The Brave New World of SEO

By Mike King
CEO and Founder of iPullRank

Mike kicked off the inaugural SEO Week and welcomed us with a powerhouse keynote on the future of search: breaking down AI Overviews, vector embeddings, and why traditional SEO is no longer enough. His talk introduced the concept of Relevance Engineering, a bold new framework that fuses AI, content strategy, information retrieval, digital PR, and UX. Now, for the first time, you can watch the full session yourself.

Is FOMO hitting you hard after Missing SEO Week 2025? It's not too late to attend in 2026.

SEO Week 2025 set the bar with four themed days, top-tier speakers, and an unforgettable experience. For 2026, expect even more: more amazing after parties, more activations like AI photo booths, barista-crafted coffee, relaxing massages, and of course, the industry’s best speakers. Don’t miss out. Spots fill fast.

ABOUT MIKE KING

Mike King is an acclaimed international speaker covering SEO, content strategy, and the creator of Relevance Engineering (r17g). Innovating and elevating as the Founder & CEO of the performance marketing agency iPullRank, Mike consults with companies all over the world, from brands like SAP, American Express, and HSBC to a laundry list of promising eCommerce, publisher, and financial services organizations. He’s about to shatter everything you thought you knew about how modern search engines work with the launch of his debut book, “The Science of SEO.

OVERVIEW

SEO Week made history, and we can’t stop thinking about it. It’s finally time! The moment everyone’s been asking about has arrived. 

We’re kicking off the official video release of SEO Week 2024 with the keynote that started it all: Mike King’s opening talk, “The Brave New World of SEO.”

This wasn’t just a welcome speech (introduced by, I’m not kidding, Redman), it was a wake-up call. Coming from inside the house. And everyone in attendance answered the call. 

In front of a sold-out crowd from multiple countries and over 230 companies, iPullRank CEO Mike King laid out a bold vision for the future of search, and the cracks in the system that brought us here. From the collapse of organic traffic to the rise of AI Overviews and agentic search, Mike pulled no punches. 

He challenged the industry to stop treating SEO as a checklist and start treating it as what it really is: a smart, scrappy mix of skills that actually moves the needle on brand, strategy, and results. And with that, he introduced Relevance Engineering (affectionately referred to here at iPullRank as r17g, and yes, we do hope that catches on). 

Relevance Engineering is a new operating system for SEO that goes beyond rankings and gets to the core of how meaning is built and delivered across search surfaces. It’s Mike’s answer to a broken status quo. 

It blends AI, content strategy, information retrieval, UX, and digital PR to help brands thrive in an environment where traffic alone is no longer the measure of success. We’re not rebranding SEO, we’re simply rethinking everything we’ve been doing. No biggie.

And this is just the beginning.

This keynote is the first in a full lineup of talks from SEO Week. Starting today, we’ll be releasing one new session every weekday (aside from holidays) through July, featuring some of the most forward-thinking minds in the industry. Names you know, some you should know, some you’ll definitely know in the future. 

Expect technical deep dives, strategic frameworks, AI research, and unfiltered takes you won’t hear anywhere else.

If you missed SEO Week live, this is your chance to catch up.

If you were in the room, relive the moment and bring your team with you.

This is not your average SEO conference recap.

This is where the next chapter of search begins.

Let’s go.

Recommended Resources:

DOWNLOAD THE DECK

Talk
Highlights

Search is evolving fast: AI Overviews, longer queries, and personalized SERPs are changing how users engage, and old-school SEO tactics just aren’t cutting it.

Relevance Engineering is the future: It’s a new approach that blends information retrieval, AI, UX, content strategy, and digital PR to engineer better results across all search surfaces.

The SEO community needs a reset: Mike calls for more experimentation, open-source tooling, higher technical standards, and a move away from checklist culture to build smarter, data-driven frameworks.

“Do you all really want to stay the janitors of the web? This is our moment to really stand up and be something different.”

Is FOMO hitting you hard after missing SEO Week 2025? It's not too late to attend in 2026.

SEO Week 2025 set the bar with four themed days, top-tier speakers, and an unforgettable experience. For 2026, expect even more: more amazing after parties, more activations like AI photo booths, barista-crafted coffee, relaxing massages, and of course, the industry’s best speakers. Don’t miss out. Spots fill fast.

What’s one thing you didn’t get to share in your talk that you’d add now?

I was in a unique position as the host of the event where I had to go deep, but also leave space for other presenters. I would have liked to have spent some time discussing measurement for the various AI surfaces. That’s an unsolved problem that a lot of people are spinning their wheels on, but we have some great ideas and metric packs that companies should be considering.
 

Has anything since SEO Week changed how you’d frame your talk on AI Mode or SEO today?

Yes, the query fan out technique was not front and center for the SEO community until Google I/O. We now have a much better understanding of how it impacts both AIOs and AI Mode. If I knew more about that then I would have dug into it as a key missing piece for how things rank in AIOs. The idea of the matrixed approaches you need to take to ranking in this surface is a really important component that should change our behavior as marketers.  
 

Transcript

Redman: Hey, yo. Redman is in the building you already know. I got a special announcement. I had to pull my car over for this shit right here.

Now we all know SAYO week is starting off right now. I hope I’m saying that right. SEO, SAYO, S-E-O, SAYO. I’m a just go with SAYO Week.

It’s starting off right now. And y’all hanging with my boy, Busta, at the after party. That’s gonna be lit. Yo, Busta what the f$% is up?

But we ain’t talking algorithms and backlinks without bringing out the king of the ranks. The king of the rank. This dude is a lyrical tech wizard in the SAYO game you already know, turning data into dominance, search into strategy. And, yeah, he’s even most nicer on the mic than most.

Alright? Without further ado, I I need y’all to put y’all hands together and make some noise for the iPullRank CEO, the one and only, that motherf$%er, Mike King. Mike King, what the f$% is up? Yo.

Let’s go. Let’s go. SAYO Week. You already know it’s time to level up.

Time to level up. And who’s gonna bring you to that level? My boy, Mike King. Mike King.

Let’s get it. Everybody SEO Week. Let’s get it. Time to level the f$% up.

And then after that, we all go smoke some weed.

Make some noise, y’all.


Mike King: I’m Mike King, and this is SEO Week.

Where’d my music go?

So welcome to SEO Week. My name is Mike King, and I’m really excited for all of y’all to be here, and there is so much for us to talk about today.

And, yeah, let’s just get into it. Is that alright with y’all?

Cool. Alright. So welcome to SEO Week. Like I said before, brought to you by the greatest SEO company in the world, iPullRank.

And so how did this all happen? About a year ago, there was a random week of events for SEO, after New York not having anything for like six years. And I was like, hey guys, we should make it SEO Week next year. And here we are.

So thanks for coming. Really appreciate it.

I don’t know if there’s a yeah, there’s a bit of delay there.

Next slide.

So thank you to all of our sponsors as well. You know, we couldn’t do this without them. So, you know, we’ve got Last Mile Retail, Yext, Moz, STAT, Previsible, Screaming Frog, AirOps, SimilarWeb, Profound, Conductor, and of course, iPullRank, because we paid for most of this.

Thank you to all of our speakers as well. You know, I’ve got forty of the best speakers in this space. Some folks you may have never heard of from before, but they’ve got so much good stuff to say for y’all over the next four days. So, also I want to thank all of the attendees. We’ve got people from twenty-one countries in this room.

That’s crazy.

We got 232 different companies represented in this room, and we got people that are one hundred percent ready to take search to the next level. And also we’ve got a lot of iPR clients here, and our abbreviation is iPR. A lot of people say, like, iPull or whatever, which is so weird to me. It’s iPR. Anyway, we’ve got a lot of iPullRank clients in the room. So, you know, you can talk to them about what it’s like to be a winner, too.

Y’all almost gave me a heart attack, though.

So these are our ticket sales over time.

The first three months, we sold very few tickets. And as you can tell, this is a very expensive production, and there was a lot more I wanted to do, but I was like, I don’t think anybody’s coming.

I’m about to put together the best show ever, and nobody’s gonna come. And we were like six weeks out. I was like, yo, I think I could get Busta Rhymes, y’all.

But I’m a throw a Busta Rhymes show, and nobody’s gonna come. So in the last couple weeks, that’s when all the ticket sales started to happen. And in fact, we’ve gotten like, it was like thirty so people that have hit us up in the last like twenty four hours saying, can I get a ticket? You can’t, because we sold out.

I’m a say something that you will never hear another New Yorker say ever. Go check out Times Square, because we have a welcome ad for you guys. I’d love for you to take pictures in front of it. Tag it with SEO Week, and then you know, get on your Internet. So what exactly did you get yourself into?

Well, I’ll tell you. Day one is the science. We’ve got theme days here. You’re gonna hear all the technical stuff that you really wish you knew about SEO. Probably more than you want, from some of the greatest minds in the space. Some folks you’ve never seen before, like I said.

And then tonight, we have our Welcome Soiree, which is sponsored by Last Mile Retail in this room or in this building.

Tomorrow, we have the psychology. We’ll be talking a lot about how you need to be thinking about SEO moving forward. And also how users are thinking about search moving forward.

And then we have our SEO Week After Dark. These are our meet ups at night that are sponsored by DemandSphere. They have one where they’ve got, Zach Chahalis, Jori Ford, and Ian Lurie that’ll be speaking.

And then you’ve also got, in the same evening, you have choices effectively, SEO Week After Dark, which is SEO FOMO from Aleyda, and sponsored by the Conductor team. So as I understand, that one may already be oversubscribed, so definitely check out the DemandSphere one. I’ll be at both, and we’ll have a lot to talk about.

Now, Wednesday, we’re talking about what’s this? The ecosystem? Yeah. We’re talking about, you know, all the different types of search. So things like local search, video search, e-commerce, and so on. And again, we’re still thinking about, what do users need to be thinking about this as well.

And then we have our After Party, which of course is starring Busta Rhymes. And I’ll be rapping too, if that’s okay with y’all.

And I got some other friends that are coming too.

And then the last day is the future. We’re talking about the future of search. Where are things going? How do we account for it?

So some quick housekeeping things. One, SEO Week is a hashtag. Two, there’s also a Slack channel. If you haven’t joined it already, please do. That’s where you get the most up-to-date information. We have a code of conduct.

Don’t be an a**hole.

Just had to boil it down for you.

And there’s also no Q & A during the sessions, or after the sessions. So grab the speakers. When you see them, they’ll be floating around.

So let’s know what I will rank. I can’t throw a conference and not tell you about what we’re doing. Right?

So uh-oh. I went too far on that one.

So we got a new website, which should have just launched this morning.

And it’s a lot cooler.

It reflects who we actually are. You know, the website that we had was like, I don’t know. It’s not as bad as like a Bruce Clay website, but it was pretty bad.

Yo, why does that jewel look like Windows 98?

Anyway, so we’ve got a lot of really cool features on our website. It’s really fast. We’ve used some things that Jono Alderson was talking about, like Edge images. We’ve improved our brand positioning.

We’ve got, it’s, like, static WordPress on CloudFlare.

There’s no third party scripts. It’s amazing. It’s incredible. Go look at it and buy stuff from us.

So here’s where we are. This is our knowledge graph, which is representing what iPullRank is at this point. You know, we do SEO, content strategy, and generative AI, of course. And like I said, we’re the best agency for this sort of work.

And this is our team, or a lot of our team. Some of them are here. Team, make some noise y’all.

Our team is amazing. You should get to know them while you’re here, because like I said, they’re amazing. Just keep looking at them. They look great.

So, you know, we’re hiring. So if anyone is looking to join the best team in the world, feel free. We’re hiring for SEO director, SEO engineer, marketing data scientist, and also a sales development representative.

So in addition to that, you should be a client if you’re not.

We would love to work with you. And we’re very excited that you are here.

So I want to introduce our first speaker. Me.

Let’s talk about the Brave New World of SEO.

Like I said, so much to talk about here.

So let’s just get into it.

Alright. So search is at an inflection point. What do I mean by that? Well let’s look at this. So if you recall, search started out very much before Google it was really bad. It was all about, like how do I shove in these keywords as many times as possible and whoever used the keyword the most was the one that ranked. And then Google popped up, changed the game, they introduced PageRank.

And at that point we had our whole like ten blue links situation and search was really easy. It was like, cool. How do I find these links and get as many of those and then beat everyone from doing that? Right?

And then things have continued to evolve, of course. So we got to a point where Google has knowledge graphs. They can understand the people, places, and things on the web. And then after that, they moved into the transformer. I know I’m skipping things, but I’m giving you the most important things to what we’re talking about today. That’s how they can better understand the context of the words being used. And then we have EAT.

This order is very specific, because EAT is not about author profiles, and we’re going to talk about that. Now we’ve got AI overviews, which is of course changing how people interact with search. And finally, we’ve also got ChatGPT, which is now conversational search. So this is where we are.

And effectively, users have so many places that they can be going. It’s not just that they go to Google, they can go to TikTok, they can go to any other channel and use it as a search surface. I’m getting hot up here.

Three thousand dollar coat on the floor.

But let’s face it, your organic search traffic is not coming back. It’s something that we’re all kinda like talking about amongst ourselves, but people are fearful to be like, yo, the traffic is going down.

They think they’re doing something wrong. They don’t want people to know that their traffic is going down. But we’re seeing this across the board. Right?

What you’re seeing in this chart here, is the traffic for Wikipedia over time. And so what you would have expected is that when ChatGPT came out, that they would have lost traffic very quickly. That’s not true. The trend of traffic after ChatGPT actually went up.

Probably as a result of people feeling the need to verify information that they’re getting, because everyone knows things hallucinate. But what you notice here on that second line on that chart, is that when AI overviews came out, Wikipedia traffic started to trend downward. So it’s not just you, it’s happening across the web. So what we’re also seeing, this is data from SimilarWeb, is that, you know, the activity in Google has gone up.

More pages per visit in Google. Specifically, around the time the AI Overview is rolled out. So when you hear things like Sundar saying, like, hey, there’s more and more people searching on Google. It’s much better and so on.

That is actually true. Users are spending more time looking at more pages. And so we’re also seeing you know, this is like another pretty big, really well loved content site seeing the same thing. So again, it’s not just you.

It’s the whole web that’s seeing things start to go down. And it’s not coming back. So you need to change your mindset about what success is in search.

And I know that, Bianca is going to talk about this as well. But on this next slide here, we’re showing that this is the the chart that everyone’s seen around HubSpot and its performance. Right? So everyone’s like, oh, algorithm update, algorithm update. What they missed out on is the fact that there’s a pretty steep drop right there that happened when AI Overviews launched. So again, there’s a big picture thing happening as a result of what Google is changing.

So unless you’re monetizing on impressions, or ad impressions rather, was traffic ever the measure?

Like, why are we always measuring on that when businesses are trying to do specific things? So on this next slide, excuse me. I talk about where Google is going. Google has always been thinking about this from a perspective of what is best for the user.

And there was a paper that came out a couple years ago by a gentleman named Andrei Broder. You may not know who that is, but you know his work. If you’ve ever used the taxonomy of navigational, informational, transactional, that’s Andrei Broder’s work. In this paper, it talks about the Delphic cost of web search.

He talks about there being a cognitive cost involved in searching. And think about it. You’ve got to first take your information need, convert it into Orwellian Newspeak, put it in Google, review ten links, and then also look at all these other SERP features, which are all coming at your attention. And then you gotta make sense of all that.

Go through a bunch of results to get your answer. Google is like, nah. That’s not a good experience for users. How do we put that answer right in front of you?

So, of course, that’s what we’re getting with AI overviews. They are summarizing that information and putting it in front of you. And so, so many users are like, cool. I’m reading through this.

I don’t have to click through anything. But they perform subsequent searches. And then when they’re at that point where they’re like, cool. I’ve read everything I need to know.

I wanna make a transaction. I wanna, you know, get more information.

At that point, they’re clicking through. And they’re a more educated or more qualified visitor at that point. That’s why we’re seeing that the relationship between traffic and conversions is no longer linear. You’re seeing a lot more people that are converting, but a lot less people that are visiting the site.

AI mode is the natural next step of that. If you’ve heard of deep research, which is popping up across all the different conversational surfaces now, they’re bringing deep research effectively to the SERP. At that point, we’re talking about reasoning and all these other things that kind of mix up what’s going to be ultimately shown to the user. So you have far less control there. Google is also training users for longer queries. So if you there was a news article about this a couple days ago, where they showed that the search box has gotten longer. Because effectively, they can better understand your queries from a natural language perspective.

There was a demo about a year ago, where they showed something called the Bespoke UI. And in that video, this gentleman was like going through, trying to figure out how to start a or organize a party for his kid’s birthday. And the interface was directly related to what it is that that person was doing. It was not your typical ten blue links. So this is where Google may be ultimately going.

And then once you start talking about agents, and agentic AI, and so on, that’s what we’re talking about. The abstraction of the user is just completely different.

There’s something that’s going out there, trying to represent you and get that information without you even being in that loop.

So the funny thing to me is that SEO keeps trying to minimize this moment.

SEO keeps trying to say like: Oh, it’s just SEO, it’s just SEO. You know, it’s not a situation where things need to be different.

I don’t understand.

Like I don’t understand why you guys want to keep being this.

And when I say this, it’s effectively, like, the people on the Internet that don’t get any real credit, aren’t really respected, aren’t really getting the value for what it is that they do.

And so there’s a concept behind this. This idea is called schema preservation. It’s when new information comes in, and you’re like, no, no, no, no. It’s not new.

It’s the same as this. And people do this out of fear. People do this because they don’t understand the situation that they’re in. And they just wanna, like, basically, mitigate everything by saying, like, oh, yeah.

It’s just like the thing that I know. Right? Just like when the leaked documents came out and everyone was like, oh, already knew all that. No you didn’t. Get the fuck out of here.

So right now, about eighty five percent of the results in AI Overviews are coming directly from the SERP. Right now. That’s the important idea to remember here.

That’s not necessarily always going to be true. Right? So on this next slide, I got this screenshot from Reddit where the CEO of Perplexity was like, oh, yeah. I don’t know what you’re going to do about SEO once all this reasoning and stuff happens.

This technology is evolving so quickly.

Just because AI Overviews overlap heavily with the SERP right now, and you can still manipulate the SERP in very similar ways that you’ve done in the past, does not mean it’s gonna remain this way going forward. This is a platform shift. They have the agent to agent protocol. They have the model context protocol, where effectively you’re going to be able to plug this stuff into any system. You’re going to be pulling data from all these different places.

SEO, as it stands, is not ready for this.

There’s another aspect here where there’s gonna be a lot of hyper personalization as these memory functions come out. So with ChatGPT and also with Gemini now, they’re able to remember your conversations over time. When that comes to search, that’s a level of personalization where the results are gonna effectively be one to one based on that user context. When we talked about personalization in the past. My kids are here. This is cool. Hey Zora.

Yeah. That’s cool. So, yeah. With memory features, things are gonna be one to one. It’s gonna be very difficult for you to slot yourself in there, and this is exactly why we’re not ready, because we don’t understand these technologies.

We don’t understand the underpinnings of all this stuff. And that’s why this day is very much about these underpinnings.

Let’s talk about the limitations of the SEO space. So what is SEO?

It’s a whole bunch of shit that we put in a box and started claiming ownership over. SEO actually should not exist. SEO is a function of the fact that all these other capabilities and disciplines don’t think about requirements for Google. So suddenly, we start doing everything.

When they say, like, hey. Core Web Vitals, we are now performance engineers.

Did we, did anybody get any more money for becoming a performance engineer?

I just wanna, you know, be clear on that. So, I feel like SEO is in its third generation.

Like 97 to 07, and I call that like Boomer SEO. That’s like that’s like Danny Sullivan, Bruce Clay. You know what I mean? Where they’re the ones who we they set the foundation. They built what we’re standing on. But one of the main things that they did wrong, was call this channel free traffic.

And then we had Gen X SEO, which is like Rand, and Wil Reynolds, and so on. They tried to level things up dramatically. They tried to bring a lot more scientific rigor into this space. And then we’ve got millennial SEO, which is like Aleyda, and Lily, me.

And we are actively changing things. Right? But there’s a lot of push back from the old heads.

What’s next is probably gonna be Gen Z SEO. You know how Gen Z is. They’re not gonna allow anything that doesn’t make sense.

So I’m trying to do my best to leave something good for them to take over.

Here’s some challenges in SEO. We don’t think big picture at all. The things that we can be doing with the information and data that we have could change businesses. Instead, we’re just talking about content links and three zero one redirects.

I know a guy who used to work at Deutsche, or one of those agencies. He worked on HTC, and he used search data, search volume data, to prove to them that HTC should have phones in colors that are not just black, white, and gray.

That is big picture thinking for SEO. We don’t do enough of that. Our software is obsolete, frankly.

Most of what SEO software does is on a model that Google kinda still uses. And what I mean by that, there’s effectively two different models of search. There’s lexical, where you’re just basically counting the presence and distribution rates of words. And then there’s semantic, where you’re effectively understanding meaning.

Almost all SEO software is still on that on that lexical model, which doesn’t make sense. So a lot of the information that you’re acting on is not right.

There’s a ton of misinformation in our space too. I mean, I mentioned EAT earlier.

It’s not about author profiles. But for some reason, everyone in this room thinks it’s about author profiles.

We’re gonna talk about that. And then, also, the value is just very much misunderstood in our space. We are the biggest referral channel on the Internet.

We get paid the least. Make it make sense.

So this myth of free traffic is something that we need to let go of. You know, there’s another channel out there, or series of channels out there, that are governed by algorithms, and you need content to play in them, and the traffic is free.

Social media. Nobody is selling them to do projections. Nobody is questioning whether or not it’s gonna work. They’re just like, hey, it’s social. Let’s just put some things out. Spend a hundred thousand dollars on it, and see what happens.

But in our space, we gotta jump through all these holes because it’s free traffic.

Search has also always been a brand channel. And it’s a disservice that we consider ourselves a performance channel. Brand has a much bigger budget than performance. Brand is able to take more risks and bigger swings.

But we are always like, oh, well, it’s zero click. We’re losing our traffic. What are we gonna do? No.

People seeing you in a featured snippet or an AI Overview and taking an action is valuable.

We need to take credit for that. You need to take credit for impressions. That is valuable.

SEO also has no standards. That’s why I can talk to every single person in this room and ask them what they do. Like, what do you do in your day to day for SEO? You will all give me a different answer.

Because there are no standards. There’s no common set of education that we all have. We all kind of do different subsets of the same thing.

And SEO has actually made the Internet worse. This was proven by a series of researchers, academics, out in Germany. They did a longitudinal study over a year, and they proved that SEO is making the web worse. So our reputation is awful.

Google’s been doing hybrid search for a long time. I mentioned that idea of lexical and semantic. They do both. Semantic, probably, the dial is up a bit more, because there’s a lot more value in there from an AI perspective. But again, the way that we operate in SEO is just on that lexical model. So again, most of our SEO software is obsolete.

There’s also this myth that generative AI content, just because it’s generative AI content, gets you penalized. That is not true. You cannot use that as a single signal in isolation to indicate whether or not content is good. Instead, what Google does, they use all the signals that they typically use. Things like user interactions, and so on. And this is also further proving that they can’t do it, because now they are asking quality raters to tell them, which means they’re trying to train new classifiers around it. So generative AI content is fine, it just has to be good.

EAT was never about author bios. EAT is how Google it’s an abstraction. It’s Google saying, hey, we do all these things with these goals, and we’re actually able to do these on the back of vector embeddings. So they’re able to generate vector embeddings for entities, vector embeddings for authors, vector embeddings for websites to understand the relationships with that topology of the web.

So do you all really want to stay the janitors of the web?

Because AI, like, this is our moment. This is our moment to really stand up and be something different.

If you wanna just keep being the janitor of the web, keep saying, like, oh, it’s just SEO. It’s just SEO. Even if it is just SEO, the framing is different. We can redefine what this is.

So remember that time when Google used us to make the Web faster? They’re like, hey guys, you’re going to get a rankings boost, so just go out and make the websites faster.

It actually happened. We saved Google billions of dollars from crawling the Web because we ran out to do their dirty work for them.

Really, being an SEO is being a power user of Google. I’m not an SEO. I do something different.

And you probably do too. You just aren’t calling it something different. So what’s the most important thing that I want you guys to understand?

It is the idea of vector embeddings. Because this is what has transformed search engines’ ability to do what they do.

And effectively, what you’re doing is you’re taking words, and then you’re converting them to coordinates and multidimensional space. And then you could do a variety of different linear algebra operations to understand the relationships between things.

So, really, those operations look like this. You’re doing what’s called cosine similarity. There’s a variety of ways to do it, but cosine similarity dot product. These are the most, what are the words, popular ways to do it. Right?

Go back.

So with, if you have a cosine similarity that’s close to one, that means it’s highly related. If you have a cosine similarity that’s close to zero, that means it’s orthogonal or not related. And if you have a cosine similarity that is, close to negative one, that means it’s opposite.

So these are the operations where we’re determining relevance. It’s really the measurement between angles in these vectors.

So when we talk about relevance, it’s not this qualitative idea, where it’s like: hey, I read this, I feel like it’s more relevant.

It is a mathematical idea. So let’s take you into the algorithm. Let’s go to vector space.

So this is effectively what Google is doing. They’re taking your queries, they’re plotting them in space. In this case, we’ve got travel points, we’ve got credit card rewards. These are close together so because they’re relevant.

Cashback bonus, these are all relevant ideas to each other. So they’re physically closer in space. And then you’ve got airline miles, another one. The concept very similar, so it’s plotted very closely.

Right? Then if we go a lot further, a lot further. Okay. Right there you got credit cards.

That makes sense. But if we go further, keep going, keep going, dog leash. That’s not related. You see how long it took us to get there?

That’s because it’s not relevant to those other subjects. So, when you put something on your website that’s not related to everything else, you have this thing that’s floating in space, that doesn’t make sense.

And they’re like why is this here? They can see this easily from the topology of the web. So ultimately what they’re doing is they’re clustering all this across the web.

They can understand when you’re doing something wrong. This, I’m just showing you this with keywords. They do this with entities. They do this with authors.

They do this with websites. They’re able to see, like, the similarity is not very close on those, but it’s very close on those things down there. I don’t know why we put that back there. Sorry about that.

But the whole point is that when these, when you zoom out you can easily see like, okay. All this stuff is related. These things are orbiting the same space. So Google is very capable of understanding where your topics live.

Understanding that an author writes on these subjects, and they’re an expert because they are very well clustered together.

But when you start doing random things, or you start building links from pages that don’t make sense, it is trivial for them to understand that.

And in SEO, we trick ourselves into believing like, oh yeah, I can get links from anywhere. It doesn’t matter. They don’t have to be relevant. No. That’s why your link building doesn’t work.

Because you’re not doing what they’re looking for. In this case, it’s really easy for you to say, okay. These topics go together. They’re close. So it’s very easy to understand what those relationships are. So this is what we’re doing. We are effectively engineering relevance, if you’re doing SEO the right way.

Lots of use cases for these vector embeddings. So one, content engineering, making content that fits into the expectations of Google, calculating your relevance, so you can use those same scores that Google is doing, building better internal linking, clustering your content better. There’s a lot that we can be doing with this.

What SEO tool does any of this stuff? Anybody?

Exactly. Because our tools are obsolete. 

So this is engineering. This is not optimization.

Optimization is like being a car mechanic. You’re like, oh, yeah. I’m just gonna fix this radio. I’m gonna make your car go a little faster.

We are building cars here. We are not fixing cars.

What I do is called Relevance Engineering.

You don’t have to call your stuff that. I’m gonna call my stuff that, and I’m actually a lot cooler than you.

I’m kidding. I’m kidding. So Relevance Engineering is really the confluence of artificial intelligence, content strategy, information retrieval, not SEO.

SEO is an abstract.

Information retrieval is the science by how you get things ranked. It’s also user experience and digital PR.

So, yeah. It sounds like SEO, but it’s a broader scope. Because we are looking for something that is far more in-depth here.

And it encompasses all versions of search. Whatever search surface you’re trying to optimize for, this is relevance engineering. So whether that’s, you know, conversational search, Amazon search, Google search.

And the main difference is that SEO is a checklist culture. Think about your audits. You have some sort of checklist that you go through every time. You’re missing a lot of stuff because you’re just like, oh, yeah. The best practices say this. How many times have you said to someone, like, oh, your page title is supposed to be sixty to seventy characters.

That’s wrong. That’s a best practice that was established at one point, and was just never reconsidered.

Every time we have appended the top three keywords from Google Search Console for a given page to the end of the page title, click through rates go up or not click through rates. Clicks go up twenty percent.

That’s not a best practice. If you looked at that, you’d be like, oh, Apple rank doesn’t know what they’re doing. They’re not following best practices. Because they don’t work.

And we need to be continually testing that, not just following checklists.

So how does it differ from SEO? One, it defines itself. SEO has largely been defined by Google. Google tells us what’s right, what’s wrong, what we can do, what we can’t. That doesn’t make any sense.

We need to define that. We need to say what the best practices are. We need to say what we cover. It also requires, or features, technical rigor. Scientific and technical rigor. Because there’s a lot of people in our space that’ll get up on the stage and be like, oh, this is what it is, anecdotally, because you saw it on one website.

We need to level up and hold ourselves to a higher standard. And it also operates self sufficiently. You don’t need to go out to other teams to be effective. You should be able to do these things yourself.

You shouldn’t be at the whim of, like, oh, well, the engineering team only has five tickets this week. No. You should be able to have someone on your team that can step in and be like, here’s how we’re gonna fix this. That’s Relevance Engineering. So where it really starts is like, okay, how do we identify relevant audience?

And if you’ve been following me for a long time, what I’m talking about is personas. Who are the people that we’re looking to talk to? How do we reach them? What keywords are they looking for?

Or if it’s not keywords, what conversations are they having? And then which channels do we want to actually be in? Is it ChatGPT? Is it Google? Where is it?

Then we need to configure the platforms. Are things actually set up to be accessible by these systems?

And then you want to create relevant experiences. Now we’re talking about content strategy. We’re not talking about content marketing, because you can just create a piece of content, throw it up, and you’ve done content marketing. But I’m talking about building content experiences, building systems, so that we can be effective in doing this stuff over and over. And then it’s about developing co-relvant experiences.

So you could say link building. You could say digital PR. But I’m saying, I want to be mentioned, and I want to be linked from sites that actually matter to what it is that I’m talking about, that are very relevant. And then ultimately, you’re measuring performance.

So what are the tenants here? One, technical standards. We are going to establish a series of technical standards that we all adhere to. We’re gonna scrutinize our best practices.

We’re not gonna just keep saying, like, oh, yeah. Sixty, seventy characters for page titles. We are gonna collectively test this over and over, and then share that information to say, like, hey. Here’s the up to the minute situation.

And then building experimentation frameworks. A/B testing. A lot of people think this is optional in our space. You absolutely need to be doing this no matter the size of your site. I don’t care if you don’t have enough pages that fit the same, similar, you know, template or whatever. You need to be A/B testing as much as you can. And reproducibility.

Reproducibility, rather.

You know, and peer review. We’re gonna have people that are gonna be, and when I say we, I don’t mean iPullRank. I mean the community that are gonna be peer reviewing things. So that we can have things that we can say this is the gold standard. This is real. This has been looked at. We’ve looked at the data, and we’ve reproduced these results.

And open source software. We can’t rely on the SEO software industrial complex.

They are not…well, the people in the room are actually pretty good. But they are not incentivized to do this the right way. They are incentivized to give you some sort of simple score that may or may not represent reality, and then you just optimize towards that score, and that score does nothing for you. So we’re gonna build open source software that we can all use.

And we can get the data that we need, because there needs to be data democracy. We can’t be like, oh, well, there’s one place where you can get rankings from ten years ago. That’s ridiculous.

That needs to be available to everyone in this space. Everything that we’re going to build is going to be feature complete, meaning that you need to come with the skill set to do these things. You can’t just be like, well, I’m a content SEO. Nah.

You need to know how to do this. And finally, the whole thing needs to be community driven. It can’t just be me telling you, like, here’s what it is. There are so many smart, capable people in this space.

A lot of them are gonna be speaking here. And there’s so many good ideas that we need to collect and pressure test and define what it is that we do.

So we’re gonna be launching Relevance Engineering.org with all these frameworks. Again, community driven, open source, lots of tools that you can all use.

And here’s some of the current projects. So one I’d started a while ago, the gateway specification. So we can have data portability across all these different tools. It’s so ridiculous that I’m locked into whatever enterprise crawling tool there is because they have a very specific data model that I can’t take to someone else. And so we need to have a standard that they all need to adhere to.

Another thing that we’re working on, because I think it’s preposterous that we have these link indices with these metrics that are effectively entertainment metrics, is and then also not having a vector database of the web. So we’re building a distributed peer to peer system so that we can vectorize the web, and then everyone can have that data. We also have the Search Telemetry project, which is an output of all the things that we learn from all these leaks. So that we can get as many of the data points that Google is actually using to then have in our data. So we can be like, okay. It looks like it might be this. Let’s do correlations.

And then finally, the open search initiative.

I want everyone to have access to all the rankings data, all the link data that’s out there. None of these tools have something that isn’t public. So we should make that available to everyone.

So I would love for you all to get involved, because we gotta do this together. This can’t just be me telling you what to do. Alright. Relevance Engineering for AI Overviews.

So I’m so frustrated looking at all the content about AI Overviews, because it’s so wrong. And I literally told you guys what to do two years ago.

And so I’m just going to tell you everything I know. And then you can do with it what you will. Here’s my strategic approach to AI Overviews. 

One, you’ve got to assess the impact. Are you actually being impacted? So basically, what keywords were driving traffic for you that now have AI Overviews? And then what’s the delta on the performance of those keywords? And then you need to look at this two ways.

There’s the performance aspect. Again, continuing to drive traffic, if that’s what you want to do. Okay. Well then we need to think about going further down the funnel on the keywords, and focusing on those.

But if you want to engineer for branding, well then we want to get you into the AI Overviews. And then just measure your improvements, and continue to engineer from there. So as we know, about twenty percent of all queries based on data from Zip Tie are showing AI Overviews. It differs across different spaces.

Health space is like sixty six percent, so on and so forth. And all of this is governed by a concept called Retrievalog Minute Generation. This is basically when you combine a search engine with a large language model.

And the way this works is they are decomposing your pages in an additional way to what we’ve historically known in the SEO space. Right? Like we know, Okay. They tokenize them and put them into the index. Now, they are generating vector embeddings, not just on a page level, but on a chunk level, so passages. And they’re storing those, and then a search happens, and those passages are then fed to the large language model to generate the response.

So there’s many ways to do this chunking, breaking things down into passages. For Google, they’re doing this on a semantic basis, meaning that they are understanding the logical boundaries within your copy. And they’re able to say: cool. This is a logical boundary, index that.

This is a logical boundary, index that. And so, the way all this works effectively, they’re like: okay. Query comes in. Do we want to go to the training data, and then give a response based on that?

Training data isn’t good enough. Cool. Go out and give me a bunch of documents, and then let’s feed those to the large language model.

There’s a patent that they published about this that tells you very specifically what they do.

And they have their prompts in there. They’re very simple. One of which is, ‘summarize content A, B, C, and D’. That’s it.

Another one is, ‘in context of this query, summarize this content’. Another one is, assuming the user is familiar with this concept, answer this query. Very simple things that you can plug into whatever you’re using to see what might be generated. And they also talk about using not just the query, the documents from the query that you put in, they look at related queries and implied queries.

So that’s a lot of times why you’re seeing like, oh, this doesn’t even rank for this query, but it’s in the citation.

And so when you think about this, they have this concept that they call query fan-out, where they’re basically going out to figure out, like, what are these other queries? So you also need to understand what those queries might be and see, hey, how do I rank there too? So it’s not just about your core query.

As far as structured data, and I know there’s been a lot of wrong conversation about this in the SEO space. Structured data does come into play here. It’s not that it’s being trained on the structured data, but the structured data can be ingested during the rag pipeline. And Crystal Carter, who’s going to be speaking on DeepSeek later, or on Thursday, proved this, which was great, because I didn’t want to argue with y’all.

So here’s an example of what this looks like. Obviously, the AI Overview has evolved a lot since the screenshot. But you know, you’ve got your citations, and when you click on your citation, it takes you to the passages that Google used for that AI Overview.

These can be extracted. This is a scroll to text fragment, also known as fraggles, as Cindy has named them. And you can get these. But again, our SEO software is not giving you these.

So what I’ve done is I’ve just grabbed all the, the citations here, all the copy, all the fraggles. And then I did some, you know, basically, some cosine similarity analysis and noted that, effectively, the higher the relevance, the higher you appear in these AI Overviews. There’s a whole paper about this. And this is why people are calling it Generative Engine Optimization, I would imagine.

And it’s probably my fault because I was one of the first to show this paper. But what they did here is they tried a whole bunch of different things in Perplexity to see what was gonna actually make a difference for ranking. And so they tried all the things. They did, like, you know, the keyword stuffing, all your typical SEO things.

But what they found is that, you know, citing sources, being more authoritative in how you speak, and also having statistics are the things that you that get you in there most. So let’s get a lot more specific about how you actually do that.

There’s this concept of what are called semantic triples. This is how content is basically understood, and data is extracted from, a given page. So in the same way that your your, structured data, you’re saying, this thing is this, you can do that within text. And so you have the relationships of the subject, predicate, object in your content, and then you can pull those out. So what you want to do is one, structure your content in clear semantic units. Meaning, you want to have a heading that’s clear, and then you want to have a couple sentences that communicate that idea very quick, very clearly. You don’t want to say something like, oh, something like, you know, a big block of text, when instead you can just be like, here is the answer right here.

And you also want to be very precise in how you’re using this language. So rather than saying SEO helps businesses get more traffic, you want to say SEO increases organic traffic for businesses. That is easier for search engines to understand and extract. And then you also want to have highly specific or exclusive insights. So rather than saying, you know, SEO helps businesses get more traffic, you want to say, our analysis of one million search queries show the pages blah blah blah, so on and so forth. So having those stats in there in a very clear way allows them to extract that content easily.

And avoid ambiguity. Don’t say something like, they improve performance. Say, our SEO strategies improve the average organic search rankings by fifteen percent within six months. Which is just generally better for you to be doing anyway. Right?

So what we’ve built is this AI Overview simulator. Using the same technology that Google does, using their prompts and so on, you can basically plug in your page or your copy and the query, and it’ll give you recommendations on how to improve it, so you can get into the AI Overviews.

That’s free. That’s yours. How it works? It uses a framework called LAMA index for the rag pipeline. It uses Gemini 2.5, so that we can replicate how Google might see it. And also vertex embeddings, which are Google’s own embedding technology. And finally, we also use the SERP API, so that you can have it compare against what else ranks for your queries as well.

That’s not SEO. That’s Relevance Engineering.

Closing statements.

It’s not just SEO. Do not minimize this moment. This is our opportunity to be better, to be more effective, frankly, to make more money.

And the future of search requires us to be intentional in order to run the show, not just be the show.

That’s all I got.

What are you waiting for? Get SEO Week Tickets. Now.

As AI rewrites the rules,

read between the lines.

AI is reshaping search. The Rank Report gives you signal through the noise, so your brand doesn’t just keep up, it leads.