Welcome to another edition of The Search Session with me, Gianluca Fiorelli. Joining me again is Garrett Sussman, a Virginia-based strategist with experience in SEO, content marketing, and AI search.
Garrett returns to discuss how SEO is evolving as AI platforms introduce new paths to brand discovery. He shares findings from his Google Personal Intelligence experiment, including the effects of emails and private context on AI Mode recommendations and query fan-outs.
We also explore brand sentiment, relevance engineering, AI-generated content, and the challenges of preparing for personalized agentic search.
What you’ll learn in this session:
Why old SEO tactics are no longer enough: AI platforms use training data, shifting indexes, and query fan-outs, making discovery far less predictable.
Why brand sentiment is harder to control: LLMs assess your entire digital footprint and can resurface negative content buried in search results.
What Garrett’s personalization experiment uncovered: emails drove a 40% visibility lift, while fake brands, unopened newsletters, and private context influenced AI Mode.
How personalization could shape users’ choice of AI ecosystem: it may tie consumers to a single ecosystem that becomes increasingly difficult to leave, despite privacy concerns.
How to build a GEO strategy that survives change: strengthen your product and show up wherever your audience spends time, instead of chasing citation shifts.
Why AI slop is a human problem: better tools raise the baseline, but useful content still depends on creativity, judgment, and audience understanding.
How relevance engineering prepares brands for agentic search: it connects five disciplines to build visibility across the entire digital footprint.
Where personalized agentic search can fail: incomplete or outdated personal data can produce irrelevant recommendations.
Tune in for Garrett’s thoughtful perspective on what has changed since our last conversation and what these shifts mean for marketers navigating AI search.
Topics covered: SEO industry · brand sentiment · AI search personalization · Google Personal Intelligence · personalized query fan-outs · GEO strategy · AI slop · relevance engineering · agentic search
About the Guest
Garrett Sussman
Marketing Director at iPullRank
Garrett has more than 15 years of experience across SEO, content marketing, demand generation, and AI search. His recent research explores how personal data and email content can influence recommendations in Google AI Mode.
Garrett joined iPullRank in June 2021 and became Director of Marketing in October 2024. The agency helps enterprise brands improve visibility through AI search, relevance engineering, technical SEO, and content strategy.
He is a MozCon and SEO Week speaker and hosts Rankable, where he discusses SEO topics with digital marketing experts, and The SEO Weekly, where he covers recent developments in SEO, content marketing, and generative AI.
Video Chapters
Transcript
Full conversation between Gianluca Fiorelli and Garrett Sussman.
Gianluca Fiorelli: Hi, I'm Gianluca Fiorelli, and welcome back to The Search Session. Today we are going to have, for the first time as a guest, someone who already participated in The Search Session, actually in the second episode of The Search Session. Quite a long time has passed already. If you are one of those who, thankfully, have been following The Search Session from the beginning, you should maybe remember who I'm talking about.
He is the Director of Marketing at IPullRank and defines himself on LinkedIn as an SEO, content marketing, and AI search leader. And it's true. Many times, we say whatever we want to promote ourselves on LinkedIn, but I can confirm that this definition is correct for this person.
He spoke at SEO Week with a great speech, and we are going to take some things from that speech and the experiment behind it. This person is Garrett Sussman. Hey, Garrett. How are you doing? Welcome back.
Garrett Sussman: Gianluca, thank you for having me. It's such a privilege to be back. I remember our last conversation. You and I go way back, and, man, we live in a weird industry, so I can't wait to dive into all that we're going to talk about today.
Gianluca Fiorelli: Oh, yes, me too. Everything is fine in New York City, on the other side of the pond?
Garrett Sussman: Yes. I mean, the weather's great. The Earth hasn't burned yet, although it's a little bit hotter summer by summer. So, the agency's based in New York. I ended up down in Virginia. I'm actually surrounded by data centers. There's one down the block being built right now, so I know that's a big topic of conversation in our world, all these data centers, especially in the US.
So it's like I'm in it. I'm literally in data center alley. Hopefully, we're safe here because all your internet is here in Virginia.
Gianluca Fiorelli: I mean, that could also be considered a military subject. Okay. No Russians or Iranian people are listening to us with bad intentions.
Quite a few months have passed since your participation in The Search Session. And let me say hi to Giulia Panozzo, who was with you in that episode.
Explore Garrett’s earlier perspective on search behavior
In his first appearance, Garrett joins Giulia Panozzo to explore how cognitive biases, authority, trust, and brand familiarity influence search behavior across the messy middle. The conversation shows how personal perspectives were shaping search journeys even before today’s deeper AI personalization.
Complexity vs. Complications in Modern Search
Gianluca Fiorelli: During this period, many things happened. I feel as if something that was exploding has exploded totally, and with a velocity that is incredible. So, the classic question is, how have SEO, AI search, GEO, relevance engineering, and whatever you want to call it treated you since that first participation in The Search Session?
Garrett Sussman: First off, I feel like the industry is starting to catch up to things that you and I have been discussing. I mean, you've been discussing this for decades because we know machine learning is not new. Consumers are becoming more and more familiar with all this.
And so I think, as an industry, we are maturing in the way that we interpret it, research it, strategize around it, and things are different. The fact that so many more people are using AI search platforms, AI Mode, ChatGPT, and still Perplexity as the starting point for their search and their discovery of businesses, brands, and information has increased significantly.
It hasn't replaced SEO, but I think more brands, marketing leaders, and SEOs understand that you can't pretend it's just the same song and dance, the same strategy, and the same tactics that it's always been. That makes our conversations and the way we're thinking about it so much more fun. A lot has changed.
Gianluca Fiorelli: Yes, I totally agree. I think a lot has changed. I think there are some nuances to consider. A part of the industry is still very attached to the old concept of SEO. Maybe they can be compared to the Luddites somehow.
Then there is a part of the industry, and I can also recognize myself a little bit in it, that is saying, “Okay, but I was already proposing many of the things that are proposed now.” Maybe it's more than saying, like Google, “Good SEO is good GEO.” But then, obviously, I'm saying this because maybe I was thinking of SEO as not just doing these specific things in technical SEO and these same specific things in content. It was something living in an ecosystem.
That's why I usually say that these are now new surfaces. Like other parts of SEO, such as local business, these surfaces have their own specificities that you have to control and build a strategy around, consistent with all the other marketing strategies or even the marketing strategy as a whole.
Then there are those who make another mistake, in my opinion, and say that SEO doesn't exist anymore. It's only GEO, et cetera. I think IPullRank, in this sense, is sometimes misunderstood. Many times, there is a semantic discussion because, for instance, I think people misunderstand Mike King when he says that SEO is deprecated. I think the old way of doing SEO is deprecated, not SEO.
Garrett Sussman: Yes and no, in the sense that I think we can all agree there is overlap. We agree that there's overlap between the way you would show up from an indexing perspective and a ranking perspective. That still exists in SEO, and it still plays a role in why a brand may be surfaced, but not in every situation.
We know there are situations, for instance, when OpenAI's ChatGPT is just referencing the training data and isn't referencing an index, or the index is changing. Sometimes it's going to Google, and sometimes it's going to Bing. More and more, we're seeing query fan-outs, where we're seeing related searches that go directly to websites. There's still a certain amount of black-box opacity where we don't necessarily know if OpenAI is building its own index as well.
We have a lot of really smart technical people analyzing different logs to understand what information we can pull out. I think about Metehan Yeşilyurt from Peec AI and all these different AI visibility tools, whether it's Advanced Web Ranking, Profound, or you name it.
Everyone's doing their research right now, but it's not the old, traditional SEO way of getting crawled, getting indexed, getting discovered, and getting recommended. There's so much more complexity to that.
Gianluca Fiorelli: This is totally true. Maybe SEO was complicated before. Now, with the introduction of AI search, the work of being visible in search—and search can be everything—is becoming not only complicated but even more complex. The two words have different meanings.
Garrett Sussman: Yes.
Gianluca Fiorelli: It was complicated because there were more bricks to examine and understand how they fit. Now it's more complex because you need to have at least a very good grasp of how this machine works. And I'm not just talking about AI search visibility in LLMs. The distinction you made between being pulled from training data and being pulled from query fan-out, RAG, GraphRAG, et cetera, is a huge difference that many people need to understand.
Then, and I think this is going to be the second very fast step, especially next year, we are going into the agentic web. How are agents going to discover whether your website's functionality allows them to interact with it on the user's behalf across any web surface? It can be a browser like this, or it can be a device like Google Home, for instance.
Garrett Sussman: There's so much.
Brand Sentiment Becomes Part of GEO
Gianluca Fiorelli: That's the big difference. Before, we were only talking substantially about Google and sometimes Bing. If you were working internationally, there were Baidu, Naver, and some others. Now, as in the very beginning of SEO, we are talking about many surfaces.
If we also want to talk about link building, brand amplification is going to become even more important because it's really entering the branding field. Sometimes I wonder whether people like us are going to become a piece in the machinery of a branding department, advising them on where and what to do, while the actual work is done by other people, not us.
Garrett Sussman: That's a really good point about the brand because that's a whole other layer where we need to reframe our responsibilities. Not only do we want to be visible and recommended, but the sentiment around our brand is so much more influenced by our entire digital footprint. It's really not just about your website anymore.
Because you're getting this generative output, you're not just getting the 10 blue links of the SERPs, where people can go to each individual link and decide. You're getting this amalgamation, this synthesis of all these different attributes and sentiments about your brand that exist on the internet. Part of your responsibility as a GEO is to maintain the brand sentiment that you want to project and the narrative that you want to control, which is that much more difficult.
Obviously, in SEO, we think about the way we represent ourselves on our websites, but I think there's a responsibility for sentiment that touches every part of the integrated marketing department. Social, email, partnerships, relationships, and all of that influence the value.
I know you've talked a lot about the whole idea of the messy middle and the loop of discovering the way that people search. The value of showing up in AI search as part of that journey—of an entry point, a discovery point, and a decision point—is something that SEOs and GEOs need to discuss internally with their marketing teams.
It's an opportunity for our industry, but that's a whole layer that needs to be part of the conversation for the value of the channel.
Gianluca Fiorelli: Yes. It's what was once called forensic SEO or online reputation management. But online reputation management was often just something like hiding the bad stuff after page two in the SERP. When you need to think fast, you just push something under the carpet. It's not visible, but it's still there.
Now you know that it is not only still there, but there is an LLM that is very good at discovery and can make it the first thing people see. Something like, “Oh, see, this brand is also this.”
Garrett Sussman: Yes.
Gianluca Fiorelli: So we are talking a lot about assuming a governance role in this sense.
Garrett Sussman: Oh, yes. There's a PR side to that. And it's not just about avoiding the negative. That's one component of brand sentiment. But one thing I'm paying attention to is whether you are a new brand or have been in your industry for a while; if you're trying to pivot or change what your products or services are or who your audience is, that is so much harder now in the world of AI.
You can't just flip the script overnight and all of a sudden have AIs think, “Oh, well, they told us that they're now going in this direction.” No, you have to deal with waves upon waves, layers upon layers. The archaeology of your brand on the web doesn't go away.
So it's a completely different challenge to change the narrative, whether you're dealing with a PR crisis or brand repositioning. It's so fascinating.
Findings from Garrett’s Personalization Experiment
Gianluca Fiorelli: Yes. Another fascinating field is something we were already discussing the first time we talked together on The Search Session: how important personalization is going to be.
During this period, you really dug into this topic and started doing some very unusual experiments. I don't know how much your personal search experience is now biased by all these experiments, so much so that maybe you need a new user ID to access Google and the LLMs to get unbiased results and answers.
But let's talk about this experiment. I know that you decided, “Okay, let's see how the personal behavior of someone like me is going to be reflected in how the LLMs converse and interact with me and suggest things to me.”
Garrett Sussman: Yes. First off, I want to say that the whole concept of personalization is not new. We know this, right? You do a lot of international work. When you think about the local element, if I search for a restaurant near me, I don't expect to get the same results that you would get over in the EU. We know that. We're used to that.
But what happens when we start to extend that to every type of search and consider both the explicit context we provide in our queries and LLMs and the implicit context, the data, that we know an ecosystem like Google pulls?
We already know that they pull location. We already know they pull the device you're using. We already know that, to some extent, they look at your search history. What happens when you start to add more implicit data points?
The big deal around that is Google Personal Intelligence, which is something they rolled out more than eight months ago. It's an opt-in feature where you're giving Google permission to say, “Okay, use my Gmail history, use my YouTube search history, use my Google Photos, use my calendar, and take everything that you know about me that I'm not including in my query and incorporate it into Gemini or the AI Mode output.”
This really is a preview of the direction that all of GEO will go. It's hard not to imagine a future where everything is hyper-personalized.
So what happens when you do a very simple product discovery search, something like, “What is the best shoe for me? What is the best hoodie? What's the best bank?” based on all that private, personalized data?
To your point, I ran an experiment. I said, “Okay, I'm going to take three accounts. I'm going to take a blank Gmail account that's never been used before. I'm going to take a second Gmail account that's opted into Google Personal Intelligence.” Right now, you have to opt in to it. Don't worry, everyone who's watching this. You are not automatically opted in by default.
Then I had a third account, which was my own personal Gmail account with decades of emails and photos, my family—don't judge me, but I gave Google… Let's be real. Google has all of this information anyway. It's all in their ecosystem. You're just giving them permission to be really creepy about it.
Over the course of four weeks, I sent a series of prompts. We checked six types of prompts across eight different products every single day. We were asking, “What's the best shoe for me? What are your top three shoe recommendations? What is the best one that's sturdy and rugged?” We asked this across all three accounts to see what the output was.
Then, after about two weeks, we identified the top 10 types of results. But what would happen if I sent an email to each of those accounts recommending a brand that did not appear in the results? Would we see that brand start to appear? It was wild. With Google Personal Intelligence, you would see a 40% lift in brand visibility for those types of prompts across every different product.
Now, all of a sudden, you start to think about the implications for marketers. If people are receiving email recommendations from friends or seeing them in newsletters, how does that influence a personalized search result?
We saw significantly more impact from emails than photos. We did the same thing with photos. I uploaded photos, and we did see some impact, but you don't have the same sort of context in a photo as you do in an email. So it makes sense that emails would have much more of an impact.
In the second phase of the study, we actually sent fake brands. We made up brands and sent emails saying, “Oh, I recommend this made-up shoe.” Even those were showing up in AI Mode results. What's wild to me is that it did provide context. It said, “Hey, your friend Garrett Sussman recommended this brand,” but it used the exact wording from the email in the AI Mode recommendation.
There were so many anecdotal examples that made this experiment even weirder, but that's the long and short of it. Once you start to opt in to these ecosystems of implicit contextual data, you see a much more personalized result that makes our jobs as GEOs and marketers even more complex.
Gianluca Fiorelli: Yes, and when I saw it, because you and IPullRank shared your SEO Week talk publicly, the first thing I thought was, “Okay, these email marketing guys will never die.” Because if you are able to create a good newsletter or different good newsletters targeting the different types of customers you may have—which has always been the best way for email marketing—then, if these people are opting in, you are substantially setting the basis for Gemini, in this case, to opt in.
But then Claude asks you, “Do you want to connect your Gmail too?” So it could also happen with Claude. I'm not sure about ChatGPT, but I think it's possible too.
So, first thing, beware of what you are saying in emails because, if your wife or your partner is using ChatGPT, Gemini, or Claude with your user ID, maybe they are going to see something that you don't want to show.
Garrett Sussman: Yes. It's really interesting.
Gianluca Fiorelli: The other thing is, and maybe this is an idea for a second kind of experiment from you, that I'm really a believer in image search and visual search because I think these LLMs are naturally multimodal.
Maybe the fact that Google Photos with just the photo, without any real title, without the context, as you were saying, must use it as neutral information because the image itself doesn't give it real context.
It would have to do too much work to say, “Okay, this photo was automatically uploaded when this person was on vacation in this place, and for the vacation in this place, in the email…” There is this whole thread of things.
The only thing I miss, and what would be the real knockout for this kind of personal intelligence, is Google Plus. Imagine if Google had continued with Google Plus and had all the social conversations there.
Garrett Sussman: Yes, Google missed that opportunity. They were not successful. They tried, but they are pulling in social to some capacity.
To answer that question, and I'm curious to see how it plays out, what I am waiting for is this: in January, Google announced a major partnership with Apple, and we know that Apple is rolling out an upgraded version of Siri that is most likely going to be powered by Gemini.
For Apple users, I think that is going to be the experience because, all of a sudden, if you have a Siri LLM that's influenced by all your messages back and forth and potentially has access to all the other social platforms that you have, Google is embedded in the Apple ecosystem.
Apple has been famous, notorious even, for its protection of privacy and data, so I'm still kind of shocked that they made that partnership. Apple had the opportunity to roll out its own Apple Intelligence two years ago at this point, which was a complete disaster. Then we heard that OpenAI had a relationship with Apple, and that kind of went away. Now we hear about the Google one, and hopefully Google gets it right.
I had a very interesting conversation with Christian Ward of Yext, and he believes that there might be a future where, in Apple, you do not have to commit to one LLM ecosystem. You might be able to choose which frontier model connects with Siri, and that seems like a smart move. But that's open. That's not what Apple traditionally does.
Gianluca Fiorelli: Maybe, at least in Europe, they are going to be obliged. But this is also true for Android because of the anti-monopoly policies in Europe, which are stricter than in the US.
Garrett Sussman: Exactly. So it's going to be interesting from the perspective of the conversation you and I had with Julia way back when: the psychology of how everyone, in some way, chooses the ecosystem they're comfortable with and use day in and day out.
As people in the industry, we know that different LLMs are better for different use cases. I know I jump between Claude, ChatGPT, and Gemini depending on what I need to do. But I think the average consumer is going to have their chosen ecosystem tied into all their personalization, and then they are very much glued to that system going forward.
I think it will be even harder to pull away your own personal data in the future. We could see it going in a couple of different ways, but we're all creatures of habit, and we do not like massive change unless we're forced into it.
Gianluca Fiorelli: Yes, and we also have history that can inform us how people are probably going to react. Obviously, there is going to be a minority of people who are very aware of privacy concerns and able to go into all the settings and not give an AI system any possibility of looking at their private data.
But we already saw it with social media. We already saw it with Google Search itself. People don't really care because there is a human condition that is always the same throughout history: “I don't have anything to hide, so I don't care.” Then, normally, you do have something to hide.
Garrett Sussman: Depending on where you live, the government...
Gianluca Fiorelli: Many times these are the people using “one, two, three" as a password, so maybe there is something you should do.
Garrett Sussman: I know. It's like we need our government or our technology to protect us from ourselves. But, depending on where you live, you also don't know who has your best interests in mind. We have a certain amount of laziness in us. I think that's part of the human condition. I don't bemoan anybody for it, but that's a reality.
Why GEO Research Is Still So Volatile
Gianluca Fiorelli: But in your experiment, apart from seeing the fake brand being surfaced and Gemini or AI Mode recommending it with the exact words you added to the email, which is also a good example of how a certain kind of passage extraction can be done and could be used for that kind of test, what was the biggest aha moment you had?
Garrett Sussman: There were a bunch. I think some of the uncomfortable moments involved references to my family. The fact that my four-year-old daughter's name appeared made me really uncomfortable.
This is a conversation that everyone needs to pay attention to right now, independent of Google Personal Intelligence. We talk a lot about query fan-out. We talk a lot about where the sources for your AI Mode or AI Overviews output come from. We have to accept the fact that query fan-out is personalized too.
For example, I asked a generic question: “What is the best streaming network for me?” It knew that I had a family, so the query fan-out and citations referenced listicles about the best streaming networks for families, despite the fact that it wasn't a connected system. So the query fan-outs are personalized now, and we have to take that into consideration.
The other finding that surprised me was that, while most of the emails cited by Google Personal Intelligence in AI Mode were personal emails, I also saw a newsletter referenced. Marco Giordano has an SEO statistics newsletter. It's a great newsletter, and I highly recommend it.
One of the questions was about the best productivity apps, and it referenced an email newsletter from him that I had received six months earlier about n8n, AirOps, and a bunch of productivity apps in the middle of the email. What was wild to me was that, when I went back to find that email, it was unopened, Gianluca. This wasn't an email that I had read and looked into. It had sat unopened in my Gmail account. So you can start to think about the implications.
Now, I don't want to say anything definitive. We know in this industry that what stands today may be different tomorrow. Volatility is the new status quo. I would caution everyone that any research you see online about LLMs and GEO is probabilistic, and it's hard to base your entire content strategy on what you see today.
You need to have a broader, open perspective and mitigation plans. But the fact that it was pulling from an unopened email was shocking to me.
Gianluca Fiorelli: Yes, and I think you're right. In fact, I often do this when I write my blog posts. I always say, “According to this research conducted at this moment, this is what Peec AI was seeing, and this is what Profound was seeing,” et cetera.
This is also because you often see substantially the same kind of experiment produce different findings from one vendor to another. Maybe they change a few things, but the methodology and purpose are the same. One says that 86% of all answers are not pulled from the first ranking position, while another says the opposite.
It's true that you can read data in the way you want. This is also why I advise people that if an experiment was conducted by a vendor, they should take it as it is, but always with a grain of salt.
It's interesting. I agree that the situation is so volatile right now. When we suggest or recommend a strategy to a client, the best approach is to find the common denominator. At the end of the day, own your entity space and build the graph around you with the things you want, not only on your website but, even more importantly, outside it. This is what you call the entity footprint.
Then make sure the departments talk to one another. If you're saying one thing on your website while social media says the opposite on TikTok, or you're creating a very academic and serious video on YouTube while an Instagram post presents a completely different brand persona, your customers will not recognize you. The models will say, “Okay, I give up,” and recommend someone else.
Garrett Sussman: It's hard. Right now, and I know I'm guilty of it, you're guilty of it, and everyone in our industry is guilty of it, we shouldn't take victory laps.
For a long time, many people were bemoaning the idea that Reddit was the dominant source of information in many LLM outputs. On the one hand, people were saying, “We have this whole strategy to get into Reddit.” Other people were saying, “Don't worry about Reddit. Don't make that your be-all and end-all.” Then, recently, we suddenly saw a drop in Reddit citations.
Gianluca Fiorelli: Yes, with ChatGPT 5.6.
Garrett Sussman: Exactly. And I see people saying, “I told you so. You shouldn't have invested so much in Reddit.” But that's not the point. That could change back tomorrow. It could be a glitch. All this is changing so frequently that you need to expand your space. You need to expand your visibility and not put your entire strategy into one specific channel.
I think the biggest thing you can do right now, and this goes back to old-school marketing as well as good practices, is to ensure that your products and services are strong. They need to be good and helpful. You need to understand your audience. You need to understand where they are and where they spend their time, and you need to be there.
Whether the LLMs are pulling from those channels or not, that's just good marketing. The hope is that, over time, the LLMs will pull from the channels where your audience spends its time. Then, as things become more personalized, you're more likely to show up.
You're giving yourself the best chance to present yourself positively across all those channels. Regardless of what happens today, tomorrow, or in the future, you're there. You're giving yourself the chance to show up, and that's the best thing you can ask for in a difficult, complicated, and complex GEO strategy.
Is AI Slop Creating an Anti-AI Reaction?
Gianluca Fiorelli: Yes, totally. I don't know if you have this kind of sensation. Well, you are very enthusiastic about everything related to AI search, but talking just about AI, don't you have this feeling? Let's say that two or three years ago, during the days of Midjourney and experiments with AI-generated video or music, there was great enthusiasm. Everybody was playing with it.
Garrett Sussman: Yes.
Gianluca Fiorelli: People were also playing with ChatGPT and Gemini to create this kind of content. But something came up in a conversation with some Italian friends of mine: aren't you now seeing the other side of the ocean wave?
As this enthusiasm recedes, some sort of rejection of AI is emerging. Maybe people, not just marketers, have been so bombarded by AI-generated content, usually of poor quality, that they are training themselves to recognize what is generated by AI. It's not just a question of seeing six fingers on a hand. It's a deeper understanding of when something is generated by AI and a growing rejection of it, alongside a demand for genuinely human-generated content.
Garrett Sussman: I have two thoughts on that. The first is not seeing the forest for the trees. To your point about two or three years ago, the quality of the output has changed significantly in just a year or two.
There are tells, but in two or five years, who knows? I'm of the mindset that there is always the possibility that we could hit what is called an AI winter, a ceiling in its capabilities. We were talking about that a few years ago, when people thought it wouldn't get better. I don't think that's the case. I think it's going to continue to improve.
Whatever you consider AI slop now will not exist in a couple of years because it will continually become more difficult to distinguish between the two.
You probably remember in the 1990s, when Microsoft Publisher had clip art. Do you remember how every business used clip art in its marketing? At the time, you thought, “This is cool.” You had some cool options, especially for local businesses. Then, eventually, you thought, “That's so bad. There are so many better options out there.”
There are artists, marketers, and designers who can use these tools effectively and don’t give you the ick from AI-generated content, and you don't know right now. I would say there are different levels of how effectively people use AI. You might be turned off by people using it poorly, but that's not a reflection of the technology. As we always say, it's a reflection of the people using it.
Gianluca Fiorelli: Of how poorly it is used. In fact, as I said in a post and a talk, AI slop is not generated by the machine. It's generated by the human ordering something from the machine. Sometimes what we call AI slop is not inherently bad in terms of quality. For instance, you can read it perfectly well. In some cases, it is spectacular.
It's still evident, especially in video, that it is AI-generated. For instance, when people talk in AI videos, they move their mouths in an exaggerated way. But I don't think the slop is really about quality stuff and not slop stuff. For me, the slop is using AI, especially for textual content, to generate things that have no utility because they don't add anything new. If this is what it comes down to, then Google is correct, in my opinion, to distinguish between commodity and non-commodity content.
Garrett Sussman: Yes. But that goes back to creativity. It's not the technology. It's a tool, and people don't know how to use it.
One of the differentiators, and we talk about this a lot at iPullRank, is how we've evolved our positioning. We believe that technology can empower authentic human content if you're using it the right way.
It's about taste. It's knowing how and when to use it, as well as in which mediums, channels, and ways. I take umbrage with the idea that the issue is simply that it was created by a machine. I think it's a reflection of the businesses and people using it poorly, and that's a choice.
There are going to be people who aren't willing to spend the money or invest in the right people, and this has always been the case. We have this weird nostalgia that things were better with AI, people live longer, people are healthier, and people have more options and more entertainment.
There's always a way to spend less and get an inferior product, even if our definitions of quality have changed. I think Rand Fishkin talked about how the baseline for being a marketer is higher than it has ever been, but there's still a difference between people who know how to use these tools and people who just want to press a button without any effort or creativity and try to monetize it.
Even though we're lazy, people are relatively smart. They can see through the BS, or at least most audiences can. Businesses will learn when they can use it and when they shouldn't. But I think the conversation is much more sophisticated than “AI slop or not AI slop,” and you know that. You and I have these discussions all the time.
Gianluca Fiorelli: Yes. Unfortunately, the discussion sometimes remains on the surface and doesn't really reach the root of the problem.
Garrett Sussman: Yes, and it's hard too. When you think about marketing in the US, and I don't know if this is the case in the EU, we joke about direct mail.
This was another insight from my study. We have this service where you can have your physical mail scanned and delivered to your email. A recommendation for a shoe store showed up in my AI Mode, and that blew up my mind.
Gianluca Fiorelli: This could also be for the packaging of things too. For instance, if you are showrooming and saving the showrooming photo in Google Photos...
Garrett Sussman: That's why your interview with Myriam Jessier is relevant. She's an expert, so go watch that episode because she talks a lot about packaging and the impact it can have on appearing in AI.
But my point about direct mail was that its design in the US has always been really poor. The design is bad, but it works. People who don't care about high-quality design still buy from direct mail frequently.
That's the job of a marketer: knowing how to speak to your audience. If your audience is put off by AI slop, or whatever version of it, it's your job as a marketer to identify that and change your approach.
Gianluca Fiorelli: Yes, to find your Trojan horse.
Garrett Sussman: Yes, exactly.
The Five Pillars of Relevance Engineering
Gianluca Fiorelli: But returning to the topic of brand, we know that it is not our job to be brand marketers. It's our job to help brand marketers create strong, effective branding.
We have a specific field within branding, which is the technology behind it: how to translate the brand for machines. I think this is something that you at iPullRank are really studying and digging into with your concept of relevance engineering.
It's not just about how to do perfect passage ranking but about how to help machines understand that this is the brand, these are the elements related to the brand, such as its products, and these are the attributes that we want to promote around the brand in a technological way.
Garrett Sussman: Yes.
Gianluca Fiorelli: Can you tell us something about this kind of study that I'm sure you're doing at iPullRank?
Garrett Sussman: The whole concept of relevance engineering, and why SEOs transitioning into GEOs are best positioned to take on this role in an organization, is that the relevance engineer is the conductor of an orchestra.
We think about it in terms of five different pillars related to marketing and AI search visibility: AI, information retrieval, digital PR, content strategy, and user experience.
Those are the five pillars that you, as the GEO or relevance engineer, need to orchestrate. This includes brand visibility, content strategy, and digital PR across all the different properties you own, the properties that you’re earning, and even paid media.
Every instance of your digital footprint that we were talking about earlier plays a role in how you appear in AI search. Our solution frameworks bring all of those together. We look at everything from how you appear technically on your website to all these omnichannel properties, making sure that your brand is represented there.
As much as possible, that's going to future-proof you for the agentic web that you briefly mentioned. Imagine a very near future, also going back to personalization, in which every individual can send a bot or agent onto the web to touch all these different channels and bring back a hyper-personalized experience.
Is your website, is your brand set up for all of that? That's how we think about relevance engineering: putting a strategic practice in place so that a brand can prepare for that future version of search. It's complex and complicated.
Gianluca Fiorelli: Yes, it's complicated, but it's fascinating, and I think you're right. And combining agentic and Personal Intelligence, you know that I sometimes connect random things. We remember that Google already had a patent about creating landing pages by itself.
Garrett Sussman: Yes.
Gianluca Fiorelli: That's one thing. Second, we know that Google is already natively capable of making comparisons. We have comparisons within its own products. If we look at hotels or flights, they are essentially marketplaces comparing different types of products from different vendors.
Filter Bubbles & Algorithmic Editorial Choices
Gianluca Fiorelli: Coming to agents, agentic search, and Personal Intelligence, it's not difficult for me to foresee a zero-click experience fabricated by Google. Someone might ask, “I have to go to the Rockies and need to buy new hiking shoes. Which ones can you suggest for the period I've reserved in my calendar?”
With an agent, Google could pull all this product data and present a link to a Google-fabricated landing page where you can compare the options if it doesn't present them directly within search.
Garrett Sussman: There are two more challenges that we need to consider when it comes to this future.
Interestingly, Google published a PDF that I included in my blog post about Google Personal Intelligence, how it works, and its problems. The issue is that we jump between devices and don't give Google all our information. So there are gaps in what it knows and how it can update that knowledge. That's a very real problem when you don't have all the data.
For example, my daughter's interests change constantly because she's four. When she was two, she was really into Sesame Street. I would ask for present recommendations for her birthday, and I received great recommendations about Sesame Street.
Two years later, she's into the Australian cartoon Bluey. She loves Bluey. But before I let Google know about that, it was still giving me recommendations related to Sesame Street.
It's not a mind reader. It still needs information about how your life is changing and the different milestones you're experiencing. Whether it can identify that information and consider it in the output will always be a problem.
The other issue is a more philosophical one that you and I could have an entirely separate conversation about: is Google giving you exactly what you're looking for within your filter bubble a good thing? That's a philosophical question.
In my study, it often recommended brands that I had already purchased. Unless I explicitly said, “Show me something that I haven't had before,” it indexed what it already knew about me. And so that is a philosophical problem: how does it determine for you, or do you have to determine how much of the filter bubble it keeps you in versus expands?
Gianluca Fiorelli: I don't know how it can be resolved. Hypothetically, it could be relatively easy for Google itself to resolve by adjusting the temperature of the answer.
Let's say the temperature is 70% or 80% based on what already works and the tastes that are explicit or implicit within a person's experiences. Then leave 20% or 30% for opportunities to discover something new. I think that would also be useful for Google because it would expand its income opportunities.
I think this was previewed in a very old patent about search entities. It was already a problem in traditional search. Search is personalized, and search history is an existing personalization factor. But if personalization was pushed too far, you would always see results from websites you had already visited.
For instance, if I searched for SEO-related information, I would see iPullRank, Search Engine Land, Search Engine Watch, Search Engine Journal, Advanced Web Ranking, my own blog posts, Aleyda Solis, and so on. I would see all my friends but never discover anything new, which is also a problem with social media streaming.
Garrett Sussman: Right.
Gianluca Fiorelli: That patent said there were ways to address this. It could look at the search history and identify the brands that usually appear, but those brands also appear alongside other brands. So let's surface those too.
Garrett Sussman: Regardless of whether you're looking at Google, Grok, Claude, or OpenAI, the way these algorithms and LLMs are built with their parameters is editorial. It's editorial. It's not just about fairness or who deserves to rank, and this has always been the case with search engines.
It's just as true with LLMs. Someone is ultimately deciding what the generative output should be. I think the philosophical implications of the different models and what they mean for us as consumers getting information and being guided are fascinating.
There's a lot of research about how your perspectives and belief systems can be nudged by using LLMs over time. A lot more research needs to happen, but we need to be aware of it.
Gianluca Fiorelli: Let's see who these big brothers above us will be, deciding things for us. Garrett, it's already been one hour. It's time to wrap up. It was great to have you here again. Do you feel well?
Garrett Sussman: Yes.
Gianluca Fiorelli: It was a fast and quick conversation.
Garrett Sussman: I could talk to you for five hours. You're going to end up creating one of those podcasts that lasts four hours, like the Rogans of the world.
Gianluca Fiorelli: The kind I put on to fall asleep.
Garrett Sussman: It's too easy. It's fun.
Gianluca Fiorelli: And when I wake up, they are still talking.
Garrett Sussman: I know. But it's fun. Thank you so much. This has been an awesome conversation.
Gianluca Fiorelli: You're welcome. I hope to have you back. We've done it twice, and there's no two without three.
Garrett Sussman: I'll be back.
Gianluca Fiorelli: You will be as welcome as you were today. And thank you to all of you for being our guests too. Remember to subscribe to the channel if you want to be notified about new episodes. If you liked this cool conversation with Garrett, give it a like. Please help us grow and grow because we want to grow too. Why not? Thank you, take care, and see you next time.
Podcast Host
Gianluca Fiorelli
With almost 20 years of experience in web marketing, Gianluca Fiorelli is a Strategic and International SEO Consultant who helps businesses improve their visibility and performance on organic search. Gianluca collaborated with clients from various industries and regions, such as Glassdoor, Idealista, Rastreator.com, Outsystems, Chess.com, SIXT Ride, Vegetables by Bayer, Visit California, Gamepix, James Edition and many others.
A very active member of the SEO community, Gianluca daily shares his insights and best practices on SEO, content, Search marketing strategy and the evolution of Search on social media channels such as X, Bluesky and LinkedIn and through the blog on his website: IloveSEO.net.
