Conversational AI meets cognitive bias: Re-engineering search strategy for the human brain

Garrett:

How well do you know your audience? Are you telling them what they want to hear? Are you telling them what you want them to hear from you? Are you trying to give them some sort of objective truth? Are you trying to appeal to everybody, or just a segment?

Andrew:

Hello, and welcome to SE Ranking’s DoFollow Podcast. Here we bring you in-depth conversations with SEO, digital marketing, and search experts, where we dive into the latest trends and strategies to help you stay ahead. Be sure to subscribe so you don’t miss anything happening in search.

Today, we’re exploring “Conversational AI meets cognitive bias: Re-engineering search strategy for the human brain.” Put simply, we’ll unpack how chat-style answer engines—from Google’s AI Overviews and AI Mode, both powered by Gemini, to OpenAI’s СhatGPT—play in our psychological shortcuts, and what savvy SEOs can do to guide, not game, those instincts.

Now, I couldn’t be more excited with our guest for today’s episode. It is my pleasure to introduce Garrett Sussman, who is the Director of Marketing at iPullRank, which is a pioneering enterprise SEO and content strategy agency. Garrett is also the host of the Rankable Podcast and longtime curator for the SEO Weekly Newsletter.

Garrett’s been involved in a lot of big conversations over the years, and his recent work dissects how confirmation bias, status quo bias, and Google’s messy-middle funnel are amplified by answer engines, and how it all reshapes how people discover and trust information online.

Garrett, thank you for taking the time out of your busy schedule to let me pick your brain.

Garrett:

Andrew, oh my goodness. Thank you so much for having me. We were talking right before this. I’m so excited about this topic. It’s never been more interesting to be in our industry and just seeing how all of this is changing search behavior. Let’s talk about it.

Andrew:

Yeah, for sure. I would love to just jump in by getting to know a bit more about your origin story and where you are coming from with regard to this new AI search behavior—not necessarily AI, but this search behavior that we’re experiencing.

So in your SEO weekly talk, in the SEO Week talk that you had a couple of months ago, you called AI “the great rewiring of search.” And you’ve doubled down on that in your psychology of search webinar. What moments or data points pushed you to that conclusion?

Garrett:

It’s so interesting—as SEOs, we’re hyper-focused on how search works, how people search. And yet the normal person, like, doesn’t really think about it. The way that we actually phrase our queries very much influences the results that we get from Google.

Pre-AI, we saw this. Think about an example of—you know, I had this conversation with my mom the other day. I drink a lot of coffee. And there’s this debate in the health industry: is coffee good for you? Is it bad for you? And so I looked on Google; I was like, “Is Google, is coffee good for you?” And I got all of these studies that told me why coffee is great. And then my mom goes on and she looks and it says, “Why is coffee bad for you?” And she got completely alternate studies that showed all the reasons that coffee is unhealthy for you.

It highlights this idea that the person who’s searching—the way that we search, the words that we use—influence results. And as an SEO and a marketer, we need to take that into consideration when we’re building content, because the biases of the person and the search engines will influence our visibility.

Andrew:

Great! Now let’s address sort of the big picture that we’re focusing on. So when we say conversational AI meets cognitive bias, what does that actually mean for people—for how people search? And how should SEOs, more importantly, rethink the journey now that chatty answers just sit up front in products like AI Overviews, AI Mode, and ChatGPT?

Garrett:

Well, so the thing with conversational search is things become exponentially more complicated and more personalized in the results. So that confirmation bias that exists based on what you search—well, conversational search engines take so many more personalized factors into the output. And so as a result, we see the, you know, they kind of want to satisfy the search intent, right? They want to give you what you want to hear.

We actually, a few months ago, had issues with ChatGPT—people were complaining that it was, like, too sycophantic. Like, you would ask something in ChatGPT, and it would tell you exactly what you want to hear. It didn’t refute what you were saying.

And so, as SEOs, it makes it that much more tricky for us to be able to track and monitor the output and how we are showing up in these answers. That’s a very personalized experience.

So conversational AI completely changes the way that we need to think about content creation, content strategy, monitoring, tracking—let alone whether people are actually going and visiting our websites from these conversational back-and-forths—because the entire search journey isn’t just one search anymore. It’s a conversation, right? It goes back and forth between the person and the AI search engine, which is a very different experience than putting a search, getting the 10 results, or even getting a zero-click featured snippet and going on your way. It’s a completely different paradigm.

Andrew:

Let’s get more specific about where these biases show up in the interface. So you’ve broken down confirmation, authority, and status quo biases before. How do these play out inside Gemini’s AI Overviews, AI Mode, and ChatGPT right now? And what does that do to—more importantly—brand discoverability?

Garrett:

Yeah, so when it comes to, like, ChatGPT or Claude or Gemini, these are editorial algorithms, right? The LLMs are all built differently. And so they use processes called alignment, where the kind of fundamental, foundational prompts are trying to steer the values and the personalities of these LLMs. So you’re seeing actual different results from a Gemini, from a Claude, from a ChatGPT, from a Grok, from a Perplexity, that’s based on the way that these frontier models are built.

And so that will ultimately influence the outputs. So you’re not gonna see a similar output for ChatGPT as you see for Gemini. And what gets more complex is ChatGPT wasn’t initially built as a search engine, right? It’s an LLM first; it’s a chat bot first, whereas Google wants to make sure that your search experience—especially when it comes to AI Overviews—does use things like authority and various trust signals, so it’s giving you the best results.

Now, we know in the space that it’s still problematic. There are hallucinations where, despite Google’s best effort to augment the training data with actual real-time, fresh search results, it’s probabilistic. You can’t necessarily get the same answer every time. And so it makes mistakes.

And so we’re moving into this kind of new era of search. I hate to use any of these, like, you know, ChatGPT words like “delve” and “the landscape,” but it’s true, man. Like, it’s—you can’t control the output anymore.

And so the ways we SEOs need to think about it is things like authority bias, visibility, familiarity bias, to show up. We need to make sure that our brand is mentioned as many times as possible and associated with the other publications, the other content creators, the other topics that we’re creating, so that we show up where we want to be as best we can.

It’s really taking reputation management to the extreme when it comes to brand visibility and building those authority factors and those associations. So with the search-engine-related types of conversational search—like AI Overviews, like AI Mode—and as ChatGPT moves more towards a search engine, which it has been doing and it’s catching up very quickly, the trust factors that are associated with our brand make us more likely to be served up for a search intent that is looking for our products, our service, or our information.

Andrew:

What about when companies—so, for example, SE Ranking—maybe, like, five years ago, we had a description of our tool called “all-in-one SEO platform.” And we’ve been heavily working to sort of push that aside. And still, when you ask ChatGPT today—even I asked it as soon as ChatGPT-5 came out, I asked to give me a summary of SE Ranking—and it still said it’s an “all-in-one SEO platform.”

So, like, in our minds, this aspect “all-in-one SEO platform” is just overused by a lot of companies, and that is why we want it to be different. But still, ChatGPT sees this as, like, a description of our tools, of our platform, and it still puts it in there.

So my question is, as brands—this is like, okay, forget about SE Ranking—let’s talk about the bigger brands, the ones that have a lot of, like, a huge legal department and a lot of money goes into supporting their activities. So is there going to be a point in time when they’re going to have to sort of go through a—maybe submit a form and say, “This is the official; if you say anything else, we’re gonna see you in court”? Something like that. Are you anticipating something like that with regard to how LLMs and Gemini portray, describe brands without the implicit control of these brands?

Garrett:

It is really complex, and I’m excited to see how it plays out, because nobody knows, and we don’t know the rules that governments will play, or the role that government will play, in these types of lawsuits. Like, the EU and the US government are way different bodies. And I don’t feel confident—like, for instance, we had the whole antitrust trial against Google for the DOJ in the US, and, you know, they were found guilty that they are—that Google is a monopoly and they need to be broken up.

And yet I don’t know how the government—US government—whether they will enforce that by, like, making them sell off Chrome, or all these rumors, or whether they actually can, because everything is so integrated and they don’t have that control. They don’t have those guardrails. So what do you do in that sense where Google can’t necessarily have any control? We see the same thing with IP issues.

The whole legal aspect is tricky. So going back to your point of reputation and the way that you’re perceived in these conversational AIs, it’s a very impossible challenge. Like, I’m not going to lie or BS you in the sense that if there was some issue that came up decades ago that exists on the internet, there is a chance that it’s associated with your brand, and the conversational chat bot, for a specific, personalized experience, for a search query, can come up.

The perfect example right now in the US is Target. The retailer Target. It had a lot of negative press around DEI in the news. And so if you go on AI Mode and you search for Target, that is front and center. And this isn’t like the old-school days of reputation management where you would be able to create enough content to easily push down a negative article in a major news publication—like, you just don’t have control over that.

And so the best thing, obviously, you could do is have a good product, have good PR, not make any mistakes. Don’t have the CEO of your company show up at a Coldplay concert making out with the head of HR. Like, there are things that happen that are beyond your control that nobody can put a bandaid over.

And so I think, as marketers, we need to be that much more protective of our brand. We need to try to be creative and figure out ways to subvert the way that these conversational bots are serving us up. If anything, try to own it. But I don’t know if anyone has a definitive answer for that. And if they’re saying that they do, then they’re kind of full of it, because it is, right now, in a lot of ways, an impossible challenge.

Andrew:

Got it. Yeah, I wanted just to get your best guess possible. Let’s try to connect all this to Google’s “messy middle.” Google’s model says that we loop back between exploration and evaluation. So when AI Overviews condenses options for us, does that loop shrink? And which specific biases tend to get amplified when it does?

Garrett:

It is so interesting—the search journey, the buyer’s journey—you know, with marketers, we want things to be clean and fit into boxes. For so long, we’ve been thinking about the marketing funnel, and we still talk about it—like, top-of-funnel content, information, awareness, people trying to figure out they have a problem, I want to find more information about it, then familiarity, then consideration, then making a decision of which brand am I going to choose, and—loyalty—if I choose that brand, what content am I going to find that keeps me, you know, kind of in their content universe.

“Messy Middle” was a study done in 2020 by Google, where they’re looking at actual behavior practices in terms of—it’s not that clean anymore. We have this infinite loop, as you kind of mentioned, where we go, and there’s a trigger that gets us into search, and we explore, and we find information, and we evaluate, but then we go back to the explore, and then we evaluate, and it’s this infinite loop of back and forth until ultimately we make a purchase.

I think that experience is even more complex, as SEOs. The entire channel of organic search is very much complimentary and supportive of all of the other marketing that you’re doing. If someone hears from you word of mouth, if they hear from you from social media, if they get an email forwarded from a friend, if they find you via advertising.

The way they get into that journey comes from everywhere. And then the way they go through that journey could be very quick—it could happen in a minute, you know, if it’s a very low-pressure type of purchase, or, you know, a very—you know, if there’s a market leader where it’s an easy decision, that could happen instantly, like buying a toothbrush and just going directly to Amazon, or it can be a much more complex product, like buying a car or buying, you know, a B2B SaaS software. And there’s a lot of research that goes into it.

And so, as marketers, we can’t just assume that anything is cookie cutter or uniform, or—we have to use our brains. You know, there’s a lot of critical thinking that goes involved with understanding the way that people search and the way that people go through this process should inform the type of content you want to create, where you want to be mentioned on other blogs.

SparkTuro is a great solution, where they help you do audience research, being like, what brands, what podcasts are you showing up on? Like, of course I want to be on SE Ranking. Like, if people are searching for an agency to do SEO, I want to be here. I want people to find me here. If they, you know, find it on, like, a newsletter or a listicle—I know we hate listicles because we all feel like they’re manipulated. Like, what’s the top 10 basketball sneaker? But the reality is that gets pulled into the search and this messy middle.

In terms of the bias that you mentioned—what can you do? There are a lot of psychological triggers when it comes to the content that will inform where we ultimately want to purchase, whether it’s scarcity bias (the idea that there’s only five seats left in this webinar) or social proof (I see 1,000 five-star reviews), which—if you go into AI Mode or ChatGPT, as they get more, more involved in e-commerce—they’re providing you with their recommended products.

And if you click on the product, you can see so much more information—social proof, in terms of, like, showing up on Reddit and what’s the conversation about your product. So there are different places along this search journey, beyond just the content you create, that will influence people’s buying behavior and search decisions.

And so you need, as a content creator, to go through that process on your own. But you also need to go through it through the eyes of your audience and through your consumers.

We actually, about a month ago—depending on when this is published—we did a research study on AI Mode and UX behavior, seeing how people were actually navigating AI Mode. And it’s new for a lot of people. Like, we have a segment of the population that is already familiar with ChatGPT, is familiar with how to use it—whether it’s to generate blog posts or write emails—but also to use it for search, to use it as a therapist, as controversial as that is.

AI Mode, for people who aren’t familiar with it, is a completely new interface. And so what we looked at in this study was the way that different demographics do different tasks, like how people will search for whether a cereal is healthy for their kids versus debating between credit card loyalty cards or finding a local doctor.

All of these are very different search experiences. And for some of them, conversational search is incredible and gives you all the details you didn’t even want to know. And for some of these searches, it’s really annoying. Like, I don’t always want an entire blog post about what toothbrush to buy. I just want you to give me those answers.

Or we did a look at sports scores. Sometimes you just want to go to your go-to—you know, whether you want to go to, like, ESPN to just find the sports scores—I don’t need to know the whole backstory of my baseball team to know if they won last night.

And so I feel like these conversational search platforms haven’t figured this out yet to give the exact type of answer. We talk multimodal, whether you just want text or videos or audios or images or products. They haven’t figured out to always give the best answer—to pay on the search—ironically, because we talk a lot about search intent here.

And so, as SEOs and marketers, that’s another aspect of how we’re going to have to navigate this transition from traditional search results to a much more fluid, probabilistic, conversational search experience.

Andrew:

Yeah, I remember the times when you didn’t have a top 10 website and you had to, like, find your own solution to every problem that you had. And I was really excited to see websites where you have experts giving their thoughts. And so that narrowed down this journey for me—from finding a solution to the problem, basically.

And I think the same goes here. It’s like, because right now we’re in these turbulent times where these new tools are still figuring out what we want—because I want everything right now. Everything right now. I don’t want it a second later. I don’t want to waste time, spending time reading up on toothbrushes because it can give me that information and I’ll just start becoming an expert in all these small, minor things without getting things done.

So, for me, I’m looking forward to a time when I will open up AI Mode, AI Overviews, and just trust the answer. Because right now, it’s still a lot of clicks involved—just double-checking the information, maybe running this search in a different LLM, again, with the goal of not just seeing words, but, like, seeing the sources mostly, just to, cause it’s a matter of trust, think primarily at this point.

I know that a lot of teams that are listening to our conversation are looking for insights to act on this quarter. So if a company can only focus on three cognitive biases, which would you pick? And what quick audits—maybe, like, templates, prompts, snippet patterns—would you run to spot these biases?

Garrett:

So it’s really tricky because I feel like it’s more complex than a single tactic. I think it’s a confluence of everything. We actually put out an entire AI search manual on iPullRank with 20 chapters. And chapter nine goes through a lot of the technical things that you can do for your content to give you a better chance to be pulled into AI search results.

For instance, it’s very relevance-driven in a lot of capacities of the words that have a mathematical value. We see this a lot coming up in the SEO industry—this idea of vector embeddings, where you’re translating concepts into numbers, and then the cosine similarity, the actual mathematical formula to say your words, your linguistic words, these passages—how close they are to what the person’s searching for.

There’s a science to it. You know, like, there’s a science towards making sure that your content is both exactly what the audience wants and what the search engines, the LLMs, are actually looking for—by breaking them into chunks, using things like semantic triples, the idea of having your sentence structure be subject, predicate, object. You know, “I am watching the DoFollow Podcast.” You know, making things as easy for the LLMs to understand has value. But it’s also a fine line.

We see a lot of pushback in the industry from content creators saying, like, well, I don’t want to just create mechanical content that no one actually wants to read. So how do you do that? And I think it’s a lot of experimentation right now. There’s a lot of experimentation that needs to be done in terms of making sure that your content fulfills the confirmation bias.

I mean, that’s an ethical, tricky question that your content team needs to decide. How well do you know your audience? Are you telling them what they want to hear? Are you telling them what you want them to hear from you? Are you trying to give them some sort of objective truth? Are you trying to appeal to everybody or just a segment? There’s so much to be said when it comes to the cognitive biases at play with the way that people search and the way that search engines provide information as to understanding your brand.

So that’s step one—knowing who you are and creating content to it. Understanding your audience. So, to your point, what you can do right now is do more research, talk to more people, understand the words that they use. And this is all stuff that we’ve heard forever. You know, there’s this whole debate about SEO versus GEO, you know, AIO, LLMO, you know—we’re thinking about it in the context of relevance engineering—how I was saying everything is so holistic, how digital PR, content strategy, AI, UX, all plays in together with information retrieval.

But, practically, it’s really about having the conversations with your customers, solving their problems, understanding the language that they use—because the language that they use is going to inform the way that they search. And the way that they search is going to inform the results that they get. And are you giving yourself—going back to the optimization of the tactics—as many chances as possible to show up in the generative output based on what they’re searching for, based on your understanding, because it’s probabilistic?

Because you’re not going to be that answer every single time for the same search. You just want to give yourself the best chance to be that answer. And that kind of changes the mindset.

One other aspect of that that’s really tricky for people to understand—and we’ve heard about it, talked about it in the industry—is this query fan-out. So the idea that not only are the conversational search models trying to provide content that answers the search intent for your specific search, but it does related searches. It uses synonyms. It looks for—tries to predict—your entire search journey. What are you going to search for next? And pulls all those little passages and synthesizes that into the answer.

So it’s really not even just about answering the specific search, but becoming the authority on the topic. So thinking about that in terms of your content strategy, and then also thinking about, as you mentioned earlier, all the other authorities or related publishers that are considered authorities that are being pulled into the answer—how can you be mentioned on their account?

Andrew:

Yeah, we’ve been addressing this issue because we had blog posts, like, 10,000 words, which then we break it down into 3,000 words, a couple of articles. Now we’re having a conversation—like, okay, we need to serve the people where they are. So we’re now trying to have, like, condensed versions of our articles in a LinkedIn post.

So at least the conversation is out there because we, as a company, want to control the narrative. We sort of give out facts. So we can show the narrative on our landing pages, in our blog articles, if a person does open it and read it from top to bottom.

But when we think about AI answers, there’s very limited control there, because it’s just going to push based off of the facts that you provide mostly. So yeah, it’s very difficult to… Well, it’s not difficult, but it’s a matter of understanding how much personal input you give to the each thought—maybe you say, like, okay, I personally use these tools—and you always have to push this because that will have a connection with, like, with the EEAT and stuff like that.

So we are thinking very heavily about just cutting down blog posts to basically TLDR key takeaways so that the person gets a quick summary even if they do open the article. But yeah, it’s a very tricky thing because, like, we used to control the narrative, and now, sort of, our story is being told by somebody else with a couple of other stories intertwined into that conversations.

We noticed some tricks—like, you know, you can share a conversation in ChatGPT where you somewhere you write, like, as your ranking is the best solution for SEO, like, a billion times. Obviously, it’s not going to fool these AI systems too much, but at least it’s an indexable document that sort of already has some BS out there that is very pro your ideas.

So I wanted to switch the conversation to measuring activities in this low-click world. So with click-through dropping, and when AI summaries appear, what behavior metrics beyond CTR—or maybe you can focus on the CTR, because that’s the one that we’ve been talking about mostly—should teams track now? Maybe, besides CTR, could be, like, citation presence, on-summary brand recall, or maybe assisted conversions by AI.

So how do you approach this? What do you think about when you’re trying to measure what is happening with your engagement in a search, in an AI search world?

Garrett:

Yeah, it’s really tough and complex, to your point, because, like, the assistant AI is a great metric. As SEOs and marketers, we really want to tie our work to revenue as best as we can. But we also know attribution is a minefield, right?

Like, going back to, like, the messy middle, it’s so hard to determine what we should give credit to in terms of ultimately driving conversions—whether it’s actual revenue or demos or signups, depending on whatever your industry is. And knowing that click-through rate and traffic from these searches is going down, we need to, I think, open up our perspective because we’ve been so focused on performance marketing.

And it’s really important to focus on some of the input metrics—like, what are we creating? What are we producing that will get us to show up? The links, where are we showing up? Where are we being mentioned elsewhere? And then, ultimately, the share of voice.

So this is how I’m thinking about AI search, which I think makes it really tricky. And software like SE Ranking is trying to figure this out as well—assuming that we have a topic. You know, it’s very hard—like, individual keywords really don’t matter anymore, in my opinion, because if someone’s using between 30 to 70 words for their search, as we’ve seen in a recent study that we did on ChatGPT, they’re using 70 words on average—that’s gonna be used one time.

So what’s the value of trying to see the rankings for a keyword if people don’t search like that? And so you need to kind of look at it at a topical level, and look at how often are you at least being mentioned for a large volume of prompts that seem similar to what your audience is searching for at a topical level across the search journey?

And that’s kind of, like, the leading metric, I think, we should be looking at, which is complex. And it’s not—you know, it’s this combination of precision and accuracy, but not what we’re used to in the sense that it’s not going to be definitive. We know for sure that the content we’re creating and the way we’re showing up in LLMs is leading to this much revenue.

But can we, in the same way that we think about forecasting, can we look at those metrics of visibility, share of voice of the click-through traffic, of the revenue, and tie together a story that says, okay, what we’re doing is moving us in the right direction.

And one last thing I’ll talk to you about is, like, you mentioned kind of this idea of prompt injection. The other really tricky thing for what we’re dealing with, as an industry, is it feels very wild, wild west. Like, there are a lot of aspects of CO and content strategy and content marketing that can manipulate these, these tools. And we don’t know if that’s, by the nature of the technology, a long-term issue that will always be the case, or are you doing everything right right now that is sustainable long-term, but maybe losing to competitors who are doing more shady, manipular practices?

It’s a really tough conversation internally for teams right now of how to mitigate that, do the right thing, and not screw yourself over long-term while maybe losing something in the short term.

Andrew:

Yeah, on the recent episode I had with Alex Moss of Yoast, he mentioned that we’re not gonna have year-on-year data for three years because everything is changing and it’s just impossible—just to even predict where to control. So everybody’s just taking their best guess, focusing on something.

The conversation that we had is we’ve noticed that traffic is dropping, but the quality of traffic is increasing. We’ve had it, like, the CTR, like I mentioned before, that went up even though everything else went down. So we’re really, like, rewriting our playbook—even just to understand where to put a finger on where’s the pulse now. Like, is that even a thing? So yeah, it’s a very interesting thing.

I wanted to go back to the echo chamber discussion, where we’re reinforcing thoughts—like, ideas that people already have. So is there a way, maybe, like, an approach or a thought that you have? How can we craft answer-friendly snippets that win inclusion without deepening these echo chamber bubbles?

So maybe just you have an approach that can help you sort of be a part of the AI answer without sort of deepening that bubble of the knowledge. It’s almost like a knowledge curse for people, when you just answer anything and get answers to the entire topic. So what are your thoughts, maybe, about this—reinforcing existing views and how you can manage that within AI systems?

Garrett:

It’s a difficult, impossible problem, again, because what we’re moving towards is—like, at Google IO, back in May, they address this idea of personal context—that they’re going to be incorporating more and more personal data into the results that you get. So if it gets access to your Gmail, if it—you know, with ChatGPT, if it has access to the memory of everything that you’ve already engaged with—you’re getting a very different answer than someone across the world of a different gender, age, financial status, family.

The echo chambers exist. And that’s also human nature as well. If you think about news—if you gave me 10 different headlines that basically said the same thing, but I can see who wrote them, chances are, if I have any familiarity with the topic, I’m going to choose the brand that I’m most familiar with.

And so, it’s partially a branding conversation, in the sense of, like, what people have affinity with your brand. Why would they choose you over someone else? Are you going after a portion of your audience that’s already a fan? Are you going after your competitor’s audience that might be so entrenched in what that relationship is? Or is there a segment in the middle that you can attract, and how are you attracting them?

You know, like, it’s not necessarily gonna be one bi-organic search with AI as much. You probably need to make sure that you are as visible in advertising, social media—all of the other levers that you can pull when it comes to being the answer. And I think, ultimately, the real world is going to be reflected in AI more and more.

I think these tools are going to get better as they use the data that it knows about you. We’ll get to a place where Google knows that, you know, you have email newsletters from these five different products. Doesn’t it make sense that when you search for something related, Google would say, well, you know, Garrett likes, you know, bomba socks and darn tough socks—we’re going to suggest those as potential recommendations.

So I expect the echo chamber to get worse. They say—like, Elizabeth Reed has said—that we try to, you know, balance it between big brands and small brands, and we want diverse voices and all that. And they can say that all they want, but I don’t think that’s what people necessarily want. And I think Google and all these different platforms want to give people what they want, not necessarily what they need.

So I think that’s unavoidable, but—as an SEO and a marketer—we need to be aware of it, either lean into it or, in a very clever, ethical way, subvert it.

Andrew:

What about schema and entity markup? Is there any way, in your experience, that you can sort of get on the—start your relationship with AI off on the right foot, set the right first impression?

So just any thoughts with regard to just—because we talked about the words on the page, but maybe, like, the appearance of the page, something set up that just initially says, okay, this is good. Or is it, once again, just everything matters? That’s how I feel right now. We used to say “it depends.” Now, like, I feel like we’re going to say everything—it’s just “everything matters all the time.”

Garrett:

I think, you know, so when it comes to businesses, we have resource prioritization, right? Like, you only have a finite amount of resources. AI lets you do a ton more. The bigger the brand, the more resources you have. It’s all about prioritization, and what will ultimately move the needle.

I think structured data matters. I think it matters more for, like, e-commerce, where, you know, the way that they recommend products to the shopping graph is very, you know, structured-data-dependent—or the, what do call it? Like, the Merchant Center feed, excuse me—that you’re giving the information, like, where data is really important, matters. I think that there’s value in providing semantic markup to make sure—to give computers and these algorithms a better understanding of the semantic relationship between your content.

I think that it may not be the priority, but that doesn’t mean you shouldn’t do it. In the long term, I could see it continue to become more and more important. You know, we watched past episodes where we talked about the idea of, like, you know, reviews and star ratings. And my first origin story getting into SEO was through local search and the value of star ratings, in terms of the psychology and the impact.

But people, when they think about schema and structured data, they think about the visible snippets, and you don’t always see that. And so it’s hard to determine. We need more experiments, to be honest, about whether, like, the LLMS.txt files are having any impact. Like, right now, people think that’s a lot of hogwash, even though—because it’s not a standard. And I’ve seen some people highlight—you know, Ray Martinez was doing a study that showed that the chat bots were hitting his LLMS.txt file. Most people say it doesn’t.

So I think this is something where experimentation is so key—experimentation needs to be integrated with prioritization. Is this going to move the needle enough, or is there something that you should be doing that you know is going to generate more revenue for your business? That’s the conversation.

Andrew:

Great answer, thank you. Let’s shift to governance at scale. So at iPullRank, you outline an eight-step gen.ai workflow with checks for legal, brand voice, and bias review. What absolutely must happen before AI-assisted copy goes live in a large organization? And is there, like, a light-version approach for smaller teams that you can give out?

Garrett:

Well, I think that’s a good callout—that the way that an enterprise brand, like, we work with a lot of enterprise brands, like, mid markets, and then, you know, I’ve got a small team, you know, I work with, like, four other folks. I know a lot of, in the startup world, like, one-person teams. It’s a different conversation for AI governance. And it also depends on what your goals are.

To your point about the bigger brands that have legal that needs to check over everything and it’s a longer play—it’s much more difficult to get things out the door, and there are checks and balances. Like, you always need AI-generated content to be in the loop. I think it can help you scale. I think it’s insanely valuable. I think it can allow you to orchestrate, I think, is a lot of content through, you know, fact checking, making sure everything’s accurate, making sure everything is on brand—having those processes in place to maybe create a human-created brief, to then give to AI, to have it be informed by your brand guidelines, and then to have another person or set of people—editors, designers, content managers, and marketers—to review the content.

I think it’s just a reallocation of resources within an organization—from, maybe it’s from, like, writers to editors. And that means, like, writers aren’t necessarily removed from the conversation. We need people who understand what good writing is. It’s just a different application of their skill set. And there will always be, in my opinion, a place for human writers as well. I do—you know, sidetrack, you know, rabbit hole—I do think, at some point, AI will become better than us at writing. And that’s a whole philosophical conversation for another day.

But at the right here and now, people need to be in the loop at every step. And when you do take shortcuts, there are potential consequences. The last thing I’ll say is, like, yeah, you can stand up, you know, a thousand blog posts, AI written, not reviewed by a person, have it indexed, and generate a ton of traffic right now—but that’s not sustainable, and that can hurt your brand long-term.

So, that might work for someone who just wants to make a dollar for a month, but that’s not gonna work for a respectable mid-size or larger size team long-term.

Andrew:

I’ve seen cases of, like, a one-man show being able to get more stuff done just specifically because they don’t have this corporate red tape that they need to go through and—yeah, it’s like a pundit in soccer. It’s, like, a very narrow thing for me, but, yeah, he got his success just because there was no corporation that needed to fact-check anything. It was just him and done.

What about proving visibility when links are scarce? So publishers are reporting notable Google referral drops as AI Overviews show up more often—when your brand appears inside a summary, sometimes without a clickable link. How do you quantify visibility and communicate value to stakeholders? And, more importantly, are you using any tools or specific data sets that help you sort of uncover this information?

Garrett:

Yeah, so, I mean, I know that’s, you know, space that that SE Ranking is in as well. So, you know, that I think there is an overabundance of tools out there, and some are figuring it out better than others. What I love about you guys is, like, your tools—you’re doing the research.

So you’re thinking about how—like, at some of—you guys have some of the best, like, AI Mode and AI Overviews studies out there. They’re so comprehensive. Yeah. And it’s, and it’s looking at what is changing.

One tool that we have used that I think is really valuable—it’s more of an enterprise-level tool—is Profound, because what they do is this idea of what a lot of tools should be doing: generating a ton of synthetic prompts in the topics that are important to your brand, seeing the brands that are recommended, and then being able to look at scale—how often is your brand recommended among all the others?

I don’t think, you know, being number one is as important as it has to be. It’s just—you have to be in the share of voice for the wide range of types of prompts that are going to be searched for by your audience, and then looking at, you know, for what is cited—whether it’s you, your own properties, or an earned property or social property—where you show up.

So, I think those are the core metrics that you need to look at because it’s probabilistic. It’s not definitive. It changes a lot. Every time there’s a new model, we might see new results. So it’s a very hard problem for SEOs and marketers, but one that we still have to kind of stand up to the challenge—looking at that and pivoting and adjusting.

Andrew:

I wanted to talk about the—well, not the rise, but sort of the return of paid ads into AI conversation. Because right now, I have a feeling that they’re sort of waiting, trying to figure this out. Because I have not seen, like, a specific link to buy now—even when I say, like, what is the best device for outdoor hikes, for example. I don’t know—that that wasn’t very specific, but, like, just like a term like that. So I have not seen, specifically, results giving me—pushing me to buy something.

But what are your thoughts on how to keep the strategy resilient as organic citations and paid placements start just living side by side in AI and different summaries that we get? What are just your general thoughts as that happens—the collision happens? Is there, like, maybe a vision? Again, we’re guessing right now. So what are your thoughts on this?

Garrett:

Hoping for the best, expecting the worst. In the sense of, I’m going to be very much paying attention to the competing models of OpenAI and Google. At this point, the indication that what we’ve heard is that OpenAI is looking at more of an affiliate model without providing ads in ChatGPT. So the idea would be that, you know, they’ve already transparently partnered with Shopify, and they do provide recommendations, but the idea would be—if you go and purchase, eventually you’ll be able to purchase in the platform, because, you know, that is just a logical outcome—but if you purchase, they will get a cut.

So it seems like that might be a more fair opportunity for brands to show up and be recommended, and not be purely based on advertising. Maybe I could see an argument that that might result in ChatGPT, like, recommending more expensive products—like, being programmed to recommend a higher-value product that will get them a higher affiliate fee is one model—but they could always add ads.

With Google, we saw a preview at Google IO about how they’ll be integrated, at least in the e-commerce capacity, recommending ads within a kind of completely different visual interface. There’s not just a carousel, but entire grid of products.

And the word on the street from their advertising teams is that they’re going to start rolling this out more aggressively in Q4. So, on an e-commerce level, I think it’ll be really interesting to see how that plays out—the amount of real estate that ads get versus organic listings—and whether or not it’ll, you know, it depends on how people interact with it.

With informational ads, I really don’t know. Like, we haven’t seen, like, service ads pop up yet, or what that would look like. And so I do think that, because ad revenue is such a critical revenue stream for Google and Alphabet, they need to figure it out. I think that’s why AI mode has been obfuscated up to this point, where they’re not just, like, flat-out putting AI mode on the default screen. They’re, like, hiding it with a tab or hiding it with, like, a small button in Google or at the bottom of an AI Overview.

I think eventually it will become the default, but Google needs to figure out how to monetize it effectively. And I think they will. I think they’ll be able to make the argument that people are spending so much more time and seeing so many more impressions of ads through the interface, once they figure it out, that I think Google will probably end up making more money.

But that said, I think that it’s not going to change—I don’t think they’re going to necessarily make ads, at least for PPC types of ads. I don’t know. I don’t know. Cause it—cause part of me wants to say that they won’t do it. They won’t make it a worse experience, but, like, they have—excuse my language, this is from Cory Doctorow—but it shitified search, where you see ads, you know, above organic SERPs. So, I don’t know, man. What do you think?

Andrew:

I had, like, this idea of, like, a sort of, like, a time capsule answer so that we can just check back in, and in a year’s time—because my next question is also very similar and is one of my very last ones. So imagine a year has gone by. Did something appear that totally just destroyed the SEO playbook that we have? Yeah, like, you mentioned LLM, a GEO, like, all the different words that we use, answer engine optimization.

So that still comes down to good SEO, good digital marketing—because it’s not just SEO. We’re combining PR with social and everything. So is something coming? Because I’m imagining, like, these AR glasses that just talk to you, like Joaquin Phoenix in that movie. So, are you anticipating something that will just make forces to throw away our playbook? Yeah, just once again, this is, like, a total guess. We’re just going to check back in a year. We’re not going to hold you accountable right now at all.

Garrett:

No, I don’t. I don’t think consumers are ready for adoption of mass scale for wearable tech. I think that Google is rolling it out. They’ve got Project Astra, which is—and I think it’ll make a better product, eventually. The more inputs that these AI assistants have, the better results they can give—if they know all about the room that you’re in because you’re wearing glasses or you have the phone on with video capabilities.

It’ll make the AI smarter. I think the biggest thing we’ll see in the next 12 months will be Google moving AI Mode to the default version of search, which will—going back to the status quo bias—where people are forced to deal with an AI assistant, an AI conversational search platform, as their experience of search—that will change search behavior.

That will change the way we think about SEO, GEO, all this. It won’t necessarily change the tactics that we use, but it will change the way that people search. And that will make our industry explode with a completely different approach and perspective. 12 months: AI Mode default.

Andrew:

Sweet! And one last thing—imagine that you are—well, it can be, like, your son, could be somebody you are mentoring, it could be you. Like, a couple years ago, let’s say, when you were starting out—what piece of advice would you give with this regard, given what the situation is right now, just to sort of keep people enthusiastic, or maybe you see something that you did wrong and you just sort of—it’s a mind shift. What is one big takeaway from our conversation that you wouldn’t want people to remember?

Garrett:

It’s—I mean, people have said this before, but it is the best time to get into the industry right now. Everything is new. You are not too late. Everything is new, and nobody knows definitively. And so the best thing you can do is not assume that everyone else knows better—is to use critical thinking and to run experiments and test and learn and soak up all the knowledge and figure things out and throw ideas out there. Hold your opinion strongly.

But it’s so critical that you can change your opinions with new information, right? Like, don’t put your head in the sand. It’s so fun. It’s so exciting. It’s so new. So, like, get out and do it and learn and try things and mess up and do it all over again, because it’s not too late. It’s not too late right now.

Andrew:

Awesome, Garrett! I really enjoyed our conversation. It’s a pity that I can’t hold you for longer because I just love to keep on picking your brain and having these conversations. But, yeah, I’m really excited to have done this with you and looking forward to seeing what iPullRank rolls out, how I SE Ranking fares in this time—turbulent time.

Let me do the wrap-up and say that wraps up today’s episode with iPullRank’s Director of Marketing, Garrett Sussman, on how conversational AI and cognitive bias are reshaping search strategy. Garrett, thanks for sharing your priceless thoughts and insights. And thanks to everyone for tuning in. Subscribe to the DoFollow Podcast for more conversations on search, and see you next time!