In this high-impact episode of the Page 2 Podcast, Jon Clark and co-host Joe DeVita are joined by Mike King, founder and managing director of iPullRank, to talk about the seismic shifts happening in search right now. Mike unpacks his unfiltered reaction to Google I/O, breaks down his concept of Relevance Engineering, and explains why rank tracking is effectively dead. We discuss AI Overviews, query fan out, synthetic queries, the role of brand impressions in modern SEO, and the urgent need for new KPIs, tooling, and frameworks. This is a must-listen episode for navigating an increasingly AI-first search landscape.
https://page2pod.com - In this must-listen episode of the Page 2 Podcast, Mike King (founder of iPullRank) returns for a deeply technical and philosophical conversation about the future of search.
From the fallout of Google I/O to the rise of AI Overviews and the evolving utility of tools like Profound, Mike breaks down what’s next for SEOs—and why the entire playbook needs to change toward .
Learn why impressions are the new KPI, how to optimize for synthetic queries, and why relevance engineering could be the only way forward.
🔍 In this episode:
• Mike’s raw reaction to Google I/O and the existential threat to traditional SEO
• The concept of “relevance engineering” and why SEO needs to evolve
• Why impressions and brand awareness are becoming core SEO KPIs
• Google's “query fan out” technique and how it reshapes keyword research
• Why rank tracking is dead and new metrics like visibility and citations matter more
• The case for treating search as a branding channel, not just a performance one
• The future of Google Search Console and what it’s still missing
• The gap between current SEO tools and what practitioners actually need
• How AI mode, reasoning, and semantic triples influence optimization
• Mike’s prediction: in 12 months, the default Google experience will be AI-first
Mike King offers a reality check and roadmap for SEOs ready to embrace change—and lead it.
🎧 Listen & Subscribe:
Follow and subscribe to the Page 2 Podcast wherever you get your podcasts. Don’t forget to leave a review and share with your team.
📎 Mentioned Links & Resources:
• iPullRank – Mike King’s SEO and content strategy agency
• Profound – Mike's next-gen SEO platform built for the AI era
• The Future of Search: A Recap Discussion of Google I/O
• Relevance Engineering Framework – Deep dive into Mike’s technical methodology
• Mike King on LinkedIn
• Mike King on X / Twitter
Jon Clark (00:00)
Welcome to the page two podcast. I'm Jon Clark joined as always by my cohost and partner in crime at moving traffic media, Joe Devita. Our guest today is the kind of person you might call digital Renaissance man, part SEO innovator, part rapper, part agency builder, and now a major conference host. He's the founder of iPull rank an agency known not just for great results, but for pioneering new frameworks like relevance engineering, which I hope to talk about a little bit today, and always pushing boundaries in SEO. I've known Mike King for more than 10 years.
And it's fair to say he's never been short on ideas, energy, or the ability to move the industry forward with a controversial opinion. It's been way too long since we've had the chance to sit down and really dig into the weeds. So today's conversation promises to be a fun, no filter ride. But before we jump into our conversation with Mike, just a quick reminder, if you enjoyed today's episode, please take a moment to subscribe to the show and leave a rating and review wherever you're listening. It really helps us keep delivering great content.
Welcome to the show, Mike. Before we dive in, why don't you give our listeners your official elevator pitch, who you are, what you do, why you do it.
Mike King (@iPullRank) (01:02)
Yeah, I'm Mike King and I do relevance engineering, is a function of content strategy, artificial intelligence, information retrieval, UX and digital PR.
Jon Clark (01:06)
You
That's awesome. I had this whole idea of like asking a whole bunch of questions that I've always wanted to know, just sort of knowing you this long. And then both Joe and I independently watched the video yesterday, sort of the future of search with you and Garrett Sussman. And we sort of left that conversation. Our heads were sort of spinning. One of our SEO directors was also at Google I.O. and she sort of came back with this
Sort of like sense of, I don't know if despair is the right word, but I sort of like picked it up in your tone as well a little bit in that conversation. Yeah. Like just unsettled maybe is a better word. And so I almost want to shift the conversation, not necessarily as a part two of that, but maybe, maybe a part two and really just sort of dive in a little bit more on the, the outcome of Google IO and sort of what that
Joe DeVita (01:44)
almost overwhelmed, a little overwhelmed.
Jon Clark (02:05)
means for our future in this industry. And one thing that really sort of struck struck me was the comment by one of the the Google folks that was basically like, continue to create great content, all and make sure it's really unique. And then there's sort of this flip side of on the publishing side, the free exchange of sort of
clicks for content has sort of gotten blown up, right? And so.
Like how do you see that evolving? Like what is the incentive now for publishers to continue to create that unique content when they're not getting clicks and returns? That makes sense? Like, I feel like that's like, we, we, we all know so many people in publishing. It's just such a tough thing to swallow.
Mike King (@iPullRank) (02:47)
Yeah.
Yeah, I don't really have an answer for people that are in publishing aside from like diversify your traffic mix. Like, you don't, you're not going to get as many clicks. And it's not to say that you won't get any clicks, but you should probably focus more on Google discover and Google news and things like that, rather than, you know, standard organic. And that feeling of despair like is real, right? Like, that's how I felt.
the after the first day of IO, even though I was around some great SEOs and talked to some engineers and had some great conversations, I felt kind of like existential dread, not for myself specifically, but for our industry, because our industry is not ready. Like, what am I going to do? Go log in to like one of these SEO tools that doesn't do anything to support where we're going?
And so it's a really frustrating situation, but that also means that there's a lot of opportunity to help one define it and also figure out like where are the edges that we can play in? And so I've been doing a lot of research on the idea of reasoning and how you can appear when a model does reasoning. And to some degree it is still like semantic triples and structuring your content in certain ways.
But I think we need more of like I showed at SEO Week, we have this like AIO simulator where it's like, cool, you put your content in, you see how it would be perceived by these sorts of mechanisms, and then you can make adjustments based on how it reacts to your content. So I think ultimately that is where we're gonna need to play more in.
And one of the big concepts that they talked about is this query fan out technique, which is where they generate a bunch of these synthetic queries and they run them in the background. And so I think the place where we need to start as an industry is like, cool, how do we understand what those queries can be? Because those queries aren't necessarily just a function of like, what are users doing? It could be what the model itself thinks is a good, sub query for this query.
So I think this is all like replicable to some degree, but the problem is we won't necessarily have a good feedback loop to say like, are these the right synthetic queries? Which means it's just gonna have to be a lot of testing to figure that sort of stuff out. So there's a lot that we can do here as far as like research and understanding and so on, but I don't think any of that is really SEO as it's been historically defined.
Joe DeVita (05:24)
Do you think that clear to stand out research
project is an evolution of a standard keyword research that an SEO team might do?
Mike King (@iPullRank) (05:35)
I mean, it could be, but it requires a skill that most SEOs don't have because it's not just like looking up and identifying keywords. It's also looking at what ranks for those keywords and then breaking down those pages into vector embeddings to figure out like, okay, which of these components is most relevant to the synthetic keyword? So there's no tooling to allow you to do that. Like, yes, if one of these SEO tools gave you some sort of like matrix
tool where you put in the keyword and then it identifies the subqueries and then it pulls all that data for you and then allows you to edit things, then sure. Then I guess it's SEO. But because that doesn't exist and we've got to figure this stuff out, we're still in a space where we're talking about engineering to get to a space where we can actually optimize for this stuff.
Jon Clark (06:27)
Yeah. I I think your background as a technologist sort of sets you up well for this sort of, I don't know, evolution in the space, but I'm curious, like, you know, you had a great picture on LinkedIn of sort of like SEO bro, right? At Google IO. I'm curious for others, that were there, like what were their reactions in the room? Was it similar to yours? Was it, you know, even deeper despair, just sort of maybe not understanding the, the technology?
Mike King (@iPullRank) (06:52)
the
Jon Clark (06:54)
technological connections of how you might tackle something like this or just sort of curious like what.
Mike King (@iPullRank) (07:00)
Yeah, don't know what everyone else's reaction was really. think we were all just kind of taking it in. And, you know, like my feeling of existential dread was just like me sitting with it after I left dinner with everybody. You know, I think we were all just kind of kicking around ideas and reacting to some of the discussions we had with different Google engineers and things like that. But I don't think we had collectively drawn conclusions yet.
Jon Clark (07:06)
Mm-hmm.
Mike King (@iPullRank) (07:24)
But a lot of people that were there were also very technical. had Brittany Muller was there, JC was there, like all people that talk about Python SEO. So I think that there's going to be a lot that comes from the group that was there that's going to be the things that a lot of us look at and say, okay, here's where we go moving forward.
Jon Clark (07:46)
Yeah, I remember, I guess about a year and a half ago at Brighton, I think we were concerned at that time about what's the best tool that can actually report against AIO results. And I think you mentioned Nozzle at the time. I'm curious, how are you thinking about tools in this future space? The traditional ranking reports, I think, are quickly getting
blown up, right? In terms of a way to really evaluate like visibility. So what does that mean for, for tool sets?
Mike King (@iPullRank) (08:17)
Yeah.
Yeah. So as far as AIO specifically, feel like Profound does the best job. And I think, you know, all the other ranking tools are doing something similar, but I think what Profound is doing is thinking less about ranking and more about like citations, visibility, share of voice, things like that. And also giving you that data directly so you can see what was actually there. And then after that, think zip tie is probably the second best and probably the better option if you're like,
You know, an SMD or something like that, but from my perspective, feel like profound is really leading the way. And a big reason for that is they're not SEO guys, they're engineering guys. So they're, solving the problem without the context of SEO and they're not limited by, you know, what the space has historically been doing. So they're looking at what state of the art, whereas in SEO, we keep leaning on like what came before and this idea that.
you know, SEOs can only really deal with like a score out of a hundred or out of 10 and, you know, optimizing towards that. So, I think that rank tracking in the AI mode space doesn't even make sense because we're talking about personal context. We're talking about memory. We're talking about reasoning and so on. So every given user will likely have very different results. So.
You know, rank tracking, I mean, we take a step back rank tracking itself kind of doesn't make sense either because you're trying to do this like depersonalized version of the ranking, which is a function of like you send a machine and act like it's first day on the internet. And the first thing that it does is search for your query so that there's no like user contacts. And you try to, you
limit the location-based personalization as well. So you kind of look at either United States or the centroid of a city or something like that. And that's nobody's user context. So it's like, it's artificial to begin with. So I think that there needs to be, you know, a new look at like how we do measurement in general, because we're not getting anything from GSC, at least not now.
Jon Clark (10:10)
Exactly.
Mike King (@iPullRank) (10:32)
And, you know, that also kind of begs the question of like, what should Google Search Console be? And that's something that I've been thinking about, you know, pretty heavily. I'm like, well, maybe I should just write about like what it is that we would want as SEOs and relevance engineers and so on, because some of those reports don't make any sense. Like, when was the last time you ever looked at that links report and used it? You know what mean? Or.
Jon Clark (10:53)
Christ.
Mike King (@iPullRank) (10:56)
Or think about the the crawl data reports right like it's just telling you the volume of crawls what would you even do with that. So there's so many aspects that are like and I was having a conversation with a Google engineer I was like when are you guys going to get some real resources on this thing like. Like it's clearly the the last product in the organization that you guys are thinking about and.
You know, then they're like, well, it's free. We don't get any money out of it. I'm like, well, Google Analytics is free. you know, before GA4 was a thing, GA was a great product. was like a very, it was an awesome product that gave you all the information that you needed. And, you know, when you think about some of the obvious features that GSC should have, like, I think we just got annotations, like, a couple months ago. Why was that not one of the first things they put in the product? So, yeah, I don't, I don't.
Jon Clark (11:30)
Awesome.
Mike King (@iPullRank) (11:50)
I don't think we're ready for most of this. We're not ready on a measurement perspective. We're not ready with the tool set. We're not ready with the strategic thinking either.
Jon Clark (11:59)
So maybe to transition into the reporting side, think.
Mike King (@iPullRank) (11:59)
Boom.
Jon Clark (12:03)
You know, so what is the KPI now? Is it impressions like to sort of lean into the, you know, SEO as a brand channel sort of thing. And, um, I think you had a comment somewhere where, you know, if someone does a search for a pair of Nike shoes and Nike shows up, you know, even if they don't necessarily get the click, that impression is still, there's still value in that. Right. Um, so is.
Mike King (@iPullRank) (12:25)
Yeah,
point was more that not if you're searching for Nike shoes, if you're searching for, you know, the one I keep using is best basketball shoes for someone with flat feet, right? Like the non-branded term. And then you show up in the AI overview or even the featured snippet and it says Nike Air Max 90 or something. Probably not that, because those definitely are basketball shoes. But anyway.
If your brand shows up there for that non-branded term and a user doesn't click, that is still a brand impression. That's no different from, you know, someone scrolling through their Facebook feed and they see an ad for these sneakers and they may not click on that. But when they go to the store, they may say like, Hey, I heard these shoes were great. Like there's a aspect of mind share that we don't think about in SEO because we just don't really respect the idea of impressions. But.
Impressions is how pretty much every advertising channel works. And of course, they drive clicks, they drive conversions as well. But those are secondary things for the brand channels. So what I'm saying is that search has always been a brand channel. We just haven't taken the credit for that. And now it is more of a brand channel because there's less clicks through to websites.
That's not to say that there aren't clicks to websites. There's not to say that there won't be conversions. There still will be. It's just going to probably become like, you know, 70 % of a performance channel and 30 % of brand awareness.
Jon Clark (13:58)
Got it. I'm curious, like, are you having these conversations with clients now? What's the reception on sort of that shift? Because I think one of the challenges with our industry in general is we've sort of trained people who are reviewing reports to like look at an individual keyword ranking, which has always been like not the best way to evaluate things or, you know, some other metric that isn't perfectly tied to what's actually
attributable to SEO. what's the, what's been the feedback from clients as you're starting to nudge them even further down this path?
Mike King (@iPullRank) (14:32)
Yeah, we actually started with this probably at the top of the year because, you know, once we saw how things were progressing with AI overviews, we were like, we got to raise the flag here because, you know, clients are, like you said, they're measuring based on traffic. They're measuring based on rankings and so on, share voice and all that. And when they see these measures going down, if you don't have the story for like what's happening here. So, you know, we, had to get ahead of it because Google wasn't giving us.
Jon Clark (14:43)
Right.
Mike King (@iPullRank) (15:00)
I feel like the right thing that I've done here would have been for Google to say, hey, search behavior is changing because of this thing that we're rolling out, like a usability study or just some sort of white paper that was like, things are changing. But we didn't have that, so we effectively had to create it ourselves from what we're seeing across clients, what we're seeing across the space in general and so on.
And we were warning clients upfront and we were saying, Hey, you know, branding is a thing here. here's how you get into AI overviews. So we proactively put that in front of people. And so with the launch of AI mode, we've been, you know, kind of doing the same thing and saying like, Hey, here's what Google is showing us. Here's how user behavior is changing. But the reaction has kind of been slow, right? It's been like, well, what do we do to get traffic in the meantime? Because.
We are still in a space where traffic is still being driven. So how do we do that? And it's largely been like, how do we go more to the mid-funnel queries or the lower-funnel queries and improving our rankings there? And luckily, I predicted this two years ago, right? Like there was a blog post I wrote where I talked about how AI mode works and the AI overviews work. And I predicted that there would be like a redistribution of search volume.
Jon Clark (16:00)
Right.
Mike King (@iPullRank) (16:23)
And then I also talked about like, here's how we mitigate. first step is like seeing, your keywords have AI overviews on them? Excuse me. Which keywords don't, and then figuring out like, do we actually need to be in the AI overview because we want that brand awareness or do we want to rebalance our keywords to go after the keywords that don't have them? So it's something I've been saying to clients for a while, but they've been like, cool, cool, cool, cool. When we get there, we'll deal with it.
Jon Clark (16:51)
Perfect.
Joe DeVita (16:52)
You had
Mike King (@iPullRank) (16:52)
haha
Joe DeVita (16:52)
a great quote, Mike. You had this quote about, you got to live for the moment, but plan for the minute. And I think that's a good summary of what you're saying. It's like,
Mike King (@iPullRank) (16:59)
Yeah.
Joe DeVita (17:01)
Organic traffic is still very important today, but it's going to become less and less important every month into the year we go. And 12 months from now, it's going to look completely different. have to, parallel things have to happen. We've got to hold on to what we have, organic traffic today, but we have to plan for something else that's inevitable, right? And it's coming out of past.
Mike King (@iPullRank) (17:07)
Yeah.
Yeah.
Absolutely.
Absolutely. I would be shocked if next year at IO, they didn't say like AI mode is now the primary version of search. And even if they don't do that, they've been very clear in like the documentation around AI overviews and AI mode and saying like, hey, the things that we pilot in AI mode, we are ultimately going to bring to core search as well. So either way, it's going to be a dramatically different environment in a year from now.
Jon Clark (17:51)
My, wife, Jamie works in, in this industry as well, and she works in-house. So maybe a little bit easier for her to sort of drive how reporting goes, but, probably about six months ago, she shifted the way that they evaluate traffic into paid traffic and non-paid traffic. direct organic, right. Referral, basically everything that's not in, trackable from a paid perspective.
Mike King (@iPullRank) (18:08)
Mm.
Jon Clark (18:16)
And I think that's one way to maybe start thinking about this, especially with the recent news that Google's not passing any refer data. So it sort of comes in as direct traffic. And how do you even get attributed for SEO in that scenario? You can't. And so you almost have to think about maybe a larger bucket of content, sorry, larger bucket of traffic to be able to.
to sort of show where that traffic's coming in from. So that might be one way to think through how to break that traffic out into more digestible formats.
Mike King (@iPullRank) (18:49)
Two things. One, I love that approach because, you know, it makes you think more about like the channels and the interplay between them rather than being like, well, this channel is dying. Well, so if you're getting a whole lot from social or a whole lot from email or whatever, then you're you're net zero and it's fine. Right. Like as long as it's still performing traffic. And the other thing is that Jon Mueller said yesterday that the no refer thing is a bug. So hopefully that will change. But
Jon Clark (19:01)
Exactly.
Mike King (@iPullRank) (19:18)
You know, right now, again, we just have no visibility into what this thing does. And it's going to be difficult to get investment if you can't prove what it's going to drive.
Jon Clark (19:30)
Yeah, I saw Jon Mueller's comment as well and I'll take it at face value until it's real. but, you know, going back to the comments around Google search console, if say in 12 months, the entire experience is AI mode. and they're not reporting on that information in Google search console today. Like does Google search console exists in 12 months?
Mike King (@iPullRank) (19:35)
Yeah
I mean, I think it'll still exist because, you know, because like, they will probably still give us like the performance, you know, the query data or whatever for whoever is using that for whatever they're using it for. But I suspect that we will see some changes to GSC is probably not going to be what we want. But I suspect we will get something that's more reflective of.
what is actually happening. And, you know, there were a couple of illusions to there being changes to GSE coming, but again, it just wasn't clear at all that it's going to be something we could actually use for what we need to do.
Jon Clark (20:34)
Got it. So I guess maybe to transition into relevance engineering is that sort of an answer to maybe some of this loss in traffic, right? Sort of thinking about things much more holistically or sort of how do you position that as a service to clients? Again, it's always hard to train an industry towards something new. And so I'm just really curious about how
how you approach that.
Mike King (@iPullRank) (21:00)
I don't know that I'm trying to train the industry towards something new. I think that we will always have people that are just going to be like, it's just SEO and what have you. And I think that those people will adapt small things to try to remain relevant and so on. But for me, I think it needs to be like a wholesale rethink of all of this. And I also think that the interplay between search channels is going to be really important, right?
Jon Clark (21:08)
Mm-hmm.
Mike King (@iPullRank) (21:27)
you know, chat, GBT, perplexity to AI mode, to classic organic or discover, like all these things have different mechanisms that you need to understand in order to get more visibility into them. And ultimately what ties it all together is the idea of relevance, right? Like measuring what is relevant for a user, what their user interactions tell you about relevance, and also just the actual relevance of content to queries. So for me, it's like, okay,
there's a lot missing from what SEO needs to be doing based on what we know. Like think about all the stuff we've learned about Google in the last year from the DOJ antitrust trial testimony, the leaked documents that I covered almost a year ago today. Actually, yeah, was like a year ago today when I got the document. And then, you know, also the...
exploit that Mark Williams Cook found where you could see like all the IR metrics and so on. So I can't in good conscious go back to just like, it's content links in 301s when we have so much information about the telemetry of how this works. And so from my perspective, it isn't about SEO because SEO is just like an abstraction. Like when you do SEO, you're really just like a power user of Google.
Jon Clark (22:33)
Alright.
Mike King (@iPullRank) (22:49)
And you're like, okay, I know how to do my site colon searches. And I know like, you know, when I do this thing, Google does this, but modern search requires more knowledge than that from my perspective, right? Like you've got to understand how large language models work. That's not SEO. That's AI. You know what saying? I'm not going to sell you my knowledge of AI for the cost of SEO. It doesn't make any sense.
So this is also like bigger picture thinking that I've seen the SEO community really be like capable of. Like SEO is not a strategic industry. It's like, it's super tactical. It's like, okay, well, you know, tell me the five steps I need to do to do this thing. Otherwise no one is happy about your talk, right? Like this idea of being more of a branding channel and less of a performance channel is a strategic change.
and how we need to approach this stuff. And there's going to be a lot of pushback on that from our community because they don't want to do anything different. They don't want to have to be like, well, impressions matter when they spent 25 years saying that they don't. So if you're thinking about this in the context of how businesses operate, they are going to say, OK, well,
Here's the top layer of how we're trying to accomplish this thing. And then what tools do we use to accomplish the thing? Whereas SEO is just like, okay, well, what levers do we pull to be number one? And it's a bigger idea than that. We need to be thinking about, me, what is the interplay across these surfaces for the user? How do you tie all this together? Because so many users will, yes, maybe start in chat, you can see.
but then go to Google or vice versa. And being able to figure out the ways to tie all this together isn't something you can do by just sitting in your silo of SEO. So relevance engineering is like, cool, how do we like look at this across everything and also position ourselves to have the right level of control of things to be effective. Whereas SEO is always like, hey, UX guys, can we talk? Like, no.
Jon Clark (25:07)
you
Mike King (@iPullRank) (25:07)
I want to come into the discussion and be like, here's how you do UX for this environment. Not like, know, well, I got to talk to this person and talk to this person and talk to this person. And then six months go by and then something new rolls out and we never got the opportunity to make something happen.
Joe DeVita (25:24)
Mike, you is SEO a piece of relevance engineering or is relevance engineering just like optimizing for AI?
Mike King (@iPullRank) (25:34)
No, I think that, again, like SEO, like the way it's defined is smaller than what I'm talking about. Like I'm more talking about information retrieval, which is of course the computer science behind what search engines are doing. And so rather than saying like, well, Google's guidelines say this and da-da-da-da, I'm more like, well, based on how the systems function,
It is likely looking at X, Y, Z. And so we need to manipulate A, B, and C in order to do this thing. So I'm never going to be the person that's going to be like, oh, it depends. I'm going to be the person that says, you know, based on how this system works, these are our three options. And I feel like that has been way more effective in doing what I do than, you know, just heavily leaning on what Google has defined about how things work.
So as an example, Google put out two documents of guidance in the last couple of days where they basically said, hey, create great content and use structured data. That's not enough for me. What is more valuable to me is understanding the concepts behind query fan out so that we can replicate that and then we can have something actionable to do across a series of keywords to then create a surround sound situation where
no matter where Gemini goes, it is gonna run into my content. So it doesn't have so many options when it's generating its response. And again, that's not defined in SEO yet. Whereas I'm able to say, based on what I understand by how the system works, this is a tactic that I believe will work based on the technical underpinnings of this system.
Jon Clark (27:16)
Almost like it, almost like this has to work because this is how the machine works. you're, you're sort of. Just no longer. It depends to your point. It's like, this is literally how it works. And we're just applying things to that process. I think, uh, uh, I think that is scary for a lot of people, right? Because they're working off the same checklist or the same process that they've developed. And that's how they've scaled up to, you know, 150 clients that they charge $50 a month for, you know what I mean? And so now.
Mike King (@iPullRank) (27:21)
Exactly.
Mm-hmm.
But
Jon Clark (27:46)
you know, that process gets blown up. so maybe that is definitely a good thing. I'm curious, like how, so I think one of the things that you've been known for historically has always been like, obviously very technical, but especially being able to come in and, and execute a technical audit for, you know, an enterprise level company, right? Whole bunch of different websites all working together. I'm curious how you guys are thinking about, technical audits in this age, right? So
Does your technical audit now include a UX component or has it always?
Mike King (@iPullRank) (28:18)
It definitely hasn't always. definitely
does now though, but the UX stuff, we more discuss on our content strategy side. And, you know, I think one of the mistakes that I made was talking about the technical stuff too much, because we actually do way more content stuff at I4N4 than we do technical. I'd say it's probably like a 60-40 split. But nevertheless, you know, when we do our audience, historically it's been...
Jon Clark (28:25)
Go.
Yeah.
Mike King (@iPullRank) (28:43)
technical content, linking, page speed, mobile usability, then user interactions. But again, it's more like the UX and content strategy stuff where we really dig into that. We'll highlight the problems in our audits, but more on the crux side where we're saying, OK, here's what to do on the granular level. And we're really just starting to expand what that offering
is because historically it's been like, you're not the UX guy. We'll shift this over to them and they'll discuss it. And what I'm saying now is like, no, we need to be able to make changes on the UX side. Otherwise, SEO is not going to work, right? Because we have a lot more clarity around the fact that Google does use user interaction signals to reinforce what should rank.
I think that's going to come into play a lot more with these rag models behind AI mode and AI overviews because they have to figure out like which of these pages should we considering more than others when so many of them are saying the same information. So they can figure out like which ones to actually cite. So I don't think we can reliably make these channels work without UX being like a germane component to it.
Joe DeVita (30:05)
you're already shifting your deliverables based on what you've uncovered in the last 6-12 months. Your deliverables have changed. They're no longer the classic, it's not like our, our industry is like 20 years old now. We all do technical audits and content audits and linked audit, but they're going to become less meaningful over time. So
Mike King (@iPullRank) (30:11)
Yeah.
Hmm
Joe DeVita (30:27)
We're going have to create new deliverables. We're going to have to sell new types of work to clients to be relevant. I feel I'm excited about the agency business. I think we're we get into it because we like solving problems. I'm more worried about the tool sets that we rely on because they don't exist yet. I feel like we as agencies will figure out how to make ourselves relevant and help clients. But without having a tool set to rely on is the part that I'm struggling to wrap my head around.
Mike King (@iPullRank) (30:44)
Mm-hmm.
Yeah, my fear is that, you know, the SEO software industrial complex ⁓ is not gonna react fast enough because they have their role maps for the reasons that they have them. And if you look at what they've been doing for the last couple of years, they haven't been like improving the products. They've just been slapping chat GBT on top of the existing products. Like,
Jon Clark (31:03)
Hahaha
Mike King (@iPullRank) (31:21)
When is the last time a link index has given us something new and meaningful about the link graph? And last year, we learned specifically that Google applies a sliding scale of link authority not just based on page rank, but where the page lives in the layer of the index. So if it lives in the Fresh Docs, which I'm assuming is like the big publishing sites, then that's like a higher value.
If it lives in the top tier, which is in memory, which means that it's like being accessed a lot, it ranks very highly, likely driving a lot of traffic. It's driving the most link equity. And then there's like two other tiers as well. And so we can use proxies like how many, you know, page one rankings does a page get or how much traffic does it get to determine where it likely lives in the index. But there has been no like new metrics around.
It's been a whole year. But instead they're like, cool, let me give you more chat, GPT and our bad data. I don't understand why we are so behind. And I keep harping on the fact that Google, excuse me, has been doing both semantic and hybrid retrieval for a number of years now, and none of our tools are aligned with that. So it's like, you are optimizing around an obsolete way of
how search works. So like we are just so behind and I think it's just gonna require some new player in the same way that Profound is coming in where they're like, cool, forget all that. What is the best technology we can build for us to be able to level up? So I don't know, maybe it'd have to be me. Maybe I gotta build my own software for that.
Jon Clark (33:01)
Hahaha. ⁓
I mean, think it's, I think it's amazing. Like Ahrefs for example, like they, introduced the ability to automatically create meta descriptions with chat GBT, which, okay, cool. But like, what about using all the link data that you have and applying some sort of AI technology there to suss out? Like, I don't know, key insights from the link graph of competitive sites or whatever. seems like a much more.
Mike King (@iPullRank) (33:16)
You
Jon Clark (33:30)
strategic use of the data that they have. But I think we saw that a ton with when ChatGPT came out, know, investors, stockholders, right? Everybody wanted to see some like element of AI involved in the tools. And so that was just sort of the easy add on. There wasn't really, it seems like a lot of strategic thinking about, okay, we could add this, but here's how to make it really valuable.
Mike King (@iPullRank) (33:52)
It seems to me if we know that vector embeddings are underpinning all of this and you're crawling.
set of the web already, why would you not also create an index of vector embeddings and make that available to everybody? If we know that the relevance between a source and target document is something that Google is looking at very closely and has the ability to do so very easily now, why do we have no relevance metrics in our link here?
the base level things that our space should have that we don't have. And so we have this whole subset of our industry that does like, you know, Python SEO that is completely a function of the fact that our tools don't do enough.
Jon Clark (34:39)
Yeah, it's true. I will say, building, being in the SaaS business, very different than the service. Well, I don't know. Maybe you could argue that's a little bit of service. the maintenance and the costs involved with that is highly different. So totally get why that might not be something you want to take on. But.
Mike King (@iPullRank) (34:57)
Yeah, for sure. But I mean, you know,
the crawling itself is already expensive. You know what mean? And so like just running an additional operation, not saying that it's free, but you might as well because it is so valuable. You know, and like, why does it take me to say like, hey, vector embeddings are a thing, and here's how you do it, and here's why it's valuable, and here's all the use cases. Like I shouldn't have to be the one that's defining that for an industry. That is like,
Jon Clark (35:01)
Yeah, yeah.
Mike King (@iPullRank) (35:25)
true in the world. You know what saying? Like it shouldn't be me having to bring that back to our industry. And what I do love is that the more agile software providers, I mean, it's really only one, Screaming Frog, like they are adding these sorts of features consistently. And then what happens is, you know, the other companies start to come in and be like, well, everyone is using that. We need to have it too. So it seems like the pipeline is like,
me is I'm the industry's product manager and then screaming frog says that's a good idea and then everyone else copies that.
Jon Clark (35:58)
Yeah, I mean, I've been playing around with some scripts too, and it's not hard. Like, it's not hard at all. And then the insights that you get, you're just like, like, yeah, like, this is great.
Mike King (@iPullRank) (36:05)
No!
Yeah, it's super valuable. you know, like, what we're talking about a scale, right? Obviously, it's going to cost them millions of dollars to do this. And I don't discount that. But it's like an absolute requirement at this point. can't keep playing with incomplete data when Google has continued to make these quantum leaps ahead. Because otherwise, what are we actually doing? You know, we're just like.
Joe DeVita (36:37)
It's all very manual right now. I wonder if you find any value in like reverse engineering. use, you use chat GPT, you ask check GPT something, you see the answers it provides. You try to understand why is it, why are those answers there? You reverse engineer from there. It feels like it's all very manual, but what else can you do without a tool set right now, aside from trying to pick it apart yourself?
Mike King (@iPullRank) (37:03)
Yeah, I mean, a lot of what I'm doing is effectively reverse engineering, right? Like, so when SGE, you know, the precursor to AIOs was first announced, I didn't have access to it. So I was like, cool, well, let me figure out how I could build something that does this exact thing. And then I came across Lama index, and then I basically like use like a SERP API to basically feed the results from a given query.
to Llama index where it basically vectorizes them. And then you have a prompt that says, know, answer this query with the information from these documents. And I got really close to what Google is actually doing. What I didn't know at that time was that they were looking at all these synthetic queries, which we've learned as a function of this whole query fan out thing. So I think one of the wrong conclusions that I made
early on in that proof of concept that I built was that, oh, you could rank like number 75 for this query and still be a part of the AI overview. I think what was actually happening is that something that ranked number 75 for the query you put in might have ranked number three for one of the synthetic queries. And so I think that is what is one of the key insights. So going back to what we talked about earlier in this conversation.
is building that matrix of keywords to figure out where are you ranking across that full keyword set and figuring out which components you need to optimize accordingly in order to appear in these different surfaces.
Jon Clark (38:39)
So is
it fair to say that the hub and spoke model that was popular previously is sort of a way to get to that, right? Like you have a topic and then you have all the subtopics around it, but there may also be these sort of adjacent topics that also now need to be included. And so you're just sort of building this content hubs essentially.
Mike King (@iPullRank) (38:48)
Yeah. ⁓
Mm-hmm
Yeah, it's still to some degree topical clustering, but it's not necessarily going to be identifiable in the same way. Because I think when we talk about topical clustering, we're like, I have this core keyword. then there's like, you know, like let's say we're talking about Disney World planning, right? I know this is an example that Simrush used. Disney World planning is like, OK, you know, what restaurants are at Disney World, which rides do I want to get on? we have all this sort of stuff around the top.
But I think that because Gemini is generating these, it's not necessarily gonna be queries that a user is already searching for. It could be queries that it just like comes up with based on that core query. So that's the key difference. And in fact, I found the patent for how they do the whole query fan out thing. One of the things that they do is they...
They look at the pages that rank for the core keyword and they start extracting features from that. But then a lot of it is like, OK, what are the recent queries that the user searched before? What are the related queries? And then what are the implied queries? And I think the implied queries is going to be the one that's not the standard stuff that we know from topical clusters.
Joe DeVita (40:21)
Can you give me an example?
Mike King (@iPullRank) (40:24)
Yeah, I think it's hard to like give you a direct example right now, but like, but the point is that the large language model can generate a bunch of queries, you know, based on the context that the user is in. So as an example, let's say, you know, your location is New York, and then you're talking about Disney World Planning. Well, part of it will be like, well, what flights from New York do I need to get to Orlando?
So that's not explicitly said, but it's informed by the other data points that they have about the user.
Jon Clark (40:57)
So it's, I think very oftentimes when we think about like, you know, content clusters or hubs or, know, whatever term you decide to use, it's often sort of like the question answer format. Is there, is, would this be an application of like a semantic triples approach where you're, getting much more specific about like the keyword or the topic that you're covering to sort of align with maybe how LLMs better understand like context? Is that?
Mike King (@iPullRank) (41:10)
huh.
Jon Clark (41:26)
a reasonable way to think about that or?
Mike King (@iPullRank) (41:26)
Yeah,
yeah, ultimately what you do is still going to be, you know, using semantic triples, structuring your content into like clear semantic units. So it's easy for these vector embeddings to be, you know, precise and, you know, using headings, things like that. But also metadata is important. Structure data is important. The topical clustering in general, because that's going to inform, you know, broadly the ranking of the site.
and so like content pruning is going to come into play. So some of the things that we already do mechanically as SEOs are still going to be important. But the aspect of reasoning is not something that we're really going to be able to control unless we are figuring out more technically, like what are they doing? So I think there's going to be an aspect of Google retrieves.
you know, all these different chunks and what have you, but then they're going to re-rank them based on different considerations. And so figuring out what those elements, aspects might be, is still going to require more than what we have in our SEO skill set. It seems like so much are going to be personalized. Yeah. Right. So the personal context stuff, you know, the historical queries, like all of that.
Joe DeVita (42:35)
It seems like so much of it is going to be personalized. We'll never have access to that.
Mike King (@iPullRank) (42:47)
is something that we're not going to have direct access to. But I think that one of the things that all these SEO tool providers is going to have to do is start incorporating more clickstream data. So for a given keyword, what is that journey most likely to be? Like, what are users most likely searching for before and after so that we can use that to inform our approach? Because yes, it is mostly going to be like a one-to-one environment. But user behavior isn't so large, right?
If you search for this before, you know, you're looking for a mortgage, you're likely to search for this next and so on and so forth. But the available data that we have and like, you know, the people also ask may not be enough because we do get those, those trees and graphs of the relationship between queries. but there's going to be a lot more long tail queries that may not be represented in there because.
users are starting to understand that Google can understand a more complex query. And there may not be enough of those to build out those graphs in the way that we
Jon Clark (43:52)
I mean, that's really fascinating. So if you took sort of like the user clickstream data, could you create a cohort of people and basically give them a task? you know, we want you to research how to get a mortgage. And then you basically just track all their collective searches. ⁓
Mike King (@iPullRank) (43:52)
Thank you.
Mm-hmm.
Jon Clark (44:11)
There would probably be some sort of general theme that would emerge from there around those types of queries. Could you then sort of, I don't know if train of model is the right way to think about it, but could you then use that data to sort of build out content around it? Like, would that be one way to think about tackling this issue?
Mike King (@iPullRank) (44:32)
Yeah, I mean, in some degree, that's what we already do, right? Like we build these journey maps, we align them to keywords, and then we say, OK, what content needs to be created in each of stages? But my point here is that uncovering that data is going to be a lot more challenging. Like, yes, you can continue to do it broadly based on the keyword research data that's provided. But the point is that these environments are far more specific. So taking the broad approach may not be enough in
Jon Clark (44:59)
Got it. I know we're bumping up against time here. Joe, any other questions you wanted to ask? I do have a curiosity question. I mean, it seems like the best way to get to what's actually happening would be to work internally, OpenAI or Google. Would that ever be in your future?
Joe DeVita (45:06)
Uh... I knew.
Mike King (@iPullRank) (45:26)
You know, it's funny, I was literally thinking about that. I said it out loud.
we went to a couple of the buildings where the actual search engineers are. And I was like, yo, maybe it's time to stop doing this. Maybe it's time to just join the dark side and be a part of building it, not just on the outside trying to manipulate it. And I've gone back and forth on that idea for a long time because my background is computer science.
Jon Clark (45:41)
All the Denny Sullivan.
Mike King (@iPullRank) (45:55)
I grew up just wanting to like build things and you know, maybe SEO is just like too small for me at this point. And maybe that is what I need to do. But right now I'm still having a good time and I feel like this is like a real opportunity to, you know, put my stamp and my mark on what things are going to be moving forward. And I don't want to, you know, lose that opportunity. But yeah, maybe like three, four years from now after I'm done with iPoRank, maybe I do join.
like an open AI or Google or something like that and, you know, work on building the future.
Joe DeVita (46:30)
You got to stick with it one more year. This is going to be the most exciting year in decades. feel like.
Mike King (@iPullRank) (46:37)
Yeah,
no, I mean, the last few years have been crazy, right? Like I started writing my book. don't know. Whenever I started writing my book, it was after a period where SEO had gotten pretty boring, right? Like there weren't any significant algorithm updates. It was just like, cool, we're chugging along. We're doing what we're doing. And then generative AI showed up and it was like, OK, I'm awake again, you know?
Jon Clark (46:42)
Yeah.
I think it's.
Joe DeVita (47:00)
You to a few chapters
to your plan for that book.
Mike King (@iPullRank) (47:04)
my God. Yeah, this book is supposed to be 300 pages. It's 900 pages. you know, every time I'm like, okay, I'm done, something significant happens. So I've decided like, this, I'm not doing any more after AI mode. Like this is it. Here's the book. Deal with it, you know?
Jon Clark (47:21)
Well, listen, Mike, before we let you go, we've started the, I guess we're trying to start a tradition on the show that gives our listeners a little bit something extra. We're calling it the PageTube Podcast Prediction. And I think we may have sort of touched on it already, but in 12 months from now, when you go to google.com, like what do you expect that experience to be?
Mike King (@iPullRank) (47:43)
Yeah, I expect the query box to be bigger, more like a text area rather than just like an input box. I expect it to default to AI mode or at least some like hybrid between what we're currently seeing and what AI mode looks like. And I expect it to be, you know, a lot more multimodal, meaning that the results will be, you know, not just showing like videos and imagery and so on, also generating it in real time.
to give you something very precise and personal to what it is that you're looking for. And I expect that it's gonna be a lot more like the movie Her, where it has all that information about you and it is very reactive to you specifically, rather than giving more broad information.
Jon Clark (48:30)
Got it. And zero clicks.
Mike King (@iPullRank) (48:33)
I mean,
I'm gonna be honest, like I actually don't even click that much anymore. You know, when I see an AI overview, I'm like, that's all I needed.
Jon Clark (48:37)
Yeah.
Love it. Mike King, everyone. Thanks again for joining us on the page two podcasts. And if you enjoyed the show, please remember to subscribe rate and review. We'll see you next time.