Optimizing the New Search: How Relevance Engineering Is Reshaping SEO | Mike King

Jul 7, 2025

30

min read

Welcome back to a new episode of The Search Session! I’m Gianluca Fiorelli and I’m joined by Mike King, the mind behind “relevance engineering.” 

Mike King and I unpack how AI is transforming search—and how he and his team at iPullRank were ahead of the curve, experimenting with language models and building tools that mimicked AI Overviews as early as 2020.

We discuss how client perceptions around AI have shifted, and what it means to optimize in this new landscape—moving from pages to passages, and from static strategies to “relevance engineering.” 

Mike explains how AI Overviews and AI Mode actually work, and why visibility now depends on building an omni media content plan. 

We also explore why SEOs must move beyond blog posts and into white papers and research to truly understand what’s next. 

As Mike puts it, we’re no longer just mechanics tweaking websites—we’re engineers building entire systems.

We also explore the rise of agentic SEO—automated, modular systems that adapt to visibility shifts—and new success metrics like branded visibility, mindshare, and directional traffic signals.

Mike King

Founder and CEO of iPullRank

Mike King is a digital marketing expert and founder of iPullRank, credited with generating over $4B in organic revenue for brands like SAP, American Express, HSBC, and Nordstrom. 

A former hip-hop artist with a background in computer science, Mike was named “Search Marketer of the Year” in 2020. 

He published The Science of SEO, created AI-focused tools like Qforia and Orbitwise, and founded the SEOweek conference in New York.

Mike King

Founder and CEO of iPullRank

Mike King is a digital marketing expert and founder of iPullRank, credited with generating over $4B in organic revenue for brands like SAP, American Express, HSBC, and Nordstrom. 

A former hip-hop artist with a background in computer science, Mike was named “Search Marketer of the Year” in 2020. 

He published The Science of SEO, created AI-focused tools like Qforia and Orbitwise, and founded the SEOweek conference in New York.

Mike King

Founder and CEO of iPullRank

Mike King is a digital marketing expert and founder of iPullRank, credited with generating over $4B in organic revenue for brands like SAP, American Express, HSBC, and Nordstrom. 

A former hip-hop artist with a background in computer science, Mike was named “Search Marketer of the Year” in 2020. 

He published The Science of SEO, created AI-focused tools like Qforia and Orbitwise, and founded the SEOweek conference in New York.

Transcript

Gianluca Fiorelli: Hi, and welcome to a new episode of The Search Session! Today, we’re joined by someone who’s been incredibly vocal and visible in the search marketing industry—whatever we choose to call it these days. Not that he wasn’t already prominent before, but over the past couple of months, he’s really been at the center of attention.

Meet Our Guest: Mike King

Gianluca Fiorelli: What if I tell you he coined the term “relevance engineer”? He’s one of the people who made the recent Google leaks public, and he’s been sharing some seriously high-level knowledge in the SEO world. Maybe you already know who our guest is going to be, Mike King. Hi Mike, how are you doing?

Mike King: Hey Gianluca, I’m fantastic. It’s good to see you!

Gianluca Fiorelli: Man, yeah—it’s been quite a while since we’ve seen each other in person. 

How SEO Is Treating Mike King in 2025

Gianluca Fiorelli: So, a question I ask all my guests: How is “SEO” treating you lately?

Mike King: Oh, it's treating me fantastically. This last year has been great. Ever since the leak, we’ve gotten a lot of new clients. There’s a lot of momentum for us. SEO Week has been incredible, and everything we’ve been putting out around AI Mode has gotten a great reception.

A lot of people have been saying things like, “Hey, my existing SEO firm is too old school—they don’t know what to do. Clearly, you guys do.” So it’s created a lot of opportunities for us. I’ve been traveling a lot, meeting great people in the SEO space. So yeah, this has been one of the best periods for me since I started doing this work.

Gianluca Fiorelli: Great, I’m really happy to hear that. Yes, it’s definitely a time of change. We’re in the middle of a transition—we still have one foot in the old way of doing things, and the other stepping into new territory. Let’s see where that leads. Maybe we’ll get into some speculation later in the conversation.

Enterprise Clients' Reactions to the Rise of AI in Search

Gianluca Fiorelli: But first, I want to ask you something. You and your agency, iPullRank, work a lot with enterprise companies. Over the past couple of years, how have those companies reacted to the rise of AI?

First with ChatGPT and Perplexity especially—though I don’t really count Claude. Claude is a fantastic tool, but it feels a bit like the DuckDuckGo of LLMs for me.

And also, with Google entering the space. What was your initial reaction, and what’s your reaction now, after all this new knowledge started appearing about AI Mode, AI Overviews, and LLMs?

Mike King: Yeah, I mean, at first it was this esoteric thing that only a few people really knew about—or something we’d bring up in specific use cases, like for e-commerce sites building out PLPs. I’m talking about using GPT-2, so this was well before ChatGPT. Back in 2020, we were already doing this kind of stuff.

And clients would say, “Cool, this is better than, you know, content spinning.” We were able to create some really interesting, unique, and valuable content that pulled in data from their internal models. I’ve been showing examples like this going back to 2020 and even earlier.

Then ChatGPT came out and entered the cultural zeitgeist. But at first, a lot of clients were like, “Oh, we don’t want to do that. We don’t want to do this AI thing—it’s just spam or whatever.”

And then what happened was, CMOs started hearing about ChatGPT. Suddenly, they were like, “Well, why aren’t we showing up?” Same kind of scenario—they’d do a vanity search and ask, “Why are we not number one for this query?” And so that sends a shockwave across the organization.

What’s happened over the past year is that the question became: “Okay, how do you optimize for ChatGPT?” And then, as Google entered the fray with AI Overviews and that started impacting things, it shifted to: “How do we get into there?”

With AI Mode and AI Overviews, we were ahead of the game. I don’t know if you remember this, but two years ago I wrote a blog post saying that RAG was going to be the future of search. I had basically rebuilt what was then called SGE—now AI Overviews—just to see how it worked.

I built a tool using a SERP API and a library called LlamaIndex. You’d enter a query, it would pull in the results, display them, crawl the top 10 results, and then build a vector index of that content. Then, using a prompt, it would generate the AI Overview-style response.

I walked through exactly how all of it worked. And I also built a projection model that showed, based on the keywords you care about, how much traffic you might lose. That model was predicting anywhere from 20% to 60%—and that’s exactly what we’re seeing now that AI Overviews has rolled out.

So, we’ve been fortunate to have that kind of forward-thinking mindset. Being able to showcase and dig into this stuff early on really helped us stay prepared. So when AI Overviews started to roll out, clients were already coming to us—or we had already been telling them, “These are the things you need to watch out for.”

We had that trust in place, which meant we could move faster than a lot of others who came in after the fact—after the impact had already hit. A lot of them are just, frankly, referencing things we’ve already put out there.

We’ve been showing people how to adapt—what to do to survive in this new era. It’s definitely been a whole hype cycle, and we’ve managed to stay ahead of it. Now, we’re just really starting to reap the rewards of all that work we’ve been doing over the past few years.

Gianluca Fiorelli: Yeah, yeah—I hear you, and I totally get it. Even at my level, though I work with different types of clients, I’ve seen almost the same thing. And the challenge is really convincing them to take that leap—to invest in what you're proposing.

Because a lot of the time, it's quite different from how they were working before. Not that the fundamentals have changed—the basics are still pretty similar—but the strategy, the tools, the mechanisms of working? Those are different.

That’s probably the toughest part: getting clients to sail with you on this new voyage.

How SEOs Can Start Learning AI Search

Gianluca Fiorelli: But let’s say—if you had a few recommendations for all the SEOs out there, whether in-house or at an agency. Many of us are starting to understand these things, but we want to go deeper. What would you suggest—not necessarily going as far as all the travel and deep dives you’ve done—but how should someone start, in a practical way, to truly understand how this new AI-powered search works?

Mike King: I think there are some great resources that Aleyda Solis has been putting together for folks—both for learning SEO in general and, I believe, she has another site focused on learning AI search specifically. There’s a lot of valuable material linked there.

But for me, I think it really comes down to reading the white papers—and maybe even the patents, though the white papers are usually more than enough to understand the current state of the art.

The truth is, this has always been the case. Engineers at Google have consistently shared how they build things. Just like we have conferences, they have conferences—like NeurIPS, and various ACM events—where they present their research and publish white papers explaining, “Here’s how we built this.”

The thing is, in the SEO world, we don’t usually pay attention to that stuff. We’re always looking for the practical, actionable, tactical tips—like, “What do I do right now?”

And for me, that’s just not enough. I need to know how this stuff actually works—because only then can we figure out what we should be doing moving forward. We need to figure out where those edge cases are. And if you’re just relying on the SEO community, you’re getting third-hand information—mostly anecdotal stuff, like, “Here’s what I saw.”

The problem is that kind of information often becomes gospel. That’s how we ended up with some of our so-called best practices. Someone once said, “Keep your page title to 60 or 70 characters,” and we’ve been running with that for 20 years. But in reality, if you actually test it, you’ll see that titles longer than 60–70 characters often work just fine—in fact, they work better.

So for me, reading research papers on arXiv, following engineers, following the computer scientists who are talking about the state-of-the-art—that’s going to help you understand what to do way more effectively than just reading SEO blog posts.

Gianluca Fiorelli: Yes, I completely agree. That’s actually something I’ve started doing too.

Let’s say, for example, that the classic Googlers—the ones we used to rely on—aren’t really active anymore. But the engineers? They’re still very vocal, especially on Twitter.

I’ve created a dedicated list just to follow them, so I don’t miss anything they share. Sometimes I don’t fully understand what they’re posting—I'll be honest, some of that information is above my level. But I keep at it. I put myself in the student mindset and try to understand, because I agree with you. 

Relevance Engineering: A New Definition of SEO

Gianluca Fiorelli: Using a metaphor, we’re like the mechanics. And we need to understand how the engine works to make our fixes to the engine.

Mike King: Yeah, but what I’d say at this point is—we’re more than just mechanics, right? I think we’re actually having to build the cars now.

There just isn’t enough tooling out there to support what we need to do for these modern, generative AI surfaces. Like, I honestly don’t know how anyone’s optimizing for this stuff without building their own tools—whether it’s in Colab, standard Python, or whatever else.

For example, passage indexing requires us to optimize at the passage level—but there’s no SEO tool out there that does that. All the existing tools are still focused on optimizing entire pages.

So I think now, we’re not just mechanics tweaking engines—we’re engineers building the actual systems.

Gianluca Fiorelli: Yes, and from there comes your definition of this new kind of work: relevance engineering.

“Engineering” is everything we’ve just talked about. And “relevance” is key—because, as you’ve said in your recent talks at SMX, and in your guides on iPullRank and Search Engine Land, it’s all about how AI Mode and AI Overviews work.

We have to be relevant for all the potential uses our content chunks might have in the context of large language models. That’s really what it’s about. Sure, it’s a bit of a verbose definition—but it works for me.

Mike King: It’s not more verbose than search engine optimization. “Search engine optimization” as a term never really made sense. We weren’t optimizing search engines.

Gianluca Fiorelli: Oh, I totally agree with you on that. I mean, I never worked to make Google better from the inside.

But let’s shift gears for a moment. Before I get to my next question—I want to talk about something else.

You come from a background of studying patents and white papers. My background is a bit different—it’s rooted in linguistics, semiotics, semantics, and so on. Not directly computer science, but still very connected in terms of how meaning is structured and communicated.

That’s probably why I can often recognize where these technologies are trying to go.

So what I typically do—using tools like Qforia, which you’ve developed at iPullRank and which I find fantastic, or the query fan-out simulator by our mutual friend at WordLift—is approach things from a different angle.

I start with a really deep audience analysis to understand potential needs, motivations, and even emotional triggers that might spark a search journey. And from there, I work downward—moving from intent to content, and even down to the passage level.

So, in essence, I’m aiming for the same outcome, but my process is different. I’ve never worked with keyword lists in the millions—it always felt a bit insane to me. I start from the entity, then move down to the specific keywords that truly matter. 

Inside Google’s AI Mode & AI Overviews Architecture

Gianluca Fiorelli: When thinking about how Google’s AI Mode and AI Overviews function, I think it’s clear to many now that there’s a query fan-out process—where one query expands into a broader set, and that expanded set generates a corpus of documents. From that corpus, Google retrieves passages—chunks of content—that are then reformulated to create the synthetic answer.

But there’s one part of your analysis of the white papers and patents that really stood out to me, and that not many people are talking about. It became clear—especially if you remember that screenshot from a few months ago, where an AI search result showed a long list of filters. Each of those filters was tied to a specific type of search.

To me, that points to an important moment in the process—a kind of pre-ranking stage, even if that’s not the technically correct term—where Google decides which specific model to use based on the type of answer it has determined it needs to give.

So, can you explain that step a bit more clearly? Especially how this works in AI Mode and AI Overviews? Because I feel like it’s a critical part of the process that not many people are paying attention to.

Mike King: Yeah, so there’s a lot of overlap between how AI Overviews and AI Mode work. But the core idea is this: you take the user’s query and extrapolate from it using a series of synthetic queries.

For AI Overviews, that might mean just a handful of follow-up queries. But in AI Mode, it could be dozens. You can actually see how many searches it performs and how many documents it pulls in.

From there, Google extracts what I call “candidate passages” from those documents. These are then fed into a series of language models.

One model figures out the structure of the response—like, does it need to include videos? Images? What should the layout be? It’s this concept of a bespoke UI tailored to the user’s specific needs.

Then all those candidate passages are compared against each other to determine which ones are the best—essentially, which should actually be used to answer the query. That set is fed to a series of language models, which generate the final response, and then that response is generated to the user..

Before that response is shown to the user, it may also be filtered based on personal context—information Google has from across its ecosystem. They’ve explicitly said this could include data from your Gmail, for example.

But, within those synthetic queries, there are nine different types. And one of those synthetic query types can include recent queries. Others might be inferred, or based on entity expansion, and so on.

SEO in the AI Era: Owning More Raffle Tickets

Mike King: So where we need to play as SEOs is, first, understanding what all of those queries are, and then figuring out how we rank for them.

You’ve basically got to have an omni media content plan. And what I mean by that is: you have to treat this more like a reputation management campaign. You want as much coverage across as many of these keywords as possible, so you have the highest number of potential candidate passages in play.

Because when you get into the reasoning chains, that’s where things become much more probabilistic. You don’t know what the language model is going to select, even if you have what seems like the best-ranking passage for a given idea.

There might be 10, 20 other passages that are just as good, and they could be chosen for any number of reasons within that reasoning chain. So the way I see it—it’s like a raffle. You want to buy as many tickets as possible to increase your chances of being picked.

In this case, the “win” is getting cited in the AI-generated response. Because ultimately, the goal is one: visibility, and two: a link. Visibility is the more important of the two, because these channels are becoming more brand awareness channels than performance channels. But in that off chance that the user is ready to click through—you want to be the one they click on.

To better understand how visible your content really is—across both traditional search and AI-generated results—it helps to have the right tools. 

Advanced Web Ranking now tracks citations in LLMs and monitors your brand’s presence across all search environments. It’s a practical way to see how often you're being surfaced, even when you're not the top organic result.

Try AWR for free and start measuring your visibility where it matters most.

To better understand how visible your content really is—across both traditional search and AI-generated results—it helps to have the right tools. 

Advanced Web Ranking now tracks citations in LLMs and monitors your brand’s presence across all search environments. It’s a practical way to see how often you're being surfaced, even when you're not the top organic result.

Try AWR for free and start measuring your visibility where it matters most.

To better understand how visible your content really is—across both traditional search and AI-generated results—it helps to have the right tools. 

Advanced Web Ranking now tracks citations in LLMs and monitors your brand’s presence across all search environments. It’s a practical way to see how often you're being surfaced, even when you're not the top organic result.

Try AWR for free and start measuring your visibility where it matters most.

Gianluca Fiorelli: Yeah, yeah. And while you were talking, I was thinking about one of the classic issues that used to torment us SEOs in the past—keyword cannibalization. In this new context, though, keyword cannibalization isn’t really a problem.

Mike King: Yeah, it’s actually a good thing.

Gianluca Fiorelli: Exactly—it becomes an advantage.

That’s something I’ve tried to explain before, even in traditional search. Keyword cannibalization only really exists when you're targeting the same keyword with the same search intent.

But if you're targeting different intents—even for the same keyword—it’s not cannibalization. Because what appears in search depends on the search intent perceived by Google at the moment they make the query. Sometimes one piece of content shows up, sometimes another. And that’s even more true with AI Mode, AI Overviews, and I suspect, with all other LLMs.

Chunkable and Multimodal Content Strategy

Gianluca Fiorelli: Another thing—and I’m curious if you agree—optimizing for chunks doesn’t mean creating millions of new pages. Instead, it’s often about making existing content more chunkable.

From what I’ve seen with my clients, that might just mean adding an FAQ to a product page. It’s really about rethinking how content is produced. It’s not about pumping out more blog posts or landing pages.

Mike King: Yeah, and it’s also multimodal—it’s not just about text. It could be videos, it could be images. That’s why I say it has to be an omni media content planning.

There are so many ways this data can be pulled in. One of the capabilities these systems have is the ability to take content in one language, translate it, and then include that translation as part of the AI-generated response.

So there’s a lot we need to consider—beyond just “what are we doing on a single page?”

Gianluca Fiorelli: Yes—and the translation aspect really struck me. As an international SEO, that touched me really deeply. I actually talked about this with Cindy, because when AI Overviews first started showing up here in Spain, and also in Italy, I immediately noticed that Google was translating English content into Spanish or Italian within the AI Overviews.

I even brought this up during Madrid Central, in conversations with John Mueller and other Googlers. Because honestly, it’s not fair.

First, those translations are served from a Google Translate subdomain, which doesn’t credit or link back properly to the original source. And second, it doesn’t happen in reverse. Spanish or Italian content isn’t being translated into English.

So, for all the people out there creating valuable content on the web in languages other than English—it’s pretty frustrating. Let’s be clear: it’s really pissing people off.

The Publishing Industry Under Pressure

Mike King: Yeah, I mean, a lot of what Google is doing right now just isn’t fair. It’s changing a lot of businesses. If you’re a publisher, honestly, it’s a wrap. I don’t really know what else to say, other than: diversify your channels, build your newsletters, and reconsider or renegotiate your deals with advertisers. Because Google is eroding traffic.

When you work with publishers, you usually rely on two key tactics. One is for the ephemeral stuff—things that perform in Discover or Google News. The other is to build up a base of evergreen content to help smooth out the volatility in traffic over time.

But now, even that strategy is effectively gone. There are entire businesses built on that whole review culture and affiliate marketing—and Google is just killing it.

So I think we’re going to see a massive shift across the web. The incentive structure is completely different now. And unfortunately, there’s going to be less value placed on paying people to create amazing content.

Yeah, it’s really just going to come down to the people who want to create amazing content.

Because, to some degree, making great content is easier now, right? You’ve got tools like VEO, Flow, MidJourney—any number of generative AI platforms that let you create incredible stuff.

I honestly think we’re not far from a world where some kid in his bedroom can make something that looks as good as what we see from the MCU.

So yeah, the incentive structure is shifting. It’s going to be less about big businesses pumping out great content for profit—and more about who has the time and proclivity to make something truly awesome.

Gianluca Fiorelli: Yes. That’s something that came up in a recent conversation I had with Barry Adams. He said something I found really thought-provoking.

In the publishing industry, a lot of publishers are probably going to die out. But maybe, in some ways, that’s actually a good thing. Because the ones that do survive might finally stop chasing the click—and start creating content that people actively seek out. Content that positions them as the source people go directly to.

He was speaking specifically about newspapers—like people opening The New York Times app or mobile site directly, rather than going to Google, searching for news, and clicking on some random news.

Mike King: Here’s the problem—that model just doesn’t work.

There’s a much higher volume of people who passively consume content than there are people who intentionally go to a specific site for a specific thing. A lot of companies have tried to buck that trend.

Take The Outline, for example. Their whole approach was: “We don’t do SEO. We don’t care about that. We’re just going to make really cool, interesting, offbeat content.” And it didn’t work—they shut down.

The point is, we need a new model to support that kind of journalism, because the current one isn’t built for it. The existing model has been heavily dependent on capturing passive demand through search engines.

So until a new model emerges, I honestly don’t know what the future holds for that kind of content creation—unless it’s subscriber-driven. But even then, that’s a tough sell, because it’s not passive.

That’s why Neeva didn’t work. People don’t want to subscribe to a search engine. The only thing they seem happy to subscribe to is, like, OnlyFans.They just don’t want to pay for things that they expect to be free.

Gianluca Fiorelli: Yeah, yeah—I totally hear you. Let’s see what the future brings. It’s a really complicated moment for the publishing industry.

Memorability Over Visibility: The Real SEO Goal

Gianluca Fiorelli: But let’s come back to the concept of brand visibility. The search engine —though I’m not even sure we can still call them “search result pages”—are clearly shifting from being click-centric to awareness-centric. But I think that awareness for the sake of awareness is just noise.

Even though we know that Google tends to anonymize your presence in AI-generated answers—sometimes improving it a bit, like in AI Overviews, where we’re starting to see more blue links—visibility alone isn’t enough.

Using the same metaphor you’ve used in your talks and articles—a billboard—I think the real goal isn’t just visibility. It’s memorability. So my question is: how can our visibility become memorable?

Because if people remember you, they might start searching for your brand directly. That’s when you become part of their journey—and even enter their hyper-personalized Chrome history.

Mike King: Yeah, I think it’s actually very different from a billboard. When you see a billboard, you’re not actively looking for what’s being advertised. You’re just driving, and—bam—there’s a McDonald’s ad. You think, “Oh, cool, there’s a McDonald’s over there.”

And you now remember there’s a McDonald’s near exit 17 when you’re driving. That’s what that does—but it’s also different when you are explicitly looking for a thing, and then you see a brand mention with that thing.

I’m gonna assume—I haven’t done any analysis to prove this—but I’m gonna assume that your recall of that brand is gonna be much higher when you’re actively looking for a thing.

So, you know, the examples I keep using are like, hey, if I’m looking for who has the best basketball sneakers for someone really tall with flat feet, and then you see a list and Nike is on that list—you are now gonna associate that brand with that need.

So I think the intent aspect of it kind of codifies it in your mind in a different way. And that is inherently incredibly valuable. And I think that’s something that we, in the SEO space, downplay—because impressions are typically associated with channels that we don’t believe in, like display.

But the reality is that display has a huge impact on what people do across the web. You know, display helps create search demand, right? It also, again, creates these brand associations.

So, despite the fact that we all hate YouTube ads—if there’s a very well-targeted YouTube ad around the thing that you searched for on YouTube, you’ll then think of that thing in association with whatever it is that you searched.

So I think we need to respect what our channel does. Because we only look at it as, like, “Okay, we drove a user to click. A user clicked and then they converted at, you know, 2% or whatever.” So you think that 98% of that traffic wasn’t valuable.

But again, that’s also not true—because that person spent, you know, X amount of time on your website, in association with what their information need was. So they’ve now collected that information about your brand and associated it with whatever it is they’re doing. So yeah, I think we just need to understand the power of our channel a lot better.

Gianluca Fiorelli: When you were talking about display ads, it reminded me of a time when I was also doing some link building—which, thankfully, I don’t do anymore.

But one of the classic tactics was to check Google Analytics, look at the referral traffic, and see which websites—through display—were sending not just the most traffic, but the most relevant traffic.

Then, we’d reach out to those sites for link-building opportunities. Because if users were already seeing you there and coming to your site, it probably meant they were primed and engaged.

Context Matters More Than Mentions

Gianluca Fiorelli: And this makes me think of something else I wanted to ask you. It’s about a common tactic that’s often discussed for LLM visibility—especially with ChatGPT and similar models, more than with AI Overviews or Google AI Mode.

The tactic is simply: get your brand mentioned on the web—on those websites that are clearly and regularly used as sources by LLMs. And okay, that makes sense.

But I always wonder—why doesn’t anyone talk about the other part of that strategy?

Because for me, it’s not just that your brand is being mentioned. It’s where and in what context your brand is being mentioned. The semantic context really matters—what kind of content surrounds that mention?

And this is something I don’t hear people talk about much. Maybe it’s assumed or taken for granted, but I don’t think it should be.

Mike King: Yeah, I think we’re still early in this space, and a lot of people still don’t fully understand what’s actually happening here.

What I’m seeing is that we’re often relegating everything back to the frame of what we already understand, rather than looking at what is.

And what I mean by that is—we’ve developed this concept of “mentions” based on things like co-occurrence and things like that. That came up when Bill Slawski (rest in peace) was writing about it. Rand Fishkin created Fresh Web Explorer. Joshua Giardino, who used to work with me, wrote a blog post on it. So our collective understanding of mentions was shaped by that era.

Now, people are thinking, “Oh yeah—mentions. We get that. Let’s go get more mentions.”

But it’s not just about mentions themselves. It’s about the context in which those mentions occur. It’s about the passages where your brand appears—and what else is being said in that same space.

It’s not enough for your brand to be mentioned across the web. It has to be mentioned in a context that actually makes sense for the query or topic at hand.

That’s why this really ties back to the omni media content planning. You want to be present across a variety of channels with the right messaging. That could include your own media—like microsites or other owned content properties.

You’re writing on Medium, you’re involved in all these different communities like Reddit and LinkedIn Pulse—it could also be PR mentions and things like that.

But again, those mentions need to be relevant to what’s being discussed. It’s not enough for your brand to have, like, 500 million mentions scattered across the Internet. If they’re not relevant, they don’t even matter.

Speculating on the Future of Google Search Features

Gianluca Fiorelli: Totally right. Totally right. So let’s try to speculate a bit. We know—because it was said by Sundar Pichai himself—that Google’s goal is to eventually move AI Mode into classic search.

Mike King: Sure.

Gianluca Fiorelli: Let’s imagine that future, how the search could be, like I did in a very long sci-fi speculative post.

In that article, I was speculating: Google already has all these existing features. And we know that AI Mode is, as Cindy Krum said, “multitasking.” 

Right now, it gives us an answer—a summary—and maybe an image or a video. But how might Google eventually integrate all its existing features—or at least parts of them—into AI Mode?

We’re already seeing some of this: for example, Translate is being used in the background by AI Mode. But what about things like People Also Ask, Related Searches, People Also Search, or even the Top news, or other features like Discussions and Forums?

How do you think those might get folded into this evolving search experience?

Do you remember the video Google showed—maybe not from I/O 2025, but from I/O 2024—where they presented AI Overviews in a way that let users go deeper into insights?

Not just follow-up questions, but refining the answer based on specific needs—like when someone searched for a recipe and filtered by “I’m vegetarian.”

I think that’s where they want to go. So, how do you think Google will slowly—but steadily—move in that direction?

Mike King: Yeah, I think it’s exactly what you just described. It’s not going to be a wholesale lift-and-shift into something completely new. It’ll be more like different components gradually added to classic search.

So, for example, maybe we start seeing things like audio-based Overviews. Or features that let you interact with content in whatever way works best for you.

I could imagine something like being able to talk to a page through Google, rather than going directly to that page. We might also see more interactive ad formats show up.

So no, I don’t think we’ll suddenly get a full-on chat box embedded in classic search. But I do think we’ll see these different components slowly pulled in to augment how we use search.

To keep an eye on how these evolving components—like SERP features and AIOs—are showing up and shifting in search results over time, Advanced Web Ranking’s free SERP features tool offers a great way to track them. It helps visualize how prominent these elements are and how their presence changes across industries and devices.

To keep an eye on how these evolving components—like SERP features and AIOs—are showing up and shifting in search results over time, Advanced Web Ranking’s free SERP features tool offers a great way to track them. It helps visualize how prominent these elements are and how their presence changes across industries and devices.

To keep an eye on how these evolving components—like SERP features and AIOs—are showing up and shifting in search results over time, Advanced Web Ranking’s free SERP features tool offers a great way to track them. It helps visualize how prominent these elements are and how their presence changes across industries and devices.

Gianluca Fiorelli: And speaking of hyper-personalization—well, we’ll have to see how the European Union allows these kinds of things. We’re always a bit different here when it comes to regulation.

The Risks of Filter Bubbles and Cognitive Atrophy

Gianluca Fiorelli: But as a user—not as a search marketer—don’t you worry about the risk of being included in some kind of filter bubble when you're searching? Almost like being stuck in an echo chamber?

Mike King: I think that’s the worst part of all this—it’s going to create a bunch of filter bubbles.

And I also don’t think this modality is the best for us as users. Because part of the value of search is that you learn as you search. When you do the synthesis yourself, you start to understand the nuance of what you're looking for. That, in turn, informs how you search next.

But now, Google is interpreting the information for you, instead of just giving it to you.

And I think that’s bad for humanity. We’re already struggling with how we understand and evaluate information. And if you abstract that even further—if people just accept these AI-generated answers at face value and not react responsively to how they understand things. 

So no, I don’t think this is a good thing. That said, I get the appeal from a UX perspective. It’s about saving time—reducing what they call “Delphi costs,” the time and effort it takes to find and piece together answers.

In that sense, it’s a leap forward in user experience. I just don’t think it’s the right leap for us to be making as a humanity.

Gianluca Fiorelli: Yes, totally. Maybe Google should revisit a very old patent—one I remember Bill Slawski used to talk about. It was about search entities, but not in the sense we usually think of today, like named entities. It referred more to things like URLs, webpages themselves, links, and other types of searchable units—“entities” in the broader sense of how search interacts with data.

In that patent, they were finally talking about combining a user’s search history and personal context with these search entities to create a sort of probability score.

The idea was to offer results that weren’t just the same things you had already clicked, searched, or viewed. It aimed to diversify the results rather than reinforcing what you'd already seen.

Maybe that kind of old concept—because it’s more than 10 years old now—could be brought back. It might help Google remain compliant with privacy regulations.

Agentic Search and Modular Search Architecture

Gianluca Fiorelli: And yes, it would be just another step. But I feel like we’re starting to grasp that there’s a bigger shift happening. There’s a lot of buzz, and we’re starting to see examples of it inside Google itself. Let’s call it agentic search.

And when you were earlier talking about modular components, I think some of those are likely tied directly to this agent-based approach.

Mike King: Yeah, I mean, I think the whole agentic idea is really just a variation of what was already happening.

When you think about how Google actually works, people often simplify it too much—like, “Oh, there’s one magical equation that ranks everything.” But in reality, what’s happening is you have a series of microservices that trigger at the same time. Each one handles a different part of the process, and then they all contribute to the final score for each document.

So when we talk about agentic, it’s not all that different. These are just different components doing various operations on behalf of the user, then coming together to produce a result or response.

That’s the essence of agentic systems. And I think the exciting part is that we in SEO can also start leveraging this kind of technology—the same way Google does.

They’re already rolling out all kinds of new features and functionality that reflect this model.

For example, in shopping, you might have something that represents you, tracks prices, and even automatically buys a product when the price drops.

That concept is really fascinating to me—because you could create something similar for SEO.

For example, imagine you’re tracking a given page, and its relevance score drops below a certain threshold, while a competitor’s score improves. You can be like, “Okay, where exactly is that happening?”

And from there, you could automatically adjust the page—not just based on relevance scores, but also on what you believe about the marketing goals of the site. What’s the brand voice? What tone should be used? What are the broader objectives we're trying to achieve?

All of that could be wrapped into this agent that can do all these operations for you. That’s why I think this is exciting on both sides.

On one hand, Google will have far more capabilities to represent individual users and bring in true hyper-personalization. On the other hand, we as marketers will have more capabilities to automate, react, and proactively optimize our sites in ways that truly support what we’re trying to do in these evolving channels.

Also Watch

If this conversation about agentic systems and their impact on SEO has sparked your interest, you might also enjoy our episode "Navigating Agentic AI in the World of SEO" with Andrea Volpini. 

In it, we explore how structured data and AI agents are reshaping the search experience—from the mechanics of modern SERPs to hands-on strategies for SEOs looking to stay ahead.

Also Watch

If this conversation about agentic systems and their impact on SEO has sparked your interest, you might also enjoy our episode "Navigating Agentic AI in the World of SEO" with Andrea Volpini. 

In it, we explore how structured data and AI agents are reshaping the search experience—from the mechanics of modern SERPs to hands-on strategies for SEOs looking to stay ahead.

Also Watch

If this conversation about agentic systems and their impact on SEO has sparked your interest, you might also enjoy our episode "Navigating Agentic AI in the World of SEO" with Andrea Volpini. 

In it, we explore how structured data and AI agents are reshaping the search experience—from the mechanics of modern SERPs to hands-on strategies for SEOs looking to stay ahead.

Why Google Will Still Win the AI Search Race

Gianluca Fiorelli: So, here’s another question. We’re starting to see the search landscape becoming more fragmented again. We’ve got Google, we’ve got ChatGPT, we’ve got Perplexity, we’ve got Bing with Copilot. So why is Google going to win?

Mike King: Oh yeah, that’s easy—they’re going to win for a variety of reasons.

First, they’re not beholden to Nvidia. They have their own chips, so they don’t have to spend a fortune worrying about whether they have enough computing power or capacity. They’ve got what they need. It’s just a matter of whether they have enough data centers and cooling infrastructure to support it all.

Second, they invented the technology. And they’ve brought back many of the original people who worked on the Transformer architecture. They’re all in on this. Google also has an enormous team—like 100,000 people. So they’re going to move faster than most others.

They have multiple products with over a billion users. That gives them a massive advantage: they can plug Gemini into those products and capture learning signals at a scale that OpenAI or others simply can’t match.

And now you’re seeing them start to leverage personal context. They’re tapping into user data from across their ecosystem—which creates a real moat around the capabilities of their models. They have access to data no one else does.

OpenAI is never going to have the scale of user data that Google has across all its products.

And then there’s the fact that they also have all this search data, right? What was it—something like it would take Bing 17 years to collect the same amount of search data that Google gathers in 13 months?

That means Google has so much more data to refine and improve its models. 

The only way Google loses is if it gets broken up by the U.S. government. And yeah, the government is trying, right? If Google were to lose all that data coming from Chrome, that would definitely hurt their ability to refine things. But even then, they have other ways to supplement that data. So no, I don’t really see a world where Google loses—unless they get broken up.

Measuring SEO Impact in the Age of AI

Gianluca Fiorelli: And one last question. In this new AI-driven search landscape—how can we really measure things? Let’s be honest: even though we weren’t given all the data for many years, we still had enough to demonstrate the value of our work.

But now, how do we do that? What are the metrics you recommend we should be paying attention to?

Mike King: Yeah, I think there’s a lot we can consider here, but the most important thing to understand is the brand opportunity. What kind of brand awareness are you creating?

So I’m telling people to start with this: for your non-branded terms, is your brand being mentioned? Is it being cited? That’s one of the most powerful indicators we can track right now.

Then, when you combine that with impression data, you start to get a kind of “mindshare” metric—similar to what you’d look at in display advertising.

Beyond that, there are other important input metrics you should be paying attention to, like: how relevant are our chunks for a given query? Is the page even indexed? How often is it being requested by ChatGPT or recrawled by Google?

These are signals that help you understand whether Google is actually seeing and considering your content. So yeah, there’s a lot of different metrics to look at—but I’d say branded visibility is the key one. And then we need to ask: how much traffic is coming from these different surfaces? Specifically for AI Mode and AI Overviews, you’re not going to get a clean number from your analytics platform. But you can still get directional data—like how many visits are coming from links that include scroll-to-text fragments, for example.

So overall, it’s a lot to consider, but the main things are going to be: how is your content performing in the channel? What sort of visibility are you getting? And how is that traffic driving from the channel, and then what is it doing?

In a way, it’s the same framework we’ve used in classic SEO, just with a new set of metrics tailored for this evolving search.

Want to track how your brand shows up in AI-generated answers

Advanced Web Ranking lets you monitor citations in LLMs—for both branded and non-branded terms—so you can see exactly how large language models like ChatGPT, Perplexity, Gemini and Claude are talking about your business. 

Give AWR a free try and start seeing what the models are saying about you.

Want to track how your brand shows up in AI-generated answers

Advanced Web Ranking lets you monitor citations in LLMs—for both branded and non-branded terms—so you can see exactly how large language models like ChatGPT, Perplexity, Gemini and Claude are talking about your business. 

Give AWR a free try and start seeing what the models are saying about you.

Want to track how your brand shows up in AI-generated answers

Advanced Web Ranking lets you monitor citations in LLMs—for both branded and non-branded terms—so you can see exactly how large language models like ChatGPT, Perplexity, Gemini and Claude are talking about your business. 

Give AWR a free try and start seeing what the models are saying about you.

The Fireside Questionnaire

Gianluca Fiorelli: Okay, okay—I mean, I could keep talking with you for hours, but we’re nearing the end of our conversation.

Before we wrap up, though, I want to ask you just a few personal questions—so that people can get to know you a bit better, beyond all the SEO and AI knowledge you’ve already shared.

Let’s do it this way: Mike, you’ve been traveling a lot recently—and as your friend, I’ve been following along on Instagram and so on.

I saw you visited the pyramids in Egypt, then went to Dubai, and then to South Africa. You mentioned how happy you were to finally check off some of the places on your bucket list.

So tell me: what are the places you’re missing?

Mike King: Yeah—so Japan, Brazil, and the Maldives are the last ones on my top five list of places I haven’t been yet. Once I get through those, I guess I’ll have to come up with a new list!

I've been to so many places, but there are still a few that I really want to see and just haven’t gotten around to yet. This recent trip was a chance to finally check off a few of those.

And actually, I’m planning to go to Japan later this year—late September or early October—for the Found conference. So three of those are definitely happening in 2025.

Gianluca Fiorelli: Very cool, very cool. So let’s do a little time travel—if search had never come into your life, what do you think a younger Mike King would be doing?

Mike King: What would I be doing...Well, my first dream was to be a cartoonist. I wanted to draw comic books.

Then that shifted into wanting to become a programmer—I wanted to make games. And both of those were things I actually did as a kid, just for fun.

I think if I hadn’t transitioned into music, I probably would’ve ended up in Silicon Valley. I was already doing internships out there at the time.

Then I got into music, and eventually transitioned into SEO. But I think at some point, I would’ve realized that the musician life was too small for me, and I probably would’ve pivoted back into engineering in some form—or at least into web development.

So yeah, I think I would’ve ended up in this world one way or another, even if it wasn’t through SEO.

Gianluca Fiorelli: I remember a few weeks ago you shared one of your drawings—you were trying to create a comic strip, and I think you mentioned struggling to maintain character consistency.

Mike King: Yeah! I was just experimenting, like everyone else, “Hey, I did this sketch something and turned it into something with generative AI”. I was trying to figure out how to do that well, because honestly, if I could, I’d probably get back into drawing comic books. It’s just such a time-consuming process otherwise.

Gianluca Fiorelli: So we’re going to get the very iPullRank Cinematic Universe!

And one last question—you made your debut on the search conference scene this year, and it was a great experience. So… are you already planning next year’s edition?

Mike King: Yeah, definitely. For next year’s SEO Week, the goal is to go bigger and better.

I mean, we had Busta Rhymes this year—one of the greatest rappers of all time—so topping that is tough. But we’re aiming for even more.

We want to make the conference itself bigger: more space, more attendees, more great content. And we’re thinking about how to integrate more of New York into the experience—maybe host something at the Empire State Building or other iconic spots, so people really get a feel for the city.

Gianluca Fiorelli: That sounds fantastic—and very promising.

Mike, thank you so much for joining me today. It’s been a real pleasure. I hope we can do this again sometime soon, especially as things keep evolving in the world of search and AI.

Mike King: Absolutely, Gianluca—it’s always great to see you and spend time. Thanks so much for having me.

Gianluca Fiorelli: And thank you, dear listeners—and viewers!

I know, I know… I sound like every content creator at the end of a video, but don’t forget to subscribe to the channel, ring the bell to get notified about new episodes, and obviously—you can’t not give a like to an episode with Mike King!

Bye-bye, see you next time!

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.

Share on social media

Share on social media

stay in the loop

Subscribe for more inspiration.