Bridging Information Retrieval and SEO Practice | Dawn Anderson
Hi! I’m Gianluca Fiorelli, and in this episode of The Search Session, I’m joined by Dawn Anderson, international SEO Consultant, speaker, and trainer. Together, we take a thoughtful dive into the realities and misconceptions shaping SEO in the AI era—a practical walk through what’s changing in search and what has never changed at all.
Here are the main topics we explore:
Myths and premature speculation in SEO: new or misunderstood concepts like “llms.txt”, "chunking", or "BlockRank", and the potential negative consequences.
SEO vs. Academic Information Retrieval: a lack of formal standards and peer-reviewed research fuels misunderstanding between SEO and academia.
GEO, AEO, and LLMO clarified: these trends aren’t revolutions; they’re just new layers or specializations within SEO, reflecting evolving tools and search behaviors, not entirely new disciplines.
The timeless foundation of search: semantics, linguistics, structure, and user intent remain essential, even as technologies like AI and LLMs evolve.
Structured data’s value: it doesn’t directly influence rankings, but it improves search visibility by helping search engines better understand and disambiguate content, which can indirectly boost rankings and increase inclusion in AI-driven results.
Dawn’s practical use of AI in daily work: AI is a valuable assistant for ideation, coding, and learning—but it shouldn't take over the learning. Human understanding and creativity remain essential.
Changing search behavior: the rise of zero-click isn’t decay but refinement, pointing to deeper, more complex search sessions.
Agentic search evolution: Agentic Search is emerging as a new layer in search, but challenges like ecosystem readiness and measurement accuracy still stand in the way.
There’s plenty to learn in this conversation—so let’s get started.
Video Chapters
Transcript
Gianluca Fiorelli: Hi, I am Gianluca Fiorelli, and welcome back to The Search Session. Today, we have a true expert as our guest. This person actually digs deep and studies all the topics that are constantly being discussed by many, sometimes without much precise knowledge.
In fact, just looking at her LinkedIn profile feels like reading a poem filled with acronyms. She holds an MSc DigM, PgDip DigM. Well, in plain words, it means she’s truly studied and earned advanced degrees in things like information retrieval, digital marketing—all the things we talk about every day on social media, with clients, and so on.
She’s an international SEO consultant and the managing director of Bertey, a small consultancy agency. She lives in Manchester and loves to travel the world. This person is Dawn Anderson. Hi Dawn, how are you doing?
Dawn Anderson: Hi! I’m good, thank you—and thank you for the kind invitation, Gianluca. I’ve really enjoyed watching some of The Search Session conversations before now, and I was really pleased when you invited me, so thank you.
Gianluca Fiorelli: You’re welcome—and thank you for your kind words. So, let’s start with, you know, the question we always begin our conversations with—the already classic question: How is SEO treating you lately?
Dawn Anderson: Oh, well—SEO is always wonderful. Even though it’s died a million deaths, it never actually dies. And, you know, while people are still searching, I believe there’ll always be a place for SEO.
It’s always interesting—we never have boring careers. It’s just fascinating. I’m always so thankful that I stumbled across it 17 years ago and started my journey in the world of SEO. I’m grateful for it every day. I enjoy it all the time—even now, during this turbulent learning period where we’re all having to, like, re-evaluate everything, and just dig in and get back on a very steep learning curve.
Probably the steepest learning curve of all. In the 17 years I’ve been in search, this is likely the time when it’s changed the most—even though we’ve seen a lot of changes before. But this is probably the biggest change. And yes, there’s a bit of confusion all around. But it’s interesting, it keeps us getting up every morning.
Gianluca Fiorelli: Yes, indeed. It really is a steep learning curve. But I like the positive vision you have, because sometimes other guests in the past have responded to the same question with more worried or uncertain answers about the future. So it’s always nice to see someone like you presenting, despite all the difficulties and the steep learning curve to absorb all the new things happening in our sector, with a positive attitude.
Dawn Anderson: Yes.
LLMS.txt and the Problem with Unofficial Protocols
Gianluca Fiorelli: Talking about the eruption of AI in search. On social media, especially, you are quite vocal, with your own style and tone of voice, about—well, against—the myths that are growing very fast and being widely spread around everything AI and search. So I have two questions: What is the most ridiculous myth you’ve read about? And what do you think is the danger of that myth?
Dawn Anderson: Well, I think a lot of it comes from—I don’t… I’m fairly confident that people are not going out of their way to create myths, as such. But there’s a lot of confusion, and we’ve already touched on the fact that there are some very, very steep learning curves that we’re all going through.
I think—it’s not ridiculous—I see that people are trying to make a case for ways in which LLMs can consume content. Because obviously, they have challenges around things like JavaScript rendering—not all LLMs are actually able to render as such. So I get that people want to find an alternative to robots.txt, so they're using this notion of llms.txt, where they literally just mark down all the content and pile it into a file.
But I think they’re kind of not getting their heads around the fact that robots.txt, for instance, is a standard. It’s a protocol. Behind it is the Internet Engineering Task Force. So, introducing anything new kind of goes beyond just a few people in the AI and SEO world saying, “Hey, let’s just launch a new protocol.”
So I think that’s one that’s being pushed, unfortunately—and it’s being pushed by some of the tools as well. Because it’s very easy just to make an additional file and say, “Hey, here’s some added value that you can offer to your clients—create them an llms.txt file, and all the AI search engines and interfaces will bring them up in mentions,” and so forth.
So I think the llms.txt one—it’s not ridiculous, but it’s just a very cheap and easy way to almost create a myth, that one.
And interestingly, Mark Williams-Cook - because, you know, he’s quite a character, smart guy—he created a file called cats.txt, which also started to get indexed. And so he said, “Oh, I’m gonna create a protocol for cats.txt.” And, you know, I’ve got a Pomeranian dog—so someone else created one called pomies.txt, and I was like, “Well, I’ll have this one then.”
So it’s things like that, where people are very enthusiastic about AI—which is good, in a way. We’re all having to learn. But at the same time, we’re forgetting that a lot of these protocols, and the things that have come before—like requests for consideration, the RFCs of the Internet Engineering Task Force—some of these protocols take years to be accepted.
So I think it’s going to take more than just a few excitable SEOs to get a new protocol for crawling, as such. So there’s that one.
And then obviously there’s the whole chunking one as well. I think, again, people are confusing chunking—which is, obviously, breaking things up into a small context window, tokenizing content, if you like, and defining where content should be separated and in what size of chunk.
I think people are getting that a bit confused with just doing strong, semantically structured content with nice little sections that are topically focused, like semantic structuring.
So I think there’s that. Again, as I said—not necessarily myths. "Myth" is just an easy way to classify them, but they’re more, I think, just confusions.
BlockRank and Misinterpreting Academic Papers
Dawn Anderson: There’s another one at the minute that’s just popped up, which is called BlockRank, which I think, again, people go off …There’s always an academic paper behind a lot of the things that Google announces on their Google Research section. So I think people go off, and I don’t think they fully read the academic papers. Because they’re not that easy to interpret, to be honest.
They have lots of metrics for evaluation. Like, if it's regression stuff, they'll have lots of metrics around mean square error, R² score, or precision at X—or, you know, lots and lots of things. So, unless you know what you're reading, it's very easy just to misinterpret it.
I know that’s a very long answer, so I hope that answers that.
Gianluca Fiorelli: Obviously, it answers everything very well. And when I use the word myth, I mean it in the sense of a classic, evergreen myth in the SEO space. Like a real classic. It’s like the idea of a myth in SEO—like keyword density, which maybe made sense in the very beginning…
Dawn Anderson: Yes.
Gianluca Fiorelli: …but then it was just so, so, so repeated—repeated and repeated—until it became something that now, obviously, not so many people still use. At least not the terminology, I think. I believe that even Yoast SEO removed keyword density from their tool.
Dawn Anderson: Yes.
Gianluca Fiorelli: So I was using “myth” maybe in that sense—like, things that are created as a kind of stepping stone when you're learning, but then become calcified into something you're supposed to do, because we don’t know how things work. There’s no justification for doing them.
So, for instance, in the case of llms.txt, people say, “Okay, maybe it’s not needed at all—but we don’t know. It’s not so difficult to create because I’m not sure if it’s used or not. I’ll just create it.”
Dawn Anderson: May as well have a go with it. And the thing is, I think there have been quite a few studies now that show that, in the main, LLM bots just don’t even bother with it, you know?
I think it was… who was it? I think Chris Green did a study where he had a lot of llms.txt files across many sites, and no bot went near it at all. So I think there are quite a few people now who have actually gone and done some testing with it.
But, you know, I don’t like the fact that some tools have actually included it in their SEO suites, as something you could sell in as a service. Because I think it’s important that we remain credible and legitimate—consistently legitimate.
Because otherwise, if we start saying, “Oh, we’ll do this because LLMs need to have an llms.txt file,” and it’s not true—or we don’t know it to be true—then I think it just makes us less credible. So then, when we do have something big that we want to recommend, that’s also not seen as something that’s needed.
So, I think we should always be authentic, and we should always be credible, and we should always just support everything we’re doing with either data or—I know people don’t always agree, but—if Google said “do it,” then, not necessarily that you have to run off a cliff if they say to do that, but at least there is some evidence to support what you’re suggesting.
If Google had literally said, “This is part of our Webmaster Guidelines,” it’s ok, but they’ve not said anything of the like with llms.txt. So really, it’s just testing. But then, blackhat SEO is also testing—so what are you gonna do, you know?
Chunking, Passage Indexing and Topical Focus
Gianluca Fiorelli: Yes, and when you were talking about chunking—I agree with you. It’s like many have translated chunking into creating content that’s simply made of one paragraph and many, many bullet points, or numbered lists.
Dawn Anderson: I think that kind of content, eventually, will get suppressed when core updates come. Because it’s just so obvious that you’re trying to be manipulative.
Gianluca Fiorelli: Now, part of it is also quite terrible, because yes, we think in terms of optimizing for AI, but then that content must also be read by someone who’s not an AI. And let’s be honest—it’s a horrible formatting.
Dawn Anderson: It’s awful, painful to read. You’d never read a book that had a paragraph with a heading. You would never read a book or anything that had two or three sentences. You just wouldn’t. It just wouldn’t be a good experience for humans. I think it will be really easy to detect that this is over-optimization.
Gianluca Fiorelli: Yes, and I think that people, with this concept of chunking, maybe would just need to think about when you click on an AI Overview, or even a Featured Snippet, and you go to the page that’s been linked.
Google usually includes an anchor link to the specific piece that’s being used. And you can see that sometimes the chunk comprises more from a paragraph—or part of one paragraph, and part of a second paragraph.
Dawn Anderson: They’re able to extract it and then put it all together.
Gianluca Fiorelli: Yes, so, returning to what you were saying, the chunk is about—using an anthropic definition—a piece of content that is clearly mono-semantic about the concept. So it’s more about clarity in what we’re writing, not about dispersion.
Dawn Anderson: Exactly. But for me, content should always have clarity—from top to bottom in any piece. It should have a topical focus. Because otherwise, it’s just… well, it’s not really relevant to anything, is it? When you’ve got something that’s vaguely about this and vaguely about that.
And I mean, to some extent, it kind of goes back to the whole paragraph indexing era—which nobody really talks about anymore, but obviously, it will have had some kind of impact.
So it’s kind of very similar in that way. Do you remember, years ago, when that first came out? Paragraph indexing? And they were talking about how you’d have a section about cats…
Gianluca Fiorelli: Are you meaning passage indexing?
Dawn Anderson: Sorry, passage indexing, yes.
Gianluca Fiorelli: Okay, yes.
Dawn Anderson: Passage indexing. Do you remember? I do believe there was an interview that Martin Splitt did—with Bartosz Góralewicz, I think? And he explained it. He said, “Look, you’ve got bloggers who don’t really have any awareness of SEO. And they don’t necessarily structure things in a way that’s very focused. But passage indexing understands that this paragraph was about cats, and this paragraph was about dogs.”
So I think that’s probably more relevant—to both LLMs and search engines—than this whole notion of “chunking” is, from an SEO perspective anyway. So it really just comes back to having some sort of topical focus in all the writing that you do, which is just… good. Good for search, good for humans.
Staying on course and organizing things well. Making sure your library is organized and you don’t end up with a bit of this and a bit of that all thrown together. Just library science, isn’t it? Like an organized world.
IR and SEO: Frenemies Forever
Gianluca Fiorelli: Yes, and I have a curiosity—it’s maybe even a silly curiosity. Because—let’s say—you have one foot in the SEO world, where so many people are trying to investigate, often without the kind of knowledge that, for instance, you have.
And with the other foot, you’re in the academic world—let’s call it so, more broadly. I’m curious—because not all people, but many in our industry, read patterns, or try to understand them. Many don’t come from information technology backgrounds. Like me, for instance—I don’t. I had to learn by myself, at least the principles of information retrieval.
But many SEOs supposedly do come from information technology, so they should know things like information retrieval. But in the academic world—the people who are always thinking about information retrieval and so on—if they think about SEO, what do they actually think of the SEO world?
Dawn Anderson: Well, I mean—first and foremost—I want to stress that I'm largely a consultant rather than an academic.
Gianluca Fiorelli: Yes, I was just joking about that.
Dawn Anderson: Yes, you're right. I do love learning all about it. And, I think, so many of us in the SEO world must learn every day.
But one thing I would say is that, funnily enough, I did a talk this summer. I went to the Summer School for Information Retrieval. I’ve been there three times now. I was first there in 2017 in Barcelona, then I went last year in Amsterdam, and this year they asked me to do a talk about the changes happening with generative information retrieval, and how SEOs are kind of adapting to it. And I did a talk that was literally entitled: “IR and SEO: Frenemies Forever”
So I think there’s, like, a friendly enemies kind of relationship that goes on between us and the IR world. I think some of them are very suspicious of SEO. But I think the problem is that a lot of people in that field don’t differentiate between what is spam and what is SEO.
And I think even among SEOs themselves—and people outside of SEO, like other general marketers—a lot of them don’t differentiate either.
Between what is good SEO—i.e., understanding people's search behavior, aligning a website with that, building relationships, promoting a website so it becomes more visible in search, building topical relevance, that kind of things—and spam, which is obviously about using dodgy tactics or just trying to manipulate things too much.
So I think there is curiosity about us SEOs from the IR world, I really do. But at the same time, I think there’s just a big misunderstanding of what we actually do. I’d say it’s probably more of a case that they’re just not quite sure what we do. And I think sometimes we, as an industry—as SEOs—we don’t help our case with some of the things we talk about.
Defining SEO as a Profession
Gianluca Fiorelli: Yes, I think it’s an original sin for us, SEOs. In very few cases, we usually weren’t—or still aren’t—able to explain correctly, and in a way that everybody can understand—from the academic to the layperson—what we actually do.
Dawn Anderson: The definition of what we do is just very difficult to pin down, isn’t it? Yes, it’s difficult to pin down. It’s very easy to confuse what we do with spam sometimes.
There’s a low barrier to entry, which is, in many ways, a very good thing.
People don’t have to have qualifications to become an SEO. They can have a really interesting and fulfilling career without any academic background, which is a good thing.
But at the same time, that also means we don’t really have standards, or a framework, or any kind of formal side to what we do. So, in some ways, it’s a bit of a wild west.
And obviously, we’ve always got new people coming in, so it really depends on who they learned from, and so forth.
Gianluca Fiorelli: Yes, but I don’t know if you agree with me—but maybe after 20 years, at minimum, if not more, of the system of SEO, we should seriously start to think about—not necessarily standardization in a very strict sense, like ISO or something like that—but at least about setting up some kind of clear definition.
A definition that’s still open to evolve—because our field is always evolving—but at least a clear definition that could be accepted across the whole industry.
Dawn Anderson: Well, I suppose that you make a very valid point there, Gianluca. Because I’m doing a module for my current master’s, and it’s around the field of learning analytics—i.e., studying using machine learning to predict student scores using different regression models and so on.
And it’s meant I’ve had to go off and learn a bit about this field that’s literally called learning analytics—which is the study of learning, and how students learn, and so forth.
So when you go and look into that, it actually says, in the early parts of my study, that there are various definitions, and people have different interpretations of it.
The problem is, I would say, that a lot of us in SEO are busy doing blog posts, but we don’t ever really submit things to academia for proper peer reviewing. And there have been conversations in the past—myself included in that—all of us. None of us, literally, submits anything for proper peer reviewing. I wonder whether, in the future, we should do that. We just don’t do it at the moment. So yes, I think that would probably be a good early step toward that.
And we have had those conversations in the past. I remember, years ago, when Ammon and Bill Slawski used to have their discussions on a Tuesday or Monday night—can’t remember which one it was—and Bill was very much in favor of the idea of people submitting things for peer review. And I think that would be a really good place to start. But we don’t do that at the moment.
The only kind of peer review we have is when people submit to one of the news websites, and the editors do the reviewing. But it’s not done by a group of people who are very experienced in the industry. And that would probably be a good place to start.
Gianluca Fiorelli: Yes.
Dawn Anderson: Then again, what would be the reward for doing it? I don’t know whether we are quite as accommodating as an industry as peer reviewers are in academia—because that’s part of their job. But I don’t know.
Gianluca Fiorelli: Yes, I know.
Dawn Anderson: Maybe some of the websites, like Search Engine Journal and Search Engine Land, could have a panel of peer reviewers. That would probably be quite a good thing to do.
Gianluca Fiorelli: Yes, you’re totally right. But at the same time, I think, for instance, that all the
GEO frenzy and somehow promotion ultimately began with a paper submitted to arXiv.
Dawn Anderson: Yes, exactly. That’s true. But that paper had, last time I checked, and it was a while ago, only about 13 citations. When you compare that with a lot of the other papers out there, it’s nothing. And some of them are SEOs as well. So it’s just somebody saying, “Well, this is just an opinion”.
I get the whole debate. I’ve kind of backed off from the debate around GEO, AEO, etc., whatever you want to call it. Because for me, I’m always expecting change anyway. Hence why I’m always learning. Hence, why I think we should all always be learning. So, while it is a steep learning curve at the moment, I don’t feel it justifies a new name, as such.
That said, you know—there is this notion in the IR world of generative information retrieval, of which the birth was in 2023. Which kind of coincides with when people started talking about the GEO thing. But they don’t turn around and say, “The whole of IR has changed.” They just say, “This is another aspect of it.”
And I think if we’re ever going to say, ‘Well, there’s this thing called GEO”, we should just say: “It’s an aspect of search”. Rather than all this “SEO is dead” stuff—which it clearly isn’t—but I love it when people say SEO is dead, because people stop doing SEO. And that just means that people like us, who carry on, just surge ahead. So, you know, we eat their lunch.
And we’ve seen it before—with Penguin, with Panda—all these things where people went off like, “Oh, I’m gonna go off and be this, that, and the other.” And then eventually, they come back, because SEO is a great and very interesting career. So eventually, they come back. But people like you and me, who just stuck with it and worked through it, we’re still here.
Gianluca Fiorelli: I agree with you, saying that GEO, AEO, and LLMO are nothing but a new layer—let’s call it so. A new vertical, if you want.
Dawn Anderson: It’s a new surface, isn’t it?
Gianluca Fiorelli: Yes, a new surface. It’s like when we have SEO as a general concept—and then we have SEO for news, SEO for local, SEO for AI search, etc. We could even go into the micro, micro category—talking about SEO for e-commerce, SEO for travel, and other specific verticals.
Dawn Anderson: App store optimization, shopping channel optimization, shopping feed optimization—we go very granular. We could do. I really wish that people would. I think the thing that hurts is that people are trying to claim that there's a whole new way of doing this, and I don't believe there is.
It’s essentially a combination of natural language processing and having to have the grounding from information retrieval. So it's two fields of the IR world—NLP and IR—having to come together. And maybe there are aspects as well of knowledge graphs and so forth, as other reinforcing sources of data or information.
So really, it's all just IR. And search has always been all of IR. Nothing really has changed—other than that all of these things are also moving forward as individual fields, but they're moving forward together as well now. So, for me, it’s a very exciting time.
Nobody in the IR field is talking about IR being dead. They're talking about the fact that there are many challenges still to deal with. Information retrieval is hard. Search is a hard problem to solve. You know? It's a long way from being solved.
Gianluca Fiorelli: Yes, and, personally—maybe it’s my own thing because of my background—I have my own way to try to understand: I try to go back even to the super basic sources of inspiration that engineers may have in order to figure out how to make machines try to “think” as humans do.
And when it comes to natural language processing, for instance, I go back to my root studies in linguistics—human linguistics. From there, you can then try to infer what they are trying to do. And it's also quite interesting because, in my case—since I didn’t do technical studies or formal studies in information retrieval—it helps me to understand what, in the technical world, the patents are trying to say.
And obviously, I always go to people who know these things better than me and ask if I’m not sure—so I don’t misunderstand something or everything. But I think that if you’re not sure, go to the basics in this sense, which is going to be: What is semantics? Semantics existed even before that. What is semiotics? Semiotics was invented in the 1940s and 1950s. And linguistics is a very old, humanistic discipline.
Dawn Anderson: Exactly. And none of that has changed. None of that has changed; the foundations still remain. There’s just a lot more automation with it, of course—but that’s the only difference.
There’s more automation. There are more probabilistic determinations. There’s more use of automations of statistics and things like that. You know, being able to predict things much more easily—and measure the predictions using tools like, I don’t know, scikit-learn and so forth, and all these different machine learning programs.
But the bottom line still is about people searching, human language, and people having information needs. Search engines and now LLMs, to some extent, are providing humans with a way of accessing information and completing tasks. And that's the bottom line.
There's always going to be a need for that to be studied. People are still going to do silly things like block robots.txt, or they’re still going to noindex whole sections of websites accidentally, or they’re still going to index their internal search results, or they’re still going to have no clue and just create spam inadvertently. So there's always going to be a role for people to guide website owners and businesses through these challenges.
Gianluca Fiorelli: Yes, it’s totally true. Because in the end, even if it's AI—especially when AI is grounding or using RAG—it needs to find things in a place that is search-indexed. So if you have to be in the search index, you still have to make your website crawlable, passable, and indexable.
Dawn Anderson: And the content still has to be good quality, and you still have to be trusted.
Gianluca Fiorelli: Yes, and that’s the definition of good quality. What does “good quality” mean? In Google terms, it's not how well you write, but how well you target the intent, and how well you're able to provide something beneficial to satisfy the beneficial purpose that the intent implies. That’s the quality.
Dawn Anderson: Yes, exactly.
Gianluca Fiorelli: That’s why sometimes pages that are sincerely a crime—in terms of graphics or in terms of how things are written, or what could be called formal quality—work so well.
Dawn Anderson: Yes, because they have the structure. They're like a library. That’s it. And that’s what it comes back to some extent.
Structured Data, Disambiguation and AI Search
Gianluca Fiorelli: And talking about the structure... I know this is a classic polemic lately. I think the polemic, many times, is based—at least for me—on the incorrect focus that is taken when it's discussed, which is the importance of structured data.
Because on one side, we have Bing saying, “Use structured data. We need it, also for AI search.” And just last week, Google itself claimed—and there was a report also on Search Engine Roundtable—“Use structured data, because AI search needs structured data.”
But then there is all this polemic coming out about how structured data: is not a ranking factor, LLMs cannot even read it, and so on. What is your position?
Dawn Anderson: Well, it's a funny one because I feel that even though structured data has never been a ranking factor as such—we know that, it's never been a ranking factor—but it's a clarification factor, it's a disambiguation factor, etc.
So for me, I think if search results have been disambiguated—so that something is seen as more precise because there's more understanding of what the content means—that actually then will, in turn, benefit what gets fetched by grounding. Because ranking more highly—there is a correlation between being included from the grounded result.
So it's like a secondary effect in my mind. It's like—if you can disambiguate, then you'll probably rank better as a result of more precision to the topic. And then, as a result of that, you're probably more likely to be fetched as part of grounding.
So I think that there is a positive impact, even though others have said, “Well, they can't see it.” It's not the LLM that needs to see it—I think it's still Google that needs to see it. And then it comes through as a secondary part of grounding.
Gianluca Fiorelli: Yes, that is exactly what I also think. I sometimes like to use my structured data as if it were a sort of script.
Dawn Anderson: Yes.
Gianluca Fiorelli: This is a classic example. I already did it a couple of times here, at The Search Session. Let’s say we have to write a biography page for one of the stakeholders of a company.
Instead of telling to the copywriter, “Okay, go to LinkedIn and try to reformulate what he wrote,” I go to Person schema, and I say: what are the things that are included in the schema as a property—and let’s try to give these information to the copywriter and, obviously, try to write the biography of this person around these properties that are associated to a person.
Then obviously, it depends also if the person we are writing the biography of is giving you information or not. But I usually try to use this kind of thing—even if then, maybe the structured data is not implemented with such granularity as is indicated by schema.org.
But I know that there is that information, that we are structuring in a way that search engines can parse better than without structure.
Dawn Anderson: Yes, exactly. Obviously, we know that the likes of Schema have been around a really long, long time—and there’s just more value around the disambiguation than anything else. So for me, there’s definitely an argument to still include it, ultimately.
And then, even if we just say, “Well, not everything’s gonna be about AI search for a very long time to come”—what if you still get into Google Search results because of the schema that you added? There’s still gonna be a lot of value there.
In fact, interestingly enough, somebody did a study, I think the other day, that said that actually AI Overviews, for instance, are dropping down the page, indicating in some instances that they're less valuable than Google had initially received them to be. So, in that instance, the use of Schema may help you to be in Featured Snippets or other search features.
So, I say: include it if you can. But it's very difficult, because obviously we know full well that there's usually only limited resources within any organization. If you had a choice between dealing with some other critical technical SEO issue or adding Schema, then it’s a little bit of a “nice-to-have,” really, isn’t it? That’s the challenge.
Gianluca Fiorelli: Yes, and this is something that I also suggest when I audit and prepare—let's say—a plan for a client for implementing actions for optimizing the website. “Let’s go first to optimize the things that really matter. First, that really makes the cut.
For instance, if you have a tremendous number of canonicalized pages—because you don't want to use noindex, because you're feeling that the value of noindex is passing through internal linking and etc.—so you have millions of pages canonicalized. And then you go to Search Console and say: “Google chose another canonical than the one indicated by the user”. First, correct this kind of mistake—things that are evident in Search Console. And then go with other things, like—in this case, we were talking about—Schema, structured data. And then another thing could be Core Web Vitals. I mean, don’t stress yourself out like crazy because you’re in the orange—or you want to bring it to green—for every page of your website.
Dawn Anderson: You have to prioritize things. That's the challenge. Nothing’s ever going to be perfect. I think I’ve never, ever looked in Search Console and seen a website that’s got no issues whatsoever, in all the years now. Nothing will ever be totally perfect. As simple as that.
There’s always gonna be that battle for priority, isn’t there?
How Dawn Uses AI and Python Day-to-Day
Gianluca Fiorelli: Yes. Now, talking about AI as a tool, how do you use AI, eventually, if you use it, in your daily work?
Dawn Anderson: Mostly I use it for ideation now, to be honest. I don’t ever take something and copy-paste it from AI at the moment. I don’t feel it’s up to the mark on that. And also, to be honest, I feel that, to a large extent, the human value-add is really key. But I do use it for ideation. I use it for “starter for 10” stuff. Like, you know: “Give me some ideas for this”, or Build a plan for this”, or whatever.
And I use it a lot in my studies as well. And potentially whenever I’m doing things with code, with Python. I’ve started to say to it: “Explain to me this code”. So it explains to me all, for instance, variables or what’s happening in a piece of code. It fully explains the machine learning side of it. I’m doing a lot with that as well.
What else? Yes, I’m picking up on a lot of other people’s scripts that are out there. I’ve looked at some of the stuff that Lee Foot does, obviously, some of the stuff that Elias Dabbas does.
So I suppose they’re all different aspects of machine learning—if you include Python.
Gianluca Fiorelli: Yes, and just as a reminder, now that Google is pushing Gemini practically in every, even more remote corner of its application and with the explore function of Chrome, you can debug or simply ask things to the web developer tool using Gemini.
Dawn Anderson: It auto suggests loads of stuff, it's brilliant. The auto suggestion aspects now are really good. I use Google Colab for Python stuff, for the machine learning things. And if I'm defining all the different libraries that I need to use, it'll suggest libraries that go together, and so I just have to tab along. It does quite a bit of the work for you. So that's good as well. Yeah.
AI is brilliant for assisting, and that's the thing. We need it for the staff for 10, as long as we don't forget how to learn, and that it's there to help us be more efficient rather than take over as such. We don't want it to take over the learning.
Measuring AI Search, Zero-Click and Query Refinement
Gianluca Fiorelli: And talking about how you're seeing, maybe, in the metrics of your clients, the impact of AI in the sense of how it's changing the way people are searching.
Dawn Anderson: Obviously, there’s kind of a whole meme that goes around about people in any kind of senior leadership being interested in AI. Nobody wants to get left behind. So I think SEER Interactive did a really, really good—very simple, but very effective—local studio report.
It just pulls in three pie charts that show the impact of AI search from analytics. It shows the impact of AI search traffic, the current impacts of organic search traffic, and the impacts of organic social. So you can literally see, at an immediate glance, the overwhelming amount of organic search traffic compared with AI search traffic.
So I think it's about quantifying it—being totally transparent, not trying to hide anything—and just using this time to define what measurement around what success looks like with AI search. Because obviously it is changing.
There is, to some extent, the whole zero-click thing. Although I still believe there are absolutely millions of clicks out there—you just have to think a bit more broadly and take different approaches. But still, I get that people are like, “Well, Google’s keeping all the traffic,” and so forth. So I think we have to be better with measurement—across both organic search and AI search—and, to some extent, organic social as well.
Gianluca Fiorelli: I have a question that’s popping into my mind regarding this. Maybe one of the biggest misunderstandings—and maybe one of the least considered—because it’s an old concept, not so fancy. But what if “zero-click,” which obviously exists—there are tons of searches that are zero-click? I mean, the ones that provide instant answers, immediately satisfying the need for information. For example: “the age of Barack Obama”—if Google is answering immediately, I don’t need to click.
But in general, isn’t it that “zero-click” is sometimes misunderstood? In the sense that it’s not that search is disappearing, but that the search session that eventually generates a click is just expanding.
Dawn Anderson: Yes, it’s query refinement. If someone lands and they search for “dresses.” So Google doesn’t know what the person means by that. Sure, it could send them to a generic dress site. But instead, the user does a bit of clicking around, and Google realizes:
“Oh, actually, this person is looking for a black dress with short sleeves,” etc. Then it works it out and does a query refinement on the actual search results—sending someone directly to a page with “black lace dress with short sleeves.” So they’re much closer to a conversion.
So I think there’s been an awful lot of: “Well, if Google doesn’t immediately send people off from the first click to the open web, then that’s zero-click.” I don’t think it is. I think it’s query refinement, really—a lot of it.
Gianluca Fiorelli: That’s why I personally really like how Google, through all this query refinement, is essentially suggesting to us what could be the potential search journeys of users—which, for me, is the real keyword research.
Dawn Anderson: You could argue they’re actually doing the work for us. They’re kind of doing a bit of that: asking people what they really mean by the query, because queries are very ambiguous in most cases.
Gianluca Fiorelli: Totally. Just one last question about the evolution. I think that everything we’re seeing right now—for instance, on Google—is sort of a big beta of what it will eventually become.
Dawn Anderson: Of course, yes.
Agentic Search and What’s Coming Next
Gianluca Fiorelli: But one of the new things—also because it’s being pushed by the LLMs and maybe in the future also by Google for AI Mode or the instant checkout in Google Search—is this idea of Agent Search. What’s your opinion on that?
Dawn Anderson: Again, I think it’s going to be a big part of search overall. I think it’ll just be another aspect. I think it’s a way off for people to trust an agent to go off and book tickets, or book a restaurant, etc. It’ll probably need a bit more work. I don’t think websites are largely prepared for agentic search yet. I think it’ll be another aspect we’ll have to be mindful of. I also think there’ll be an aspect of Google, and so the search engines, having to get the measurement right, so that the data doesn’t end up being completely skewed by bots all over the place. So there’ll be that. I think there’ll be some challenges, but it’s just something that we are going to have to accept is coming—and again, more adaptation on their side—and seeing it more as an opportunity rather than a threat.
Gianluca Fiorelli: Yes.
Dawn Anderson: Everything’s an opportunity.
Gianluca Fiorelli: Yeah, always try to see the glass…
Dawn Anderson: …half full.
Gianluca Fiorelli: Exactly.
Dawn Anderson: Yes, and bear in mind as well how many things search engines have tried to implement and then decided were rubbish ideas—like AMP, like the Google Graveyard products, etc.
Gianluca Fiorelli: Yes.
Dawn Anderson: We also have to remember that even though search is a massive, massive part of search engines, there are now other product offerings of theirs that are massively profitable. Like cloud computing, for instance. You know, that’s huge. That’s an absolutely huge part of Google’s offering: cloud storage.
Gianluca Fiorelli: It’s one of the contributors to their very good earnings.
Dawn Anderson: Exactly. The same with Microsoft and the whole Azure network—or the Azure platform, rather. So I think we forget that, yes, search is massive, but Google also has other fish to fry as well.
The Fireside Questionnaire
Gianluca Fiorelli: Yes, well, almost one hour. So let’s stop talking about AI search, information retrieval, search engines, and so on. Let’s talk about you. As I said in the introduction, there’s a specific type—subcategory—of travel that you really like, which is cruising.
One: Why do you like it so much? Two: What is the cruise that still stands as the best one you’ve experienced?
Dawn Anderson: Well, what do I like about it? I actually like being in different places every day and being able to just go and take a snapshot of somewhere and then deciding later if I want to go back there in the future. So having that varied vacation, if you like. I'm not one for staying in one place. I find it gets a bit monotonous. So there’s that.
Obviously, I love lying on the deck of a cruise liner and listening to the band playing—with a cocktail in my hand—so there’s that as well. And, you know, I like wearing the nice outfits. I’ve been on a lot of cruises over the years. Not so much recently, because we've had struggles getting somebody to look after our dogs. We have two Pomeranians.
But I’ve been on a lot of really nice ones. I always liked the Caribbean ones. We went from Florida many years ago, and that was great. I really liked the Caribbean ones.
We’re considering going on one next year that’s going to go from the UK, actually, because my hubby is not a huge fan of flying. So that one will go to Northern Europe—Denmark, Amsterdam, Belgium—which doesn’t sound that exciting, but as long as we get the weather, I think it’ll be nice to go around that side for a change.
Gianluca Fiorelli: Have you thought about the river cruise?
Dawn Anderson: I have, and if I did go on one, I’d like to do one in Germany. When I went to Bonn years ago—that is an absolutely lovely place. But the hotel we stayed at was on the Rhine, and there were river cruises going backwards and forwards up there—the Viking River Cruises.
So I already did that—and I have thought about it—but the challenge is, there are so many places you could possibly go, isn’t there? We’re actually also considering going to Sorrento next year, maybe in Italy. Friends of ours have been there, and supposedly it’s very nice. So there’s that. Just deciding, isn’t it?
Gianluca Fiorelli: And there’s another side of you—which is the Dawn Anderson who really, really runs, during the year, hundreds of kilometres. How did this passion for running start?
Dawn Anderson: Well, that actually came along during COVID. Because obviously everybody was busy baking banana bread and whatnot, and like everybody else, I put on a few pounds. So I decided I was gonna do the Couch to 5K program, which is the UK’s NHS app for getting fit and going from not doing any running whatsoever to running 5K over nine weeks.
So I did that and found it was a really good structured running program. I literally couldn’t run around the block to start with! But I did the Couch to 5K, and as it got to the end, it said, “What you should do next is find a running club.” So I found my local running club—and that was nearly four years ago now. That’s the Harriers—my Middleton Harriers.
So I run with them like three or four times a week, and then I do parkrun on a Saturday. I’m on my 160-odd parkrun, I think. And I did a half-marathon on Sunday—and I’m doing a 10-mile race this Sunday.
Gianluca Fiorelli: One year, you have to come here to Valencia for the real marathon.
Dawn Anderson: Well, that would be nice, lovely idea. But I’m going to Benidorm, doing the Benidorm Half next month. I did it last year and it was great. So I’m doing that one again next month with my friends from the running club. So, yes, I love it. And the thing is, I have to do it, because our work—as we know—is very sedentary. You know, once we start analyzing a crawl, we don’t move.
Gianluca Fiorelli: Well, you can’t really analyze a crawl when you’re running.
Dawn Anderson: That is true! Yes, exactly. So we need to have balance, don’t we?
Gianluca Fiorelli: Totally. Thank you, Dawn. It was really a big, big pleasure to have you.
Dawn Anderson: And like. Thank you again for the invitation. That’s very kind. I appreciate it. It’s been good, I’ve enjoyed the chat. Thank you, Gianluca.
Gianluca Fiorelli: You’re welcome. Let’s promise to have a new chat in the future.
Dawn Anderson: Yes, absolutely! Sounds good. Thank you.
Gianluca Fiorelli: You’re welcome. And thank you to all of you for being here until now, listening to us—and especially to Dawn—talking about what’s happening in the world of SEO.
Remember to ring the bell to be notified about new episodes. And if you want to help us, subscribe to the channel to help it grow. Thank you, and bye-bye!
Dawn Anderson: 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