Get Your Business Found by AI Search: Practical AEO with Natalia Rice

Charlie Cowan: Alright, let me move everything across here. And move this up onto my main screen. And… present… And we're good to go. Well, welcome, good afternoon, good morning, depending on where you're dialing in from. My name is Charlie. And if this is the first Kowalah Wednesday webinar that you have joined, then you are very welcome. And that's whether you are joining us live, thank you, or whether you're watching on catch up on YouTube. We get together every Wednesday at the same time to spend an hour to dive into one topic that is top of mind for us, either in terms of our own usage of AI within Kowalah, stuff that we're working on, or stuff that is top of mind with the clients and the projects that we're working on.

Charlie Cowan: It's such a fast-moving space, and it gives us this sort of safe space where we can have a conversation about something, and it's useful for you to then take back into your business. And today, I'm absolutely thrilled, because we've got one of our spotlight webinars, where we welcome in a guest who I'm going to introduce you to in a very short moment. Over the next hour, we're going to go through our sort of three-step process for our Wednesday webinars. The bulk of our time today is going to be spent on our main topic, getting your business found by AI search. We will then have a little bit of AI in the news.

Charlie Cowan: Normally, for a spotlight webinar, we wouldn't do AI in the news, because we're focusing on the main topic, but as we shall talk about later on, a new model was launched yesterday, which is shaking up the AI world, and so we wanted to just touch on that very quickly. And then we've got an opportunity for you to ask questions. Now, what I would say is that we love these sessions being interactive as we go, and so whilst I've put this Q&A at the end, do not worry about asking your question as we go, depending on what I or Natalia are saying.

Charlie Cowan: And you're in a Zoom event, and so what you will find is that you have got, depending on whether you're on what screen you're on at the top or the bottom of your screen, you'll see a chat icon, and so you can chat there, and then you can talk to others that are in the session, or to ourselves. And there's also a Q&A bar as well, which might be on the More item, and that is going to give you, yeah, let me open it up on my screen here so that I can keep an eye on it. And then the chat as well is there. So I will try and do a good job of keeping an eye on the questions as they come through. But then, my lovely human colleague, Caitlin, is also online as well, and if Natalia and I are busy speaking, then she'll point us in the right direction.

Charlie Cowan: And so with that, let's get on and introduce Natalia, who's going to be leading us through this session and guiding us through. Natalia, maybe you just want to say hello and say a few words.

Natalia Rice: Hi, everyone. First of all, Charlie, to you, you have this professional podcast host quality voice. I tried to match your level, you know, you sound amazing. Well, I'm really happy to be here. Thank you for inviting me. I've been fascinated with AI visibility, probably for the past year, when I realized how quickly search changes. And so, yeah, I just started going deeper and deeper into the subject, and I have 3 clients now I'm working with. And today, I'll just share what I've learned over the past few months, being specifically zoomed in on that, and how the space is changing, what it means for brands. Even, should I say, if the brand is yourself. I know a couple of people here are running their own solopreneur businesses, and so when I say brands, don't think about, like, Unilever or anything big. You can be a brand, you can be searched for online, so it relates to everyone.

Charlie Cowan: That's fantastic. Well, just to set the scene then, in terms of today's topic, getting your business found by AI search. And let's just make that real in terms of what we're talking about here. You know, 5 years ago, 10 years ago, when people were trying to find something, whether that is a product for them as a consumer or as a business, you know, they're going to Google, they're going to Google search. Increasingly, and I know Natalia's going to walk us through this in a minute, they're not going to Google, they're going to ChatGPT, they're going to Claude, they're going to Gemini. And, you know, a simple question might be, I'm doing this in incognito, because it's not actually our situation. Hey, I'm the CFO of a thousand-person manufacturing company. We've outgrown our accounting platform. I'm about to start a buying process.

Charlie Cowan: I'm interested to know what are some of the steps, what are some of the considerations that I should be thinking about. And, with that in mind, who should we be putting on our shortlist to start thinking about who we arrange meetings with? That would be a very typical question that someone might be asking. Here's my problem. In the past, I may have gone out to an external consultant or contractor to help me with these things, but right now, the path of least resistance, the first step is to go and talk to Claude or to ChatGPT. And the whole focus of this session is, how do we get our companies referenced and cited there so that we're the people that make it onto that shortlist? So with that, having explained the challenge and the problem, let me stop sharing my screen. And, what do I want to do to stop sharing?

Charlie Cowan: Was I sharing?

Natalia Rice: No, you weren't, Charlie, and I'll just start sharing mine. Yay!

Charlie Cowan: Yeah, go on. Okay. Well, never mind. What was on my screen looked really, really good, so there we go.

Natalia Rice: We'll get back to it.

Charlie Cowan: Yeah.

Natalia Rice: No.

Charlie Cowan: We'll get back to that. I'll hand over to you.

Natalia Rice: Thank you. No, it's a nice segue to my little talk, guys. So, I'll start with this, it's literally 7 slides. I want to maybe just remind ourselves how we even got here, what's been happening with search, with information, and where we are now at the moment, and why this shift is so, dare I say, fundamental when it comes to search. So, let's just look, I don't know, 2,000 years back, right? If you want to find something, you just went to the market and talked to people, and, you know, exchanged information, and it was very human to human, human to community, and it was on you to find who you can trust, who is a reliable source, who then maybe has some ulterior motives. So the agency of sourcing information, verifying information, acting on the information was on you.

Natalia Rice: Then, maybe a thousand years later, a book came into the game, and it changed quite a lot. So, it doesn't have to be present in the moment, this information, right? It can be manifested in a certain form, in a physical form. It can be passed from one person to another, and in a way, the knowledge was preserved. This was huge, right? So you didn't have to have a one-to-one connection with a person. You can get your hands on certain knowledge, on information, without knowing the source personally. Massive, right? So this is the knowledge preservation. And then let's move another, what, 500… no, 900 years, right? And enter probably last century, when the modern media arrived. So we have newspapers in the beginning of the century, then we have TV and radio. So you don't have to go to the market, right?

Natalia Rice: So you don't have to pass this one book from one person to another. The information started coming to you. Right? You can sit in your chair, relax, and information gets delivered to you. So that's quite something. And this is when influence started also taking shape. Of course, you can think of advertising agencies and their hard work to package information that could be useful to a consumer of the information. So, ways of playing with narrative and information, this is when it's happening. And so, another 50 years, comes the internet. I don't know if you remember, I do, you probably, Charlie, don't remember, you yawn. I remember those first searches in the late 90s, it did feel like a miracle.

Natalia Rice: You just type in a keyword, and you have everything, so it's not information coming to you, it's, like, coming specifically to your demand, the essential keyword that you're trying to find. This was huge, that felt groundbreaking, right? So, the internet was with us, and still is, but in a way, that's how we conducted our research for a good 20 years. And then, of course, the big moment is when LLMs came in 2023, and when I first, probably the first week ChatGPT was released, I was like, oh, holy moly, the search is never gonna be the same. This is something else. And I haven't changed my opinion since then. It has become different, right?

Natalia Rice: So, we moved on from this, like, put in a keyword, I don't know, shoes, to having this massive, long-tail contextual question that is so unique to you, to your situation, to your content, to your context, and then the LLM brings back the answer specifically for you. This is huge. I was this morning at London Tech Week, I was telling the guys, and one of the speakers was a lady from Trainline who's responsible for AEO. And she said, goodness, 3 years ago, people would just be typing London to Paris, that's it. Now people are saying something, they open the app, and they would say, I'm traveling with my whole family, I have children, 8 and 13, we have barrels of skis. And this request is, like, 6 words… well, much longer, and this is your context that you request. And, of course, the LLM returns an answer, so it's a dramatic change, extremely personalized.

Natalia Rice: The next era will be agents, when you probably will be taking even less part in the search process, so if your personal agents will know you so well, they'll probably run search in the background, so you won't even have to worry very much. And even Google announced recently that that's what will be happening. You will have dedicated agents running search constantly in the background, so if you're waiting for a sale to happen, you will be notified without having to go and check. So, now let's zoom in into our practicalities, into what technically has changed. Who is searching, another one before we go there. So, just to capture what changed, and why it's different. So, pre-internet, the human was in the center. They would source information, comparing, ranking, and hopefully deciding.

Natalia Rice: With the internet, still, information is more organized, but it's still you, the human, who is searching. You're clicking the blue links, right? You're like, okay, I trust, don't trust, looks good, doesn't look good, so… and still, the decision is up to you, right? And comparison is up to you. LLMs have changed that, so the answer comes in a kind of prepackaged way. A lot of things have been done in the background, you haven't seen them. How LLMs rank information, why are you being shown these specific 5 recommendations? What's happening to the others? So, there are a lot of things that you did not take part in, right? They've been done for you, and specifically for your context. And, of course, in the next agentic era, we'll have AI agents that will be acting without even you being in the loop. So what I mean, for example, now.

Natalia Rice: And we are here on level 3, so if I'm searching for a summer camp for my daughter, I'll say, hey, LLM, I'd like to find a summer camp for my daughter, she loves dancing, and so on, and I have a list of results, it's still up to me then to go and act on the results. If I have an agent, I'll say, hey, if you find a camp that qualifies according to these parameters, just go and book it, don't ask me at all. So I won't even be part of that. So that's the next step, and the expectation is we are getting there, obviously with different paces in different areas of life, but that's kind of where we are. Let's talk about the stage we're at today, right? So that's where today's work happens. What is the categorical difference between traditional search and AI search?

Natalia Rice: The first is really how search is done, as we already discussed. So, we used to compare web results, right? So I have links, I click, I compare, I make a decision. LLMs interpret things for you. They understand your request, and then make a judgment, we'll talk about that, and then return already predetermined kind of answers. If you are a brand, your goal was to rank in search. Remember how everything was about where you are in Google search? And this famous joke, where is the best place to hide a dead body, is on page 2 of Google search. So, this is out. We're out of that. So it's still a noble goal, but we know organic website visits are really, really down. So now we aim to become that thing that the LLM recommends. That is our goal as a brand. We want to be in that answer.

Natalia Rice: So what Charlie entered at the beginning of this conversation, when he asked about accounting software. We want to be that accounting software that has been recommended. Brand focus used to be keywords and metadata, so we worked with this equity on our website. We optimized for keywords, we worked with metadata, really teaching Google that we are that thing that the person needs. So now, we're moving to what LLMs call entity associations. Very quickly, it's not a difficult concept, so… LLMs are essentially a representation of our physical world, right? So they try to map where things are and how they relate to each other. So, an entity is pretty much anything, your product, your brand, your service. So, the LLM is trying to understand how those entities are connected, and where exactly you sit, right? So, you need to be very articulate in that.

Natalia Rice: So, that's where brand narrative and messaging is important. So, it is believed that when an LLM scans your website, it acts as, it's a good metaphor to remember, it acts as a busy editor, right? It needs to very quickly understand who you are. If it doesn't, then it struggles to place you in the world of connections. Where exactly do you sit? What is your role? What do you do, right? So, it's important to think about LLMs as a representation of the physical world, and make it clear where you are in relation to everything else, because why? Because an LLM is a vector-connected kind of universe of semantic vectors, right? So relations are everything for an LLM. If you're confused, don't worry, it's not as difficult, promise. So, trust. In the past, we used to rely on backlinks.

Natalia Rice: The more sites refer to us, the better, the more authority we have, the more trust. LLMs want pretty much the same, but a slightly different form. They want more signals from the whole ecosystem, from forums, from discussions, from third-party reviews, to prove that your brand is what it says it is. It wants consistency between what you say about yourself and what signals from everywhere confirm or don't confirm, right? So, if you're saying you're the best software for accounting, we need more than you saying that, right? We need reviews, we need other people praising you, we need case studies somewhere, we need an expert talking about you, then you're good to go in an LLM response. And in this ecosystem, there's the biggest shift of all. If you remember one thing after this webinar, it's this. It used to be that the website was king, right?

Natalia Rice: And we optimized everything on the website to be ranked. This is gone, and I'll show you by how much it's gone. So, what we call now is omnipresence. So, an LLM would be pulling from everywhere to make an opinion about whether your brand is what the customer is looking for. Approximately the website accounts for, well, different research says different things, but, like, 40% of the whole success. So, think about this like, yes, that's about half of the deal, maybe a bit less, and the other half is coming from elsewhere. Synthesis used to be on you, you compare your site, now the LLM prepackages it for you. And the last one, I really think a lot about this now. The LLM recommendation is not neutral, right? So, remember how in Google you just had links, that's all. Everything was on you, on a brand, to convince the person. It's not the case anymore.

Natalia Rice: LLMs have very, very strongly held opinions. When I was building my consultancy, I asked an LLM, shall I buy this software? It said, don't buy this by any means. So it's a strong one, right? It's like, if I was the brand, I would be super upset, but it comes with opinions on how it works, right? Remember I said it pulls from different sources? It does. This is a real case, as many people here who know me, they know I like dance. So this is a real case, I was helping to find a CRM for a dance school, so I put in, what's the best affordable CRM for a dance school that doesn't require a lot of technical knowledge? When I put this question into an LLM, it doesn't just run one request.

Natalia Rice: It sends different requests into different areas to collect this really informed opinion on what brand to recommend for this particular request. For example, when I say best affordable, it probably will go into review websites and find whether it is true, whether this CRM is really best and affordable. Then, CRM for a dance school, it will probably go to websites to scan for information, but also listings of third parties about this particular entity, and technical knowledge is something about product documentation and discussion forums, so it will search for documents. Is it easy to set up? What do people say about it? So, in order to answer this very normal question, it will probably send, not probably, it will send a variety of requests into different areas, and then collect the information. So, this is a company called Yext that likes publishing research on the topic, and I find this very useful.

Natalia Rice: They published this graph recently, basically saying, when it, it's called this, this multiple searches called fanouts, when an LLM sends its fanout to different sources, how much of which is it pulling in? So, according to Yext, this is basically… dark is a source that you have full control of, medium is some control, and no control. So what is full control? Full control is your website, is all your blog posts, and all your kind of press releases, official information written by you. Some control is third-party platforms where the brand can manage its profile. Think about, like, G2. Think about Google, Google profiles, right? So, anywhere where you can place your information, but you can be responsible for it, right?

Natalia Rice: So, then we go into limited control, which is everything where, like, say, Facebook, it's your brand, people talk about you, you can't stop them, but you can be part of it, so you can somehow be there. And the third one, no control, is like independent sources where conversations happen. Think Reddit, think forums that you cannot enter, think everything where you're not present as a brand. So, this is, roughly, how it's done. That's representation. So, when you work with your AEO, it's really two sets of minds, your own equity with your website and everything, and the whole universe talking about you, and whether you can do something about it, and whether you can put your best foot forward in that second part. So, two more minutes, guys, I know I can talk forever, but I'll stop after a couple more slides. What are we measuring?

Natalia Rice: When we talk about AI visibility, what do we even measure? First is presence, which is, are we appearing at all? So, for Charlie, when he typed this request, if I'm an accountancy company, do I even appear as a response to this, like, in a recommendation to this question? Second one is my favorite, sentiment. Sentiment is really about the emotion, and the vibe, you know, what color is it? Is it positive, or negative, or neutral? How exactly are we being talked about? And this is a massive area, and most of the difficulty when it comes to AEO is there. And how the LLM interprets your brand, because you'll see, it is just so different. Share of voice, how often is your brand mentioned compared to competitors? Great sign of whether LLMs favor you for certain things, and maybe not for others.

Natalia Rice: Intent-specific visibility, I do love this one, because for any brand, some use cases are more important than others. If you're the brand who's specifically working with, say, the hospitality industry, you really, really want to come up for requests about, if I'm a cafe and I need a CRM, what do I do? If this is your target, you really care about that stuff. So, intent-specific visibility is important, because it forces you to focus your efforts. And portability. I borrowed this one from an influencer called Kevin, and it's a good one. You will see that, say there are 4 major LLMs, your presentation will be very drastic. They are not the same, they're really different. Gemini is different from ChatGPT, which is different from Perplexity, which is different from Google AI mode, and you will be wondering why, and if we have time, we'll talk about it, but you will be represented differently.

Natalia Rice: So, the question is, is this what we have? Is this success? Is it portable from one system, from one LLM, to another? Or does only one favor us, while the other three dismiss us? Also a signal? Very last one, what if you want to get your hands on, like, what do I do now? So, very easily. Of course, there is specialized software, like Profound, that you can use to measure your success in LLMs. But really, it can be easy. When I want to run a manual, quick research, I will take 12 to 15 prompts and run them across the 4 major LLMs. You will start noticing positioning and narrative gaps. You will start noticing visibility, sentiment, where you're missing, and then what to do. Remember, we have two branches of effort.

Natalia Rice: On-site, and it's clarity. You need to be very clear who you are, and you have to be consistently repeating it. SEO is still important, no broken links, no misdirects, no repeated titles. This is still good practice, this is still important. And, again, an important one, try to balance your brand authenticity for humans with still a crisp narrative for an LLM. Happy to address that later. And off-site, again, notice what's missing. Is it visibility? Is it clarity? Are LLMs confused about who you are, so they're not offering you as a solution to a request? Or maybe it's just being offered for other use cases. See where the problem is. If there's a persistent theme, you'll see it, and then start thinking about how this can be unpacked and maybe offset with signals. And, take care of your third-party assets where you can, where you have influence.

Natalia Rice: So if your Google profile hasn't been taken care of for a while, or the information is old, and it's easy to change, do it. If you can pull in a few positive reviews, do it. So the whole universe is making a huge impact on how you're appearing in LLMs. Sorry, Charlie, I talked for longer than I expected, I'll stop.

Charlie Cowan: I know.

Natalia Rice: Let's go into a conversation.

Charlie Cowan: Yeah. No, I think, well, maybe to start that, if I share my screen, and we can talk a little bit about what you were just touching on, which is, what's being shown to people? So this was the prompt that I put in at the start there. So, I'm the CFO of a 1,000-person manufacturing company. We've outgrown our accounting platform, about to start a buying process, and so on. With that in mind, who should we be putting on our shortlist? And so, this is a typical type of question, and then, I'm in Claude here, and Claude's given me guidance about how I should think about my buying process, we should be doing a defined phase, evaluate, select, and implement, so far so good. And then, this is what I thought was very interesting, is your shortlist.

Charlie Cowan: And it's not some of the companies that I would be thinking of. So when I was putting that prompt in, I had NetSuite in my mind, thinking, right, NetSuite would be an accounting platform. Maybe, a growing company could be Workday Financials. And so what comes up is SAP S/4HANA, Oracle Fusion Cloud, Microsoft Dynamics, Infor, and a few other things. And so I thought this was interesting, because I'm thinking that the companies I would have thought would be coming up are not, so it's a good example of AEO not necessarily working for some companies. So I guess with that in mind, maybe that's just a question about what are the types of questions that we're thinking people are asking, you touched on it a little bit, that it's maybe different and more detailed questions these days than they would have put into Google.

Natalia Rice: People are asking questions with a lot of long context, right? So my guess here, Charlie, if I may, is, are you using your regular account that you use for everyday work?

Charlie Cowan: I'm not, I'm using incognito.

Natalia Rice: Oh, okay, I'll see…

Charlie Cowan: But it doesn't remember that I'm not a CFO of a… third.

Natalia Rice: Yeah, so some signals… it's actually not bad, I think, this list. It's not bad. So, a thousand-person manufacturing company does prompt these big names, such as Oracle and other heavy enterprise solutions. I bet you if you change it for a 250-person company, it will be a different response, so…

Charlie Cowan: Let's try. Let me go.

Natalia Rice: Yeah, so see, it's the context, context matters, yes. So, while Charlie is typing, yes, the way people ask questions is now extremely specific, so that's why when you think about your brand, it's never been as important to know exactly who you are, just because you need to make it easy for an LLM to recommend you, here we go, to recommend you and put you in the right place to match with that context. When I run a manual test, for example, for a client, I will be testing prompts, first to identify a category. Just say, okay, I want accounting software, was there for me. But then the more precise I…

Charlie Cowan: economy.

Natalia Rice: …the narrower the recommendations are. So, I mean, oh, here we go, see? You have Xero now, so…

Charlie Cowan: So, yeah, what I love about this is that on your slide, you talked about how Google search is not opinionated. Here's a list, you go to them, you make your decision. This is very opinionated. Like, if I'm not familiar with this space, I've immediately ruled out Xero and QuickBooks, because Claude has told me to rule them out. And, Claude knows more than I do, so you've likely already outgrown these. So very opinionated, and I've been on the receiving end of this, where I use Claude Code to help develop the Kowalah platform. And, you know, as we were embarking on that, Claude made decisions for me. I mean, I was the one that said, yes, let's do that, but Claude's saying, let's use Supabase as the database for these reasons. Let's use Clerk for authentication for these reasons. I'd never even heard of these companies before, but I'm like, you know better than I do, so this is the direction that we're going. And so, very opinionated in terms of that direction.

Natalia Rice: I want to give you an example from the other side, as a person who looks at it. So, I was working with a company that positions itself as an intelligence layer over a bigger SaaS system, right? So it's a smaller offering on top of the bigger offering. And so, I ran research here across 5 LLMs, and only one said, this is actually great. This intelligence layer is amazing. It will solve you these issues. It's economically, the ROI is amazing. Once you have it, things are gonna look different. For others, it was almost treated as a decoration, as if, okay, you have this big system, this one can probably be dismissed, companies are trying to buy less SaaS. Anyway, don't worry, you'll be fine.

Natalia Rice: So, I'm like, wow, so I have the same product, and one LLM, by the way, which is ChatGPT, and we can talk about why, ChatGPT is much more perceptive than many, actually, which surprises many people. So, ChatGPT recognized the value of that thing, while others didn't. And so, it prompted me to think about the language we use to talk about a product, right? So, if it's so confusing for LLMs, and they see it as an unnecessary addition, then maybe there's something in our positioning, in our language, that basically doesn't do a good enough job to solve that. So, yeah, so if ChatGPT goes everywhere, into Reddit and Quora and all these questions, it can form an opinion, which is actually correct. But if…

Charlie Cowan: Awesome.

Natalia Rice: …someone like Gemini, that tends to use websites and own properties more, it's just not getting it from our own website. So see how it's already changing on different models, and prompting us then to take actions, us marketers, I mean, take actions to be understood by LLMs the way we want to be.

Charlie Cowan: Yeah, yeah. Let's, you were talking about this earlier, so let's go, ChatGPT, and I'll go to a temporary one here as well. And ask that, and let's go to Gemini as well.

Natalia Rice: I love that.

Charlie Cowan: And, do we have a temporary one here? I don't know, maybe not. Anyway, we'll just go. And whilst that's building out, I think, okay, we're looking at 3 different platforms here. Gemini even… Gemini doesn't even work, that's my experience. Anyway, so this is ChatGPT. Bottom line, Dynamics, Oracle, NetSuite, a few other different ones there, maybe Sage Intacct, okay? This is interesting because, depending on what tool you use, Claude, Gemini, Copilot, you're gonna get a different response.

Natalia Rice: Wow.

Charlie Cowan: But then within each one, the model that someone is using as well, whether they're using Sonnet, Opus, they're gonna get a different experience. And, just considering where your customers are most likely to be, the types of tools they're using, I think is an interesting dynamic.

Natalia Rice: Very, and I was looking at, we don't have a final stat on model preferences, but we still know with certainty that most of the market will be using Google search, so it will be, if you want to optimize, Google AI mode will be your number one, probably, target. So then ChatGPT comes second, probably two-thirds of the remaining market. And after that, you'll have Gemini, Perplexity, and Claude. Strangely, Claude is not that big for search. It's big for other things, right? For automation, for agents, but not for search. I know Perplexity is favored by journalists and people in marketing, so who wants to do this factual research data? So, but that's your kind of levels of popularity. I would say, Google AI mode, then ChatGPT, then Gemini, then Perplexity, then Claude.

Charlie Cowan: I mean.

Natalia Rice: Again, not solid market shares, but kind of approximate market shares. And yes, they'll have different preferences, and you will be featured differently in all of them, yeah.

Charlie Cowan: Do you have any sense of the volume of traffic that's coming to websites through LLMs now compared to Google search?

Natalia Rice: So, it's a very good question, because only about 2 weeks ago, actually just this month, Google introduced this new feature in the Google Search Console that shows you not the traffic, but how many times you were showing in Google AI mode, impressions of your brand, and people have been getting crazy about it, like, Google withholding this information, it doesn't want us to show, like, what… because the traffic is so bad. It's a tricky one. You've probably everyone heard about zero-click, right? So, zero-click is a situation when a user asks a question, learns everything they needed to know without ever visiting your website, like Charlie is doing now. Charlie probably learned everything he wanted to know. He doesn't need to click on this website, because everything is here. So, there is a reason for Google not to show you website visits, because they are declining, we won't lie. The numbers are not exciting.

Natalia Rice: So, that's why, that's what I started with. We're now optimizing for appearing within LLMs. We want that. We want to be here, we want to be on that list, we want to be recommended. We want the sentiment to be positive, even if it never results in a click. When I work with brands and they say, well, how do we know? I say, honestly, just buy inbound, like, requests, because if you see some, well, if you just have an inbound request, as if out of nowhere, please do take care, if possible, where possible, to just ask. And it turns out people say, yes, I did find, ChatGPT recommended me to you. I've heard these stories from all over the place, some small design studio in Spain, saying, what is happening? Why are the customers… but apparently somehow it was favored by an LLM.

Natalia Rice: About being recommended, you know, and so, yes, you will see traffic, you might not know where it's coming from, so do ask at this point. We're getting better at measuring it, but again, like Charlie today, for example, you read these things, you probably now close this thing with a few brands on your mind, right? Tomorrow, you take another step. Attribution will suffer. We will not know for sure you were influenced by this discussion.

Charlie Cowan: Yeah, yeah, yeah. I mean, if I am in this fictional, now I'm 250 employees, these searches, but, you know, we've got NetSuite, we've got Dynamics, we've got a few others coming in, but across all of them, it's the same sort of 4 or 5 platforms, and so basically, as a potential buyer, my mind is probably influenced. This is what the market is sort of saying, and there may well be some nice challenger brand with a great salesperson and a great website, but already, I'm being influenced by what I'm seeing, and that's probably going to be my shortlist. And then I'll get in contact with one of them, and it will appear that attribution is inbound, or maybe responding to an eager SDR's email, like, oh, the SDR wrote a great email, but it's all of this research that's happened beforehand.

Natalia Rice: Exactly. And actually, what I'd say is, this situation, when you keep seeing the same brands, it's not that common. Yeah, because they differ. So, no, this is good. This is probably the models kind of merging, not emerging, like, the results are getting closer, but more often than not, you will see different lists of recommendations, which, as I say, some, when you work on the brand side, you will see some LLMs favor you so much more than others.

Charlie Cowan: Yeah.

Natalia Rice: Yeah, like, for example, I have a brand that's done a really good job in the past, working with bloggers and publications in editorial content, and we know for certain, Google loves the tutorial content, so…

Charlie Cowan: Oh, cool.

Natalia Rice: Google, go and search websites, more like traditional media, if that makes sense, or, validated third-party, good, edited articles written by humans. Google loves it. So, if you did a good job as a brand in the past, they will really bring you up in search.

Charlie Cowan: Yeah.

Natalia Rice: And an AI recommendation. So, while ChatGPT shows very different behavior, it cares less about editorial websites, it cares much more about discussion forums, and, strangely, Wikipedia.

Charlie Cowan: And it's.

Natalia Rice: Very high on the ChatGPT list. It cares about Reddit, and all this kind of specialized discussion around your brand, more than clean editorial content. And so, if you didn't feature well in those discussions, you will not be on the top 5 ChatGPT list.

Charlie Cowan: But if.

Natalia Rice: You did feature well within those discussions, even if you didn't have money to invest into brand visibility in a traditional sense, you'll get mentioned, which I think is good news for brands.

Charlie Cowan: Yeah, yeah.

Natalia Rice: Like, for smaller ones, yeah.

Charlie Cowan: Yeah. I'm sort of touching on that, and then moving us into what people can do. This report's come out the last couple of years, but 6sense, an account-based marketing platform for B2B. They've done this research a couple of years in a row now interviewing, I think it's about 4,500 B2B buyers. And by B2B, they're talking about deals that I think were more than, I want to say $100,000 was the threshold, so it's enterprise-type deals, not people buying a pencil. And, they asked them about their buying process, and, very much recommend reading the whole report, but the summary is, from the buyer's perspective, they've gone 70% of the way through their buying process before they speak to a vendor. So, 70% of the way through, the average buying cycle for a buyer was 11 months, or down to 10 months this year.

Charlie Cowan: Now, this contrasts with the salesperson thinking about their sales cycle, which is 4 months or whatever. So the buyer is in the game for 7 months before they speak to vendors. So the second thing, right, the buyers consider that they make the first move 70% of the time, I think, is that right? Anyway, I'll say I know this because I read the report. So I think 80% of the time, the buyer says they made the first move. So this is important, because you've got a sales team, you've got SDRs, they're writing the most amazing email. No, they just happen to have landed their email in the inbox of the person who is now ready to speak. So the buyer thinks, 80% of the time, I pick up the phone first.

Charlie Cowan: And then the third number that is really interesting is that 80% of the time, the first vendor that the buyer speaks to is the one they buy from. So, like, having come from a sales world, when you think you're in control, and we've got our campaign, and we've got our sequence, and we know what we're doing, what the buyer says is, I'm doing 70% of the work myself, 80% of the time, I'll then pick up the phone, and 80% of the time, the person I pick up the phone to is the person I'm gonna buy from. And so, where this feeds back into AEO is, right, if your buyer is spending 70% of the time doing their own thing, not wanting to speak to you, how do you help them in that phase? And it's, for me, it's all of the things they're trying to research.

Charlie Cowan: How do I build my team? How do I build a budget? How do I understand this category? How do I build an RFP? All of the stuff that has got nothing to do with buying the solution, but it's about understanding their world. And so maybe I'll pass that back to you, Natalia, in terms of, as you're thinking about the types of content, not necessarily just where it is, but what's the type of content that is feeding these responses.

Natalia Rice: This is fascinating, I haven't seen it, Charlie, thank you very much. No, that really supports this whole idea that, and I bet you, even when they pick up that phone, they still, in parallel, go and check everything in a little.

Charlie Cowan: Yeah, yeah, yeah, yeah, yeah.

Natalia Rice: You know, so it's definitely an advisor there. Types of content. Really, okay, my big question, the one I'm still answering myself as a consultant in that field, is how do you keep your website and your brand being authentic and still human-oriented, while ticking all the boxes for LLM readability and discoverability, right? So, we want to be so understood by LLMs, so that they kind of know where to position us. There is a good piece of research saying, and I think you can probably trust that, that LLMs make decisions based on the first third of your copy on the website. So, if you didn't say by then who you are, what you're for, you probably lost your game, right? So, for example, if I work with a, we start with a very simple thing. Is it clear, at first glance, who you are and what you do? I'm in technology. You probably, Charlie, noticed that so many times, even the human, I read it, like, what? What? It just takes me…

Charlie Cowan: Transformation! You know, yeah.

Natalia Rice: Everyone, like, what?

Charlie Cowan: Strategic Global Leader.

Natalia Rice: Yeah, exactly. Yeah, so this is out, this doesn't work. So, I spend time with companies to really just say it, just say what it is. So, I would probably keep the main tagline. If you want some brand, some aspirational tagline, I'll keep it, because I'm all for keeping humanity, but the second line needs to say what you are and what you do, then you really have to be clear, if you're talking about technology, where you sit on a tech stack, right? So, that is super important to say, so the LLM doesn't go around trying to locate you, where you are. So, like, remember, entity association, where you are in relation to everything else, right? So this is important to articulate. Google published a recommendation a couple of weeks ago on what is good for LLM search, what does a good site actually look like?

Natalia Rice: And they are very reasonable recommendations. One really caught my eye. It says, publish non-commodity content. Non-commodity, useful content. What is non-commodity content? Well, if we assume the LLM has been trained on the world of knowledge, it knows quite a bit by now, right? So, in order to be cited, in order to be brought up in a conversation, say something that isn't already there in the training data, right? What is it? Well, these are new, original research. These are case studies. They are interesting takes on an existing problem. This is some novelty that you are bringing into this world. If this is the case, then of course there's a higher chance that you will be brought into a conversation. I'll again go off Google because it's still the biggest kind of search engine. The recommendation is, don't go into this game of, okay, step back.

Natalia Rice: Maybe, like, half a year ago, it was like, yeah, let's publish a lot of AI-generated content for each use case, because the LLM will pull it in, great. Well, it's not that great, so the systems, they're not stupid, right? So, they are good, so, don't try to cheat, don't publish 500 pages of the same thing for a tiny differentiator, that's not gonna work. So, content that is genuinely understood, well-structured, clearly written, ideally has some novelty to it, really solves an issue, and the LLM understands where to put it in the entity association. When it comes to the combination of human/machine, I do advise keeping the voice. However, this is still such a murky area, even for me, because, for example, Google says, okay, publish your authentic content, make it super interesting for humans, but then the other recommendation is, make it clear for LLMs.

Natalia Rice: So, which one is that? Humans like narrative, they like some storytelling. LLMs are busy, right, dealing with all of that. They want meaning, they want it quick, they want it now, so you need to be much more crisp, more concise, more, like, condensed. So I'm trying to solve that for my clients. I want to keep their voice, but I still want these bits of the content, like paragraphs that could be easily extracted, and I just imagine questions in my head, like, what… I don't have to imagine, there are tools to know what questions are being asked, right? So I imagine a question, and I'm like, okay, is it a good answer? Like, you say, I am a, let's come back to my dance studio, right? I'm a dance studio, I need a CRM system, is this affordable? Is it easy?

Natalia Rice: So I imagine, is this the question? How does a good answer look like? So it's like, boom, and it's brought as an answer straight into the LLM. So, yes, it's still, I remember it will be read by humans, but it will be processed by LLMs, so it's a kind of interesting balance of being you, but being clear, crisp, and understood. And sometimes it involves some soul-searching, because some brands are not as clear about…

Charlie Cowan: Yeah. Yeah. It's, you know, to bring the conversation to a close, we're obviously in the AI world, and, you know, people would think that we're all in on AI, and it's anti-human, but actually, I'm super bullish about the future of humans. I've got 4 kids, and I'm like, if you get this right, then you've got a great future ahead of you, and this is a good example that, when SEO first came out, and search, people tried to game it, we're gonna come up with these little hacks or whatever, and then Google just clamps down on that. And the golden rule for SEO is, if you write really good, valuable content that helps your target reader or buyer, they will find it, they will link to it, the machine will do what it needs to over a long period of time. And it's exactly the same right now. Like, if you try and game it, they're gonna lock down on all of this stuff, but if you write really good, valuable content that could only have come from you, because it's related to your product or your unique view of the market, then over time, good things are gonna happen, and you're gonna get referenced and shown in these tools.

Natalia Rice: 100%. I just want to add one more thing. So, definitely, all the best practices of SEO are still valid, and on content, I am 100% with you. The new thing with AI search is all the signals we're getting from the rest of the world, right? So, into the responses, so, I'm noticing, and all these LLMs, they are not very forgiving, in a way, so if, I don't know, there was a comment in 2018 left on a popular website about your brand, it may come back and haunt you. So, things like ChatGPT, they will be pulling it into a conversation. So my recommendation in those cases, if it's something that's there, that's bugging you, and LLMs keep bringing it up, try to balance it with new, positive content. If you can, I don't know, submit your technology for a review, send it to an influencer. So we need, essentially, if something happened in the past, but it's reappearing, you need a new signal. You need a new positive signal coming from somewhere. And, yeah, let's, there are ways to be creative. There are review websites, as I said, influencers, even editorial reviews, so we…

Charlie Cowan: We need…

Natalia Rice: Some novel signal that will be louder than the old signal that is, maybe, not very good for your brand reputation.

Charlie Cowan: Yeah.

Natalia Rice: That one's…

Charlie Cowan: I'll give you a little test, and we'll see what happens, because this is a live situation. My wife and I ran a babysitting platform called Koala, and for years we've been trying to get AI to stop confusing it with our company. I have to do it in a temporary chat, because my Claude knows exactly what it is. Okay. Well, we haven't got…

Natalia Rice: You did a good job.

Charlie Cowan: Which is good. I've spent a lot of time going to websites where we were listed as a babysitting platform and saying, please, can you…

Natalia Rice: That's…

Charlie Cowan: It's all…

Natalia Rice: Well done.

Charlie Cowan: I have a…

Natalia Rice: I have a fun exercise for our group. So, there is new research about changing your brand proposition. Like, how long does it take for an LLM to change information about your brand? Say, for example, you change something, either on your website or in the outside ecosystem of your brand. How long does it take for an LLM to present that information to the public? Give me the minimum and maximum number of whatever you think it is. Are they days? Are they weeks? Are they months? Put it in the chat, let's just see what people say. The minimum and maximum amount of time.

Charlie Cowan: Yeah.

Natalia Rice: That it takes an LLM to update the knowledge about your brand? What do you guys think? Like, what do you think it is? I have an answer, I read the research, so I'll tell you.

Charlie Cowan: Yeah, I… Caitlin, you can show me.

Natalia Rice: There you go, yes.

Charlie Cowan: Yeah, I think it used to be long, because it was all about training the models. My experience has been it's getting a lot quicker, because it's supplementing its pre-training knowledge with either post-training, or with what's going on, but, you know, as I'm experiencing, it's taken a while to get it to forget that we don't do babysitting.

Natalia Rice: A couple quarters, okay. Anyone else on August? So, yeah, so today, Caitlin, my colleague, so it's getting insanely fast, and it's really good news for everyone. They update information like this. So, the shortest time for updating information on a website to appear in LLMs is 6 days. And the longest time is 37. So if you updated something, be sure you will know within a month.

Charlie Cowan: Yeah.

Natalia Rice: It will be there. So it's, compared to SEO, it's light years, you know? Super fast.

Charlie Cowan: That's amazing. Natalia, I'm gonna wrap up, I have 2 minutes on AI in the news, so that we finish on time, but thank you so much for everything that we've talked about. I'm not gonna go into detail on this, because we will finish on time, but just yesterday, Anthropic launched a new model, which is called Fable 5. Now, you might have heard in the past few months that they had this sort of secretive, super powerful model called Mythos, which was too powerful to release to the general public, because it was so good at uncovering bugs and cybersecurity issues, that if it got into the wrong hands, it was going to destroy the whole planet. And so, they've still got Mythos, which is in a limited preview for a subset of companies. But Fable 5 is a safer version of Mythos that is able to be released.

Charlie Cowan: Each time we show you these benchmarks, benchmarks, benchmarks, all you need to know is that the numbers keep going up, and so if I just take this agentic coding benchmark up here, ChatGPT 5.5, 58%, Opus, 69%, and this is now at 80.3%. So this is really a frontier model that has come out. It is able to tackle way more complex tasks, way more long-running tasks. And I think if you're in engineering and development, you're going to find this super powerful. If you're working in complex analytics, or biomedical, or healthcare-type research, you're going to find it super, super helpful. You're going to find this in Claude. Just in your model picker, it's up at the top there. Now, you would not be using this day-to-day in the same way that you shouldn't be using Opus day-to-day. Sonnet 4.6 should be your daily workhorse.

Charlie Cowan: But if you are working on very complex tasks, then it's worth knowing that it's there. It uses tokens at twice the rate of Opus 4.8, so there is a cost to it. And, it's included in Team and Pro accounts, certainly for the rest of June, and then they might limit usage for it, and through the API, you can get access to it as well. So, with that, thank you so much, Natalia, for joining us. Thank you for all of our guests that have either joined live or are watching on YouTube. Next week, and this is good timing with Fable coming out, we're going to be talking about token costs.

Charlie Cowan: You might never have heard of a token a year ago, if you're in finance, but now it's going to be one of the biggest items in your P&L, and so we're going to be talking about how you report on that and how you can minimize that. And we'll see you same time next week. Thanks, everyone!

Natalia Rice: And guys…