anyway.
There are countless times when we as humans go through the world, hear something from someone, and repeat it. Then it becomes a falsehood that we think is correct. The number of times in history that has happened… we started killing witches because they must be bad. Because you can’t drown them. That’s the kind of thing that happens when facts go wrong.
The Maybelline Pattern: Why Even Big Brands Fail at Topical Hubs
Gianluca Fiorelli: Yes. Last question, because I know you have to go. One last question about all these things. You started to present on LinkedIn some case studies.
Dixon Jones: Yes.
Gianluca Fiorelli: I like them because you take a brand and, using Waikay, you explain what is happening to that brand.
Dixon Jones: Oh, you like that? I’m glad. That’s great, thank you!
Gianluca Fiorelli: One of the most recent ones is the case of Maybelline. That was particularly interesting for me because I collaborated with a competitor of Maybelline, an Italian brand. So I had a special interest in that case.
Obviously, you write about certain cases, but surely you have seen many more. What kind of patterns do you see in terms of mistakes that even big brands make? Maybe they are over-relying on their brand visibility and recognition.
Dixon Jones: I don’t know if this is a mistake or whether it’s a foible of our technology, but creating hubs seems to be something that we never really got around to doing. It’s like we say, WordPress creates tag hubs or category hubs, so we don’t need to worry about it.
But we weren’t creating proper information hubs about the topics that we actually want to rank for. I see it even in the biggest companies. I did one yesterday on Maybelline and mascara, and it turned out that they hadn't gotten it. Of course, they have a product page about mascara, because that’s what they sell. But they didn’t have a proper page about mascara as a topic, and becoming an authority on mascara.
Gianluca Fiorelli: Well, I think that more than not having pages, because this was the problem of this client of mine, the competitor of Maybelline, it is that they have the pages, but they are scattered, totally isolated, and not connected.
Dixon Jones: And that’s what I mean by a hub. Clearly, Maybelline is an authority in mascara or it should be. Maybe they are an authority in marketing; I don’t know.
But it’s all of us, really. The more you are an authority on something, the more you talk about it in different places and different contexts. And the less you spend time concentrating on that one thing and showing the world that this is the authority place for someone to go to see all of that.
So I think the hub idea is probably the simplest thing that people miss.
Internal Linking Is Still the Most Underrated Weapon in SEO
Gianluca Fiorelli: And that’s why, returning back to InLinks, internal linking—apart from the classic technical SEO hygiene—is maybe still showing that it is the most important weapon SEOs should learn to use.
Dixon Jones: It’s certainly an important weapon they should learn to use. I think it gets underrated consistently.
Gianluca Fiorelli: There are many people talking about it, and it is strange because it is not a topic that nobody is talking about. It’s a recurring topic. For instance, when people say these are the things you have to do for a migration, for instance. There are twenty years of recurring topics and one of them is internal linking, but maybe people are not paying as much attention as they should.
Dixon Jones: I think too many people just think about menus. But menus are just navigation.
Gianluca Fiorelli: Yes, they confuse internal linking without understanding the nuance between navigation and architecture.
Dixon Jones: Exactly. And sites like Search Engine Journal, for example, have very strong internal linking. I’m quite impressed by their internal linking. They don’t use InLinks, by the way. They should. But their internal linking is very strong. But they're very unusual.
The number of even many news websites that are not using internal linking properly is criminal. The BBC should be using internal linking. They actually used to, back in 2003 or 2004. I’m really old. Back then they had a website where internal linking was already working very well, and they were at the top of the search engines for everything.
Life moved on for them for sure, but they should get back to an internal linking algorithm.
One reason internal linking stays underrated is that its effects are hard to attribute - you restructure, you interlink, and then you wait.
Advanced Web Ranking makes that feedback loop tighter:
Its Pixel Position & Trend metric tracks not just rank position but where results physically appear on the SERP - above or below the fold - so you can see whether a consolidated hub is genuinely pulling authority pages into more visible territory.
Click Share and Estimated Visits then translate those position changes into traffic impact estimates, giving you something concrete to put in front of a client.
If you're doing the structural work Dixon and Gianluca are describing, you need a measurement layer that's precise enough to show it working. AWR's 7-day free trial is a good place to start.
Gianluca Fiorelli: Let’s stop talking about these things. Maybe just one last question. Are you already thinking about other features? Is there something you are testing in beta for Waikay or InLinks? Or something you would like to build next, when all the things work fine?
Dixon Jones: Absolutely. I can’t tell you too much about what they are, though. But we’re not stopping development. We’ve been launching things pretty much every two to three weeks. Sometimes, small things, for example, the ability to change competitors for every topic every time you run a report. That’s actually a big change. There are also bigger things coming, but I can’t tell you about the big one yet.
Gianluca Fiorelli: Okay, we’ll live to see it.
Advice for the Next Generation of Tool Builders
Gianluca Fiorelli: So, stopping talking about it and just talking about you, I don’t want to make you sound like an old sort of…
Dixon Jones: That’s all right. I’m in my sixties, Gianluca. It’s okay.
Gianluca Fiorelli: Yes, but I don’t want to make it sound like you’re an old Gandalf-type SEO guru. When you look at all these younger generations creating tools, what is the mistake you made that you would recommend they not repeat?
Dixon Jones: Some of them are brilliant. Some of these kids are great. But if you ask me to name a mistake, I think there is a lot of value in proving yourself as an individual first before you dive into building a tool.
If you build a tool in an industry where you are not known, your chances of getting heard and seen are quite small. Waikay, InLinks, and even Majestic still benefit from the fact that I have been around for a long time.
Now, you don’t have to be around for a long time, but you do have to demonstrate that you are respected in what you are trying to build and sell. If you can’t do that yourself, then you need to team up with someone who can.
That’s what happened with me and Fred, and also with me and the guys at Majestic. Often, the developer doesn’t want to be that front person. But if you don’t have that front person, then you’re in danger.
It also helps to have two people challenging each other. You become a stronger business because you don’t make mistakes, because you have someone checking your sanity.
Gianluca Fiorelli: Okay, thank you. I’ll let you go, and thank you very much for this conversation.
Dixon Jones: I enjoyed it, as always.
Gianluca Fiorelli: Dear friends listening and watching, let me do a YouTube influencer thing: remember to subscribe to the channel and click the bell so you’ll be notified about new episodes of The Search Session. Thank you and bye-bye.
Podcast Host
Gianluca Fiorelli
With almost 20 years of experience in web marketing, Gianluca Fiorelli is a Strategic and International SEO Consultant who helps businesses improve their visibility and performance on organic search. Gianluca collaborated with clients from various industries and regions, such as Glassdoor, Idealista, Rastreator.com, Outsystems, Chess.com, SIXT Ride, Vegetables by Bayer, Visit California, Gamepix, James Edition and many others.
A very active member of the SEO community, Gianluca daily shares his insights and best practices on SEO, content, Search marketing strategy and the evolution of Search on social media channels such as X, Bluesky and LinkedIn and through the blog on his website: IloveSEO.net.
stay in the loop