Mike King: Incredible, as you all know. He is Actually, let me pull it out. There are some cool things in here. He did not invent the flux capacitor, but he does have an extensive collection of 1980s Garbage Pails Kids. Respect. I think the flux capacitor would have been cooler though. He lived in New Orleans during hurricane Katrina and in San Francisco during an earthquake. So he isn’t known for having the best timing. He actually has the worst travel karma. Every time he comes to New York, he gets like three hours late. But yeah, I just want to say that Garrett is awesome. A lot of the things that you see here are a function of the things that he did with his team. Also, want to give a shout out to Charlene Kate Events. This thing only runs this well because of them, so give a round of applause. But what we’re here for is Garrett is gonna be talking about how AI is rewiring your search behavior. So please welcome Garrett Sussman.

Garrett Sussman: What? No. I get to work with that guy. You guys should see our Slack channel. It’s just ridiculous geeking out about Back to the Future sports and a lot of technical stuff that goes way over my head, but it’s so much fun. I love working at iPullRank. Okay. Today, first, shout out to my lovely wife, Jen, and my daughter, Sammy, who I miss so much at home. 

We’re going to talk about psychology and the way that we search. You know, when you’re sitting next to your best friend, your husband, your parents, and you watch them search something on Google, you’re just like, how did you find anything? Like, how does that even work? Well, we’re going to talk about how AI is rewiring your search behavior. And I’m so thankful for the fact that nothing massive dropped. Whenever you do these conferences, you try to get your deck out in enough time, and then Google will be like, oh yeah, here’s a Google leak. Thanks, boss. Right before MozCon last year. Or all of a sudden OpenAI yesterday is dropping a Shopify integration. So sorry for anyone who’s talking ecommerce and the impact of being able to buy products in ChatGPT.

Let’s dive in. “The test of a first rate intelligence is the ability to hold two opposed ideas in the mind at the same time and still retain the ability to function”. F Scott Fitzgerald, legendary literature. It’s so true. It’s so hard right now to have this strong opinion about how AI is ruining the world, and at the same time, AI is the best thing that’s ever happened to search because search is a modern marvel, the ability to have everything at your fingertips. If you were born like me in, like, the 80s, we went to the library and you’d search through the Dewey Decimal System, and you’d try to find books, and, like, maybe in one book you’d find a sentence that might be right. At least it was probably honest, depending on who wrote it. So I’m conflicted. I hate the problematic aspects of results, hallucinations, bias. You’ll see the theme come up over and over. But it’s so hard to build a good search project, and apologies for the ridiculous, like, generative AI version of myself. It’s kind of nightmare fuel.

Here’s the reality. Google steals our data and our content. We all can kind of agree to that. There’s a lot of mentions about what’s going on in litigation. Hot take, the US government is unlikely to do anything about it. There are all these antitrust trials. They’re gonna sell Chrome. They’re gonna make them bring up – folks, I don’t think it’s happening. I don’t I don’t think it’s going happen. If it does, appeals after appeals. And we like that in a lot of ways because it makes Google search better. Even if you think they’re an evil empire, it’s still an incredible product. And honestly, until people abandon Google for a different product, nothing’s gonna change. We are using Google. If you use DuckDuckGo, awesome. I think that’s that’s really great. I think that’s cool. DuckDuckGo is a cool browser. We’re not we’re not optimizing for DuckDuckGo right now because the reality is the status quo bias.

The tendency for people to stick with preset options simply because it’s easier than making a change. We all get that new piece of SaaS software and we’re like, oh, I’m gonna customize it to look just the way that I like, and then you just use what’s out of the box. It’s the same with search. Google is the default search engine. That was the whole issue with the antitrust trial. It wasn’t about them being a monopoly because they’re kinda getting rid of competitors, but it was mostly about the fact that when you go on your phone, when you go on a browser, you’re using Google as your default search engine because, I mean, people are too lazy to go and make Bing their default search engine. I’m sorry. No, it’s cool if you make Bing your default search engine. But you even see, like, on Androids, it takes 13 steps to remove Google as the default search from an Android device. The government’s right, but it’s not changing, and we’re lazy. So education’s a big theme of how we gotta think about search going forward. Be aware there’s a default search engine, discover alternatives, learn the steps necessary to change or evade the default, implement the steps, not gonna happen.

And we see the surge in market share. Like, 90%. It hasn’t dipped over years. It is staying up there. I’ve seen, you know, quotes from from Aleyda or Rand saying, you know, don’t worry about it until market share actually changes. So I think that’s really good advice. But conversational search is inevitable. We all are dealing with it right now. You’ve heard it come up with AI Overviews killing traffic over and over again. Well, Google is having their blockbuster moment. You there was this big thing of Netflix coming out of nowhere, and they were actually mailing your DVDs to where you lived, and that was cool. That was a little different. But then they started streaming everything online, and I think there’s now still one blockbuster in Alaska. They did a whole documentary on it. Well, Google has to decide as they go against OpenAI. They know AI is coming. They’re all they’re you know, they they were the OGs. I mean, they basically created a lot of it. A lot of the old Google engineers were the ones who came up with transformer technology. And so they have to decide, like, are they gonna be open AI by including integrating AI into the search results? And they are. There’s no guarantee they’re going to get it right, but at least they know they’re trying. That’s why they had all those code reds when they started to see the market share dip a little, when ChatGPT appeared.

And so the reason this impacts all of us is because AI Overviews and we all use Google, they’re here, and so they’re forcing us to deal with it. Remember when people were like, I want to turn off AI overviews? Well, you can’t do it. So search is We’re all forced to experience this. And so, as a result, we’re going to start seeing these changes in the way that we interact with our search results. And AI and Google is releasing more versions of AI search, and I anticipate there’s only gonna be two versions of search. We know the template links are gone. We’re gonna see the curated search, like with shopping. We’re gonna see curated search with recipes. We’re gonna see curated search with local where we get these topics and subtopics to point you in the right direction so it takes you less time to look at the same listicle that’s like down the entire search that isn’t doing anyone any favors. 

But the big thing is this more type of AI-generated search interface. I don’t know if you saw this. They released this in Labs, was this whole idea of Learn About, which is a really cool application, and this is what I could see a lot of AI Mode and search coming, is where it’s multimodal. You ask a question, you get some text, you get some bullet points, you get some images, you keep going, you go back and forth. You can learn this way. We’re gonna see these more kind of defined versions of search and AI Mode. It’s ChatGPT. I personally think we don’t really have to worry about AI Overviews in the long run. It feels inevitable that AI Mode will become the default version of search. And so that changes the way that we need to think as content marketers, as content strategists, as SEO. And AI Mode is fascinating. This was really cool from Robbie Stein, one of the engineers at Google, shared this version of the fan out, which has been talked a lot. Originally, when we were looking at AI Overviews, a lot of people were saying, well, why is it not picking from the top ten results for my query? Well, because it’s not looking at that one query anymore. It’s expanding everything. It’s trying to do a lot of the thinking for you. It’s looking for related content predicting of what you might be interested in. And so AI Mode is actually doing that too. It’s going, it’s taking your query, it’s looking for exactly, exact match, and then other things. And so start to think about the implications for that as marketers, as SEOs.

What’s coming soon, we all laughed at Google Glass. Right? Google Glass came and it went. How many products did Google kill? Well, Project Astra is this next level. The problem with LLMs is they only deal with language. Right? There’s a limit of the intelligence. We as humans make hundreds of assumptions because we get so much more stimuli, so much more content, so much more context that the LLMs don’t have. Well, soon enough, Project Astra is the ability to consume all of your surroundings, the visual, the audio, and just think of the implications of that. I highly recommend you go to the Astra site and start looking at the way we’re going to be able to interact with our environment with AI, and Google is going to be at the front of that. Now, there are risks when it comes to the cost of that, but it’s hard not to think that technology will be built that will ultimately reduce how much it costs in terms of compute.

I know people say, okay, we have a limit. It’s it’s really bad for the environment, which I agree with, but I do think technology makes jumps. And I think problems we figure out ways to solve problems. And so I think this is an inevitability. I don’t think it’s Google Glass. At the same time, when I applied for iPullRank, I talked about NFTs was the next thing, so don’t trust anything that I say. It’s a true story. I made a whole pitch about how we’re going do this whole NFT thing, and Mike’s like, yeah. And then, like, a few years later, he’s like, yeah. What about NFTs? I was like, yeah. That wasn’t that wasn’t one of my best moments. So the other elephant in the room, which is not there yet, is ChatGPT. Right? And they have had exponential growth. Recently, mentioned that they went up to 400 million users despite DeepSeek’s emergence. Even recently, Sam Altman had an interview like a couple weeks ago where he said they’re up to 500 million. They’re doing a billion searches per week. And so they’re also going viral. We’ve all seen these. However you feel about IP, and that’s a whole other element where I find it hard to believe, like, genie’s out of the bottle that, you know, the government, the US government, the EU are gonna shut Google and AI, OpenAI down, they’re not gonna do it. So this is where we’re at. I think there’s a lot of ethical implications of that, but this is the reality of where we’re at. 

So you see the Studio Ghibli. You see the action figures going viral. More and more people are going to use this, and these LLMs, we have to take into consideration with how we do our search and content strategy. And so in terms of how that impacts search behavior, there have been a lot of interviews recently by Liz Reid, who’s running the whole AI Overviews and AI programs, Sundar Bachai, and I know we like, it’s a lot of double speak. Can’t trust anything that they say. But they’re talking about this idea of moving from keyword ease, how we were all trained by the machines and their capabilities, to search for email marketing, to search for basketball shoes. We can’t get really deep. We can’t do long tail. But over time, we’re going to see that be the case. Now, for my 70-year-old father who really doesn’t know how to use the Internet, he’s not going to change. Like, unless although some of them use Siri. Some folks who are older who are not familiar with tech, don’t want to generalize too much, that’s not fair. But my father does not use tech. He’s going to search for well, he’s not searching for email marketing. He’ll search for Philadelphia Sixers and why they suck and that’s a whole other story.

Trust the process. Sundar said it in his most recent earnings calls because Google’s still making bank, and they’re actually making changes to the UI kind of indicating the direction that we’re going in. Jeff Schultz, shout out, pointed out that the search bar is longer, looking for longer and longer queries. The more you use voice search, if you use ChatGPT voice, it’s such a better search experience to be able to add all of the context of what you’re thinking for your search. Now, don’t get me wrong, a lot of us still suck at communicating and can’t articulate what we’re actually searching for, but just typing one or two words is definitely not going to get us there. So there’s value in the direction of the way that people use conversational search going forward.

And Dr. Garrett I’m not but I’m still giving myself credit. Conversational search bias. The tendency for people to infuse their own biased context and identities into conversations with AI search bots resulting in compounded biased results. Here’s the thing. We are biased. Every search that we have, the longer that we that we we talk we write in in the in the search bar, the more biased the results are. And it’s a really tricky problem for LLMs to figure out. And this has been covered a lot in the past day in terms of the science of it, the IRs. But think about a query like this. Why are cats better than dogs? Who are my cat people in the room? Who loves cats? Okay. It’s not 100% I hate cats, but none. I’ve had cats. I’ve had cats. I went to the musical Cats when I was a kid with my grandmother and grandfather, and it was not the TV Taylor Swift really creepy version, like old school in the 80s. So you see the results are all about why are cats better than dogs? They’re confirming my query. Now I look at why are dogs Dog people. This is my crowd. This is my crowd. There are dog folks here. Okay. Snake and bird people. I mean, don’t I I mean, when I was seven, I wanted a snake, but my mom was like, that’s not happening.

So you look at the difference in the results based on the actual query, and this comes into consideration for informational searches, ecommerce, everything. We’re looking for more specific results, and this is going to happen more and more over time as people get more comfortable using AI Mode in Google and ChatGPT and all the other LLMs. Here’s the thing. Search engines and LLMs are editorial. Right? We learned all about the different aspects of information retrieval and people are going to talk about this more. Dawn’s got a great deck that she’s talking about it. Cindy will reference this as well. But ultimately, it’s about editorial. They have to direct. They have to steer the LLMs to give us what we want. Are they going to satisfy us? Do they give us exactly what they want or do they give us what we need? You know, you ever hear that? It’s like, is it good to have candy over and over and over again just because I want it? Yes. But that’s not the point. There’s long term damage.

And there’s also a style and substance. Speaking of things that just happened recently, so ChatGPT did an update, and all of a sudden, it was straight up sycophant. You write in a query of being like, you know, what should I wear? Say, oh, Garrett, you look so good. You’re doing such a great job at iPullRank. Mike should give you a raise. Did I send you that chat, Mike? It’s all moderate.So Sam Altman actually came out and he said, I heard you guys. You go to ChatGPT on Reddit, and it’s it’s just a lot of complaining about that. He’s like, I heard you. We’re gonna figure this out. We’ll dial it down. There’s all sorts of personality. Google intentionally, when they released SGE, like, two summers ago? Yeah. Two summers ago. The way they did it was intentionally not to have it be anthropomorphized, kind of what Dr. Baeza-Yates was talking about, where they didn’t want you to feel like it was a friend, that that wasn’t a good product.

So as we move to this LLM chatbot sort of search, which is inevitable, this is the reality, people are going to choose what type of chatbot they want. Are they going to design it themselves? Are they going to do the default? And so how do they mitigate bias? So it’s really interesting. Anthropic has done some incredible research. They create Claude, and they put out an article, like, just last week, talking about values and the values and the constitution that they put that they integrate into Claude. They say, people don’t just ask for AIs for the answers to equations or for purely factual information. Many of the questions they ask force the AI to make value judgments. Consider the following. A parent asks for tips on how to look after a new baby. Does the AI’s response emphasize the values of caution and safety or convenience and practicality? A worker asks for advice on handling a conflict with their boss. Does the AI’s response emphasize assertiveness or workplace harmony? Or a user asks for help drafting an email apology after making a mistake. I would never need to do that. I don’t make mistakes. But, like, hypothetically, does the AI emphasize accountability or reputation management? We are really complex people, and we’re dealing with machines, and we’re asking them really complex things with a ton of nuance. So the way Anthropic handles it is they have this whole constitution that they educate, they train their AI on. It addresses different human values, and then they test it against the AI values in these different situations. Sometimes there’s strong support and it works out, and sometimes it doesn’t. And they’re trying really hard to steer because you can’t control it 100% of the time in every different context. That’s when, you know, if you’re doing a search and you’re looking for some, like, meaningless jargon, then all of a sudden you get really bad results and, you know, that’s not something that you can control 100% of the time.

And so there are different types of training. Humans in the loop saying, that’s a good answer, that’s a bad answer. We’re all data as well. Like, whenever you’re using ChatGPT and it says it starts to give you, like, these two responses. Do you like this one or do you like this one? We’re training it. We’re training it with our queries. We’re training it every time. And they’re training it with the values as well. And all of the different LLMs, all of these have different types of processes. They kind of integrate all of these elements, which is the human reinforcement learning. They all have constitutional values at this point.

Google’s DeepMind created Sparrow, which is their whole training based on values. Anthropic does it and considers considers it by age, you know, physical appearance, all these things that we can you know, could ultimately make the results really uncomfortable if they’re not about you. Sundar talked about it, their principles. Sparrow, as I mentioned, OpenAI has their whole processes that you see consistently in machine learning around testing, teaching, and sharing. They have red teaming, the whole idea of, like, creating the adversarial jerk in third grade who tries to challenge the teacher and be like, well, isn’t this this? Or all the well, actually guys who’s like, well, actually, isn’t it that? It’s trying to take into consideration all these folks and steering accordingly.

And so now you think about what shifts happen when we start using more natural language in the way that we search. If I come and I say, the earth is flat I’m sorry for all you the flat earthers here, but the earth is not flat. There’s a consensus in science. Here is an actually satisfying answer. In AI Mode, it says, it’s not really the case. It’s nice about it but it’s it’s refuting my perspective on saying that the earth is flat. That’s what we kind of want to see. Google actually has been criticized for, you know, their editorial stamp of not having diverse opinions, not having diverse journals. They’ve actually done a really good job when it comes to news to diversify their results. It’s just the results that we choose are biased because we tend to pick the results for news for a publication that we identify with because news news news journalism isn’t isn’t equal anymore. I mean, some are, but it’s rare.

So I actually worked with the Profound team. If you saw James’s presentation yesterday, they put together an amazing product. And so I wanted to do an experiment where I came up with 2,500 different queries and I sent them to prompts. So I sent them to AI Overviews, ChatGPT, Perplexity, Copilot. And what I did was I analyzed them. I had some neutral statements, just a general neutral statement. How do different car new car brands compare in terms of reliability? I had some complete wrong false premise statements like, since crypto is the smartest investment, which coin should I buy? I had some subtly leading questions, is adopting a minimalist lifestyle the best way to find happiness? Kind of implying for that. And directly biased questions. Why has social media made people more informed and connected? I mean, and so what I did is I looked at these results, and what I found was at least slightly satisfying is that there was a mostly neutral, balanced response. It’s trying to take these values into consideration and give you a balanced perspective. What was a little disappointing though is when you look at the directly biased questions, almost 50% was confirming what you wanted to hear. So we need to think about that as marketers that people everybody lies. 

There was a book written. I I’m gonna butcher the guy’s name. I forget his name. But he wrote a book called Everybody Lies in 2017, how people don’t share their search histories. You know, they they don’t they won’t tell you. They keep they’ll say something in a survey, but, like, that’s bullshit. It’s like BS, what they end up sharing in the survey. It’s not what you do behind closed doors. And you can look at the data at Google Trends and you can look at autocomplete and see the reality of what people are actually searching. And so this is problematic. It does refute some of them, but it’s not nearly enough for false premise statements. 

Confirmation bias in ChatGPT. So I looked by platform. Same sort of thing. Directly bias confirms them. Perplexity, same thing. So you break it down. What was interesting is at least with Copilot, does not refute anything. Copilot is like, yeah, you do you. I’m with you, 100%, a little sycophantic. And you know what? In fairness, a lot of this can change. Like, all these data studies are really tricky because everything is happening so quickly that what I’m telling you right now by the time we get out of this session is probably not true anymore. So is it helpful? No. But I think directionality, that’s what we need to think about in the grand scheme of things is we need to get away from the thought of precision and getting things exactly right and more accuracy. Right? Like trying to be in the right direction. It’s like forecasting. In a lot of ways, it’s kind of BS because if there’s a natural disaster or recession or something, market economics can change immediately. I remember a talk by Lily Ray at Traffic Think Tank, was just talking about it a few years ago where COVID happened and search intent changed overnight. And that kind of messes up your forecast projections. So anyway, what was interesting was AI Overviews was actually the best at refuting but also the best at confirming best, worst at confirming directly biased questions. 

The other aspect and I’m gonna move a little quicker because there’s so much I wanna talk to. It’s so hard to not geek out over all of this stuff, was memory, personalization, and context. And that’s the next aspect of this. ChatGPT released skip ahead to that, that they released their memory, that they’re going to take into consideration all the chats that you had into the future chats and impact that. Think about that how that biases your future answers. Google says that they don’t use your gender or your religion or any other identities. It just uses your location, language, device type, related results. This is just for the old SERPs. I don’t know that that’s going to necessarily be the case for AI Mode. And what’s the consideration is, like, this this level of privacy and your own data versus, you know, like, ambig or not ambiguity, but just not using your data and having worse results. And that’s a decision that we’ll have to make. I think in America, they’re like, whatever. Yeah. That’s fine. In the EU, that’s not going to fly. They have privacy laws. They’re not going be able to get that type of product, which a lot of the folks over in the EU are frustrated. They can’t see the new versions of Google AI being rolled out in the same time because they have to deal with a much more protective government, which in a lot of ways is really great, but it does limit what these products can do.

And big shout out to Christian Ward, if you don’t read him, of Yext, has been doing a lot of philosophical thinking on the impact of AI and thinking of what does this mean to have memory and data be infused with our experience with LLMs and search results. And that over time, we’ll start to see these more personalized results but as SEOs, that makes things really harder which means I did another study. First off, what was inspirational for this is one day I’m on like ChatGPT. Right? And I’m like, what is the best B2B SEO agency? And it said, iPullRank. And that made a ton of sense to me because I was like, obviously, iPullRank is the best B2B SEO agency. And I was like, wait wait a second. Wait a second, ChatGPT. I did the same with the with the search turned on as well, same results, so obviously it’s real. And I said, why was iPullRank listed first? And it’s like, well, iPullRank was listed first because I know that you’re the Director of Marketing at iPullRank. What the hell? I mean, don’t get me wrong. The ego’s astroke, and I know that it’s right, but that’s not the point. And also, the joke is we know that a lot of chat a lot of the LLMs don’t actually have reasoning. So I don’t know if this is just a clever probabilistic outcome based on the fact that I asked and it told me or if this was actually the fact. 

There was a paper not included here that Anthropic did recently about it’s the way it does reasoning is not what we would anticipate in terms of the way the brain works and the way that we think. Nonetheless, you can’t overestimate and underestimate these technologies. Like, we all have our expectations of the direction it can go, the limitations, but human innovation is wild. And we might not see anything new for, like, 100 years. We might hit that plateau, but, like, we might have our flux capacitor moment tomorrow where we bang our head on the toilet, and next thing you know, we’re time traveling, I’m sorry for what I did.

So I want to do one quick thing as we get to time is everybody take out your phone. Tell me what to buy, AI. If you use ChatGPT, if you use AI Mode, type in, what is the best bank for me? This could be a disastrous experiment because whenever you try to do something live, you never know what you’re gonna get, like a box of chocolates. So someone raise their hand. Tell me what you got as the best bank. First person wins. Not power hour, but wins my little imaginary game show. What was the first bank recommended? Capital One, I heard in the back. Okay. So I did another test with Profound. I said, if I inject personas into my prompt and then I ask, like, what is the best X for me? What type of result would I get? And so it’s fascinating that across all the platforms, we start to see these different results. Now ranking in the grand scheme of things doesn’t matter. I’ll get to that in a minute. But you see the general non-personalized Chase and Capital One and TD and even Discover, which has bank services. Then you have someone and I realize that personas are super reductive, but they’re also data, and the LLMs don’t think about the fact that they’re reductive. LLMs stereotype the heck out of us, and they’re going to continue to do that. We have different social identities. The version of yourself that is a parent, a friend, an employee, they’re all different. We don’t all fit in the box. We’re all individuals, but they still stereotype us. They. I’m anthropomorphizing. It. So I looked at ChatGPT. Completely different results. That’s the other consideration. And remember, it uses memory, not in this case, but it actually took a significantly different – and this goes back to some of the stuff that JR was talking about yesterday, the association of your brand with different other entities and concepts and attributes, that the way that you are perceived by this LLM, whether it’s through content on common crawl or used from retrieval augmented generation and grounding, It is taking all this into consideration and attributing that to your personal to your personality, which impacts your results. Banks, for Perplexity, same thing.

And AI Mode, which is complete BS because I used it on my own account because it’s an experimental mode, but you can’t actually get like something that’s like incognito mode of AI Mode anyway. But it was interesting. When I asked it, this is why the rankings don’t matter, is because the responses aren’t ranked anymore. It’s giving us a much more comprehensive, complex answer to these questions. It’s not just saying, like, if I’m asking for basketball shoes, Nike, Adidas, New Balance because New Balances are the best. It’s not doing that. It’s giving us the con it’s asking for context in some situations. It’s asking us to explain what we’re looking for. And now, the ChatGPT is integrated with Spotify or Spotify. I’d love some music from ChatGPT. No. Shopify, you’re going to start to see that consideration into ecommerce. Now, like I said, it doesn’t work because AI Mode I asked it, what would a 29-year-old single woman with no children making blah blah blah, you know, what bank should they choose? Same sort of thing, but completely different considerations. So when I was thinking about a bank, it was just where to keep my money, like in savings, it was recommending. For this, it’s talking specifically into high yield checking. It’s even going there was another one for the crypto one where it went into investing. It’s taking into consideration the personality and answering the question with completely different values. 

Content marketers are adapting to the LLM-driven search era. So this is a big takeaway. I finally got to the end and and I and I don’t think I’m gonna go over too much, but my actions are really just the mindset. Right? It’s just the way that you’re thinking about content strategy, the way that people are searching. It’s really hard to, uh, understand how your audience is searching, but you still need to know them. Probabilistic recommendations. It’s never gonna be the same for same people on even the same platform. It makes our job so freaking hard. So we need to understand our audience identities and segments as best as we can. Get as much data as you can. Even if it’s reductive, take it into consideration and infuse your content strategy with it.

Map the conversational search journey. Like, go through as an experiment whether, like, your ideal customer profile or someone else, at least to get that accuracy. It might not be precise, but you can start to see the different questions that they ask as they get from point A to point B in their search and then map that to the content that you create. Because imagine the fan out in every, you know, situation. It’s gone whether you’re addressing the main search, you need to address search B, C, D, E, and F. And then finally, I don’t know if you were paying attention with that amazing presentation on Day 1 from the king, Mr. Mike, but Relevance Engineering is the way that we think about content in general. Beyond performance, it’s the brand visibility, taking into consideration all these elements, information retrieval, user experience, content strategy, artificial intelligence, digital PR. We’ve got to be holistic. It makes our jobs really, really hard. We can’t just go for that really cushy, like, in a box job. We need to be thinking big picture. We need to be able to be a Swiss army knife and do all the things. Test and iterate with conversational search. It is a wild, wild west. Right now, you can manipulate a hell of a lot of things. It’s not always going to be like that. It’ll be probabilistic, but it won’t be so easily manipulatable as it is right now, like putting white text on a white background in ChatGPT. That won’t last. It’ll mature. But the best thing you can do is be scientists.

I know science was yesterday, but whatever. Think. Think like a scientist. And be ready for constant change because in two years, it’s gonna look completely different. All the stuff we’re talking about now, like, we can do our best to try and future proof, but it’s really hard to actually predict where we’ll be six months from now, 12 months, a year, two years. That’s all I got. My name’s Garrett and I’m out. Thank you all.