← Using AI at Work: AI in the Workplace & Generative AI for Business Leaders
Using AI at Work: AI in the Workplace & Generative AI for Business Leaders
107: Using Generative Engine Optimization (GEO) to Win the AI Search Race with Cole Casperson
00:55:06
Loading summary
Transcript94 lines
- [00:00]
A
AI is better at putting the right products in front of the right customers.
- [00:04]
B
Are you seeing that Google searches are now being also influenced because more people are using the large language models?
- [00:10]
A
People are not using Google search less, they're just using AI models more.
- [00:15]
B
But Cole, obviously you've got tens of millions, if not hundreds of millions of dollars that you guys have tested and spent on SEO. But this is not just SEO with new tools, is it?
- [00:24]
A
No, it's really a new paradigm. Google itself is actually changing behind the scenes. Like for the last two decades, search has averaged 3 wor. Now with AI, people are loving this ability to navigate based on meaning. And that's the big difference. Queries are now 16 words long and voice queries are now 29 words long.
- [00:43]
B
What about the companies that are on here? They're not selling on Amazon, they're not selling E commerce. Does geo matter to them?
- [00:49]
A
Yeah, totally. AI discovery isn't just a retail problem, Right? But it's a buyer behavior problem. But with AI, if you are not early on in making your content easily and substantially understood by these large language models, you're probably going to fall behind in the same way that other brands fell behind 30 years ago.
- [01:07]
B
Cole Casperson is a leading voice in AI powered marketing and chief Data Officer at Crank Tank. He helps brands understand how AI is transforming search, E commerce and customer discovery, showing companies how to get found when buyers ask AI what to trust. Welcome to Using AI at Work. I'm your host, Chris Stagle. Each week we'll be learning how today's business owners, entrepreneurs and ambitious professionals are getting more done with smart use of tomorrow's tech. Let's get started. Right now, every business leader is asking the same question. What are we going to do about AI? If this is you, chiefaiofficer.com has the answer. We give you a simple path forward where we provide executive and team training so your people know exactly how to safely use generative AI in their day to day. We also manage the deployment and implementation to make sure tools actually get adopted and deliver results. And we'll also guide company wide transformation so AI becomes part of your operating system, not just another shiny object. The companies that act now will increase productivity, cut costs and grow faster than their competitors. Those that wait will get left behind. So if you want to make AI work in your business, visit chiefaiofficer.com and see how we're helping companies of all sizes finally get results from AI. Welcome back to another episode of Using AI at Work. My name is Chris Daigle. I'm the host of Using AI at Work and today our guest is Cole Casperson. And Cole, we're gonna be talking about how they're doing some pretty sophisticated stuff with AI and marketing. Cole's currently the chief Data officer and a partner at a company called Crank Tank. And they reached out to me saying hey, I think we've got some stuff that would be of interest to your audience. I got on the phone with them, had a great conversation before we, before today. And they're in a really cool niche. They're certainly world class marketers, primarily focused on E commerce, digital commerce, those sorts of things. But for the outdoor sports industry, mountain biking, skiing, things like that. So regardless of what your industry is though, today we're going to be talking about how an expert level, you know, marketer that's dealing with well known brands, how they're, how they're like thinking about AI and the, the you know, full stack marketing spectrum, how they are seeing the impact of it and what they, what they would suggest your company does, what your company doesn't do and just really give you professionals opinion on where you should be leveraging generative AI in your marketing. So Cole, awesome to have you on the show. Thank you so much for taking some time out of your day to, to share this stuff with our audience and say hello.
- [03:42]
A
Awesome, thanks for having me, Chris. Yeah, look forward to having a nice conversation here.
- [03:47]
B
Yeah, yeah. So let me ask you Cole, what had you guys reached out to us about being on the podcast?
- [03:54]
A
Yeah, well, we really, we liked the, you know, I've liked some of your shows. I liked your Bruce Clay episode obviously as a marketer.
- [04:02]
B
Yeah.
- [04:02]
A
When Bruce Clay talks, you listen.
- [04:05]
B
Yeah.
- [04:05]
A
And yeah. And we've just had so many of our clients and so many, you know, auxiliary conversations, even conversations with other agencies about AI. And it just feels like, you know, in some form or fashion everybody out there had an AI itch to scratch. And from the marketing perspective, what we'd really seen from our side of things is that AI is really changing online platforms in that for years the front door of the Internet has been search engines. Now what AI represents is basically multiple new front doors to the Internet. And those new front doors work quite differently. In some ways they work the same. As Bruce talked about during thing is that SEO still matters, but with these new front doors and these new layers, there's a lot of really cool opportunity for brands. Again, whether you're E commerce or any kind of brand that's looking to market or get yourself known and noticed Lots of different ways, lots of new exciting ways for them to actually engage on this new generative engine doorways.
- [05:14]
B
So because I guess for the last 25 years anybody who's wanted to kind of game the system, quote unquote has been focused on Google and just in general, search engine optimization. Now what are you calling it? Answer engine optimization, Generative engine optimization. What do you call it? Aeo, geo?
- [05:32]
A
Yeah, yeah, it's been bouncing around quite a bit. It seems to be the word the terms that I'm hearing the most, I think the target CEO used it in their earnings call is Geo. Seems to be like geo is the one that again with this market specific Geo tends to be the one that everybody's looking at it. But yeah, we've heard the terms AIO, AEO, answer engine optimization, SEO 2.0 is a nice little cachet to it. When I heard that said the first time. But really, yeah, I mean it's the fun way that we describe it is just that like you know, no more paying, praying and obeying to the googlebot is again and that stuff's still important and Google is a very still important part, a still important front door. But now there's just layers above that front door and brands are working hard and if you've worked hard in the past to find your keywords, stuff your keywords, all these other games you've played in the past, if you're a brand that's been putting in that hard work, you're going to like how easy and transparent these new layers can be when you approach it from the right angle.
- [06:41]
B
Okay. And for those of you listening, if you're not sure what we're talking about with aeo, Geo, whatever we're talking about how in the past if somebody went to Google and search for something and ideally in your industry, your website or your advertisement would pop up. The people are doing that in the large language models in their chat GPT account, in their Claude account and their Gemini account. They, they're not necessarily just. Well, a lot of people still do treat it like a Google search term. They just go in and type in a search term and get a result. But just in general as especially decision makers are using the models as part of their ideation and strategy process and they're digging in and they're finding resources. You want to pop up when they say oh well, who's the best at this? Or who could help my company with this? If it's something that your company does, you, you want to show up and that's A thing. Now it's called generative engine optimization and that's what we're talking about here. But Cole, obviously you've got, you know, a lot of like tens of millions, if not hundreds of millions of dollars that you guys have tested and spent on SEO. But this is not just SEO with new tools, is it?
- [07:49]
A
No, no, it's, it's really, is a new, it's really a new paradigm. Is, is that. And Google itself, what's really interesting is, is that Google itself is actually changing behind the scenes. For anybody else in your audience out there who understands how search works, whether it's the co occurrence modeling or the string matching. Is that for years the search, I mean like for the last two decades, search has averaged three words. You know, somebody types in men's running shoes, right? And then you start clicking blue links to do your research. Yeah, now, now with AI, people are loving this ability to navigate based on meaning. And that's the big difference is that before you were navigating based on words, now you're navigating based on meaning. So if I type in women's mountain bike shorts or feminine biking bottoms, those in the past were different keywords. And maybe your Google PMAX campaign might match up a little bit there, but those represent the same idea. And so now we're navigating, we call it concepts over keywords.
- [08:53]
B
Interesting.
- [08:54]
A
And customers. I mean, and this has made a huge change in consumer behavior. Queries are now 16 words long and voice queries, which are even more growing in popularity are now 29 words long. And they're so important in what's going on is that just, I think like seven, eight days ago. So mid, you know, mid bay here, Amazon rolled Rufus into Alexa just because that voice search is so important as they wanted the power and the semantic matching of Rufus within the Alexa aegis. So yeah, so that's the really thing is the whole idea now is that you can navigate based on meaning. And that's fantastic because you know, in the past it, you know, it could be kind of enough like we'll stick with that. Men's running shoe example is that like you type in men's running shoe and then you're starting to do your research, you know, and you're like, oh wait, but I want a zero drop shoe and well I'm running, you know, I only run for like five miles at a time and I like to run in wet weather and I, you know, instead now you can navigate based on meaning. Say hey, I'm, I'm a 40 year old guy who's training for a marathon. I, you know, I'm in the Seattle area and I prize comfort over speed and okay, boom. This is the exact, these are the exact three products for you. Then which color would you like these three products in? And yeah, and brands that can and as your goal as a brand or you know, whether you're a manufacturer. Right. And you're trying to manage a customer relationship to specifically and naturally put content that answer these questions, you know, and queue up the data so the AI can get their hands on this information to say, you know, so if you've designed, if you're a shoe brand and you have multiple running shoes, like, you know, make sure that you're talking directly about who and what this shoe was designed for. That's what we've seen in our own research actually is that you've probably heard many times that AI driven recommendations convert at a higher rate. And the single biggest factor in that that I can find is that AI is better at putting the right products in front of the right customers. The, it's almost it. The way that I describe it to some of the SEO professionals is like, it's like having the best negative keyword you've ever seen because you can keep away unqualified customers. In the past. Right. Google Search would just say, oh well, let me put our most popular shoe in front of you. Right. Because that has the most indexing.
- [11:23]
B
Yeah.
- [11:24]
A
Now AI is smart enough to say like, well, you don't want the Air Jordans. You want, you know, this one that's, you know, designed for an older foot or whatever it may be.
- [11:33]
B
Yeah. I got a couple questions. You mentioned a few things, you mentioned that, that now search and I'm assuming like, I get it, Google men's running shoes. Everybody had just had short quick query. Now it's context they're putting in the window. And that's what as a trainer, that's what we encourage people, more context. Right. Are you seeing that Google searches are now being also influenced because more people are using the large language models and now the average Google search is no longer three characters, it's now context search.
- [12:04]
A
Yeah. Yeah. A couple really interesting data points about the Google search is, is that what we've seen is, and this is larger industry stuff is, is that people are not using Google Search less, they're just using AI models more. And so they're doing a lot of their research up on, you know, so, oh, I found the perfect running shoes. It's going to be ASICS versus Brooks here right now, I'm going to go to Google, I'm just going to go, I'm going to search both of those specific shoes and then I'm going to jump on the website and you know, I'm much further down the funnel now. And the other way, the other really interesting thing is, is that, you know, over the last couple, over the last few years, Google's been rebuilding that, their foundation of search. And so their new system doesn't even look at words anymore. It actually looks at meaning as well. And so when someone searches something like that, it understands, it understands the idea of the question. In fact, a lot of AI overviews was like 6 out of 7 times. The AI overviews isn't even answering the question you typed in, it's just answering the idea that you typed in. And one of the things that we've really seen for our brands is that by optimizing their content for these further AI levels of retrieval, Google's doing the same thing behind the scene. I mean, Google search cannot work for long queries. It just does not work for these really long tail queries. And so Google is using what's called BI encoding, modeling behind the scenes to help them with search. And what we've seen is that as we solve some of the buying coding problems for small challenger brands for this AI retrieval, it's having a recursive effect and vaulting them ahead of some of the titans that they've been behind, always showing up like, hey, we're happy with third place. We've been in third place for years. As long as we're in third place because we're not the biggest ones and Adidas, we're fine. But now Challenger brands can vault ahead because Google is using these same concepts to keep search accurate, you know, and good for Google, right? I mean, yeah, it's like they, they want to, they want to stay in the game for as long as they can and so, and they want to give better search results.
- [14:08]
B
I know that at least during the super bowl there was some dust up between Claude and OpenAI related to ChatGPT is now going to be charged or be advertising. And they were kind of presenting that as a negative characteristic of OpenAI. When does that start and what is your thought on how that's going to impact like a user experience for sure, but also maybe bias in the, the output that I get from my, my, my input.
- [14:41]
A
Well, I mean, Ian, to, from Chat GPT's perspective, they have said very clearly, you know, and whether you believe them or not, but so, and what we've seen is that, yeah, the advertisements do not have any effect on the results. That has had no effect on the results. And the reason that I think that that won't happen is that what I've seen, our agency handles Amazon and Amazon has been ahead of the rest of the world by a couple years on AI based retrieval. And that's simply because Amazon only deals in nouns. Right. The only reason you're on Amazon is you're searching a product.
- [15:20]
B
Yeah.
- [15:20]
A
And so Amazon within this contained sphere has been able to really build a bunch of a nice infrastructure of machine learning based retrieval that now, now we're seeing, you know, that looks and smells a lot like what now we're seeing with AI retrieval on the open web. And what Amazon has is that Amazon has a point where it can say, hey, you know, Amazon can make money two different ways. Right. There's Amazon ads and then there's the referral fee on selling the product. And so Amazon has had always had a choice to say, well you know what, we could show the product that has the highest bid or we can show the product that's going to get a sell. And Amazon again, whether you like them or not, to their credit, they said the value proposition for us is to get the right product in front of the right consumer and not try to optimize for revenue. And given the highly competitive landscape out there among all these LLM models, I think for at least the near future and probably the medium term future, is that their value proposition is going to be in giving the most accurate answers possible. Especially when it's like the first time that I sniff that Gemini is trying to push me. Procter and Gamble is paying Gemini to tell me something I don't want. I'm just jumping over to Claude. Right, That's a good point. Yeah. So that would be. My thing is based on what I've seen in the marketplace, I don't think that, I don't think that's going to have an effect yet. But historically speaking, based on trends, you're going to see it in foods before you see it anywhere else. So as soon as you start getting recommended breakfast cereals that don't apply to you, that may, maybe that's, that's going to put your radar up.
- [16:58]
B
So let's talk about this. So for the listeners who maybe they've, they've had a budget for search for search engine optimization as well as like Google Ads and ads on other platforms, do you have any? I don't. Is the ad Platform for chat GPT. Like can I go buy ads today or is it still in beta and small group or what?
- [17:18]
A
Yeah, it's still in beta. You know, personally, our agency has, has applied for the beta and we've talked to a couple other large. Some of our larger clients have already applied for the beta as well. And again, it's going to be for just certain levels of accounts. So if you're, if you're paid or you have like a professional access like I do, you're not going to see the ads. It's more, you know, you're using the free version. So it's kind of like same thing on, you know, like on Netflix, Right. Or Amazon Prime. Sure. Some, some big, you know, the super users don't even actually see the ads yet.
- [17:51]
B
Yeah. Or YouTube or whatever, right?
- [17:54]
A
Yeah, yeah, exactly.
- [17:56]
B
Okay, so obviously you guys are going to be experimenting with ChatGPT ads. Do you, do you. Have you heard anything that would tell you that all of your years of SEO expertise may not necessarily be applicable in a paid search environment in GEO or LLM?
- [18:15]
A
No, I haven't heard anything like that. It's more just being like, again, SEO is still a major base layer of what's going on and then just kind of seeing behind the scenes on the questions that they're having and how they want you to syndicate your information. I would say what the. If you're an SEO person, I would be high, you know, or if you're a brand, right. And you're saying like, how's my SEO experience? I would say the, the big thing that's going to be driving these things, and I think the term is going to be platform AI is, is that the, all of these. If I'm chat GPT, you know, and I, and I, not only am I doing ads, but I'm hopefully doing clickless shopping. Right. At some point down the way.
- [18:58]
B
Okay, yeah.
- [18:59]
A
Is that I want this stuff all laid out on a nice platter for me. And so.
- [19:04]
B
And a brand user or as a company.
- [19:07]
A
As the company. And so in the same way that as a brand, like, you know, you're going to be, you know, you get a big benefit by having your product catalog or your services catalog match up with Google Merchant center and you're laying your products out there exactly how it wants. It's going to be the same. This exact schematics and architecture, like, you know, that is going to matter more than it ever has, you know, because it's going to be syndicating your data outwards. So architecture for syndication.
- [19:38]
B
Okay, another question I've got. I know that at least early on when, when paid advertising started being offered on Facebook and Google and stuff like that, it's kind of Wild west clicks were super cheap and the people who nailed it early like cleaned up. Do you expect that there's going to be some lessons learned from that that are applied to the, the, the paid environment in chat GPT that will like, you won't really have that opportunity as an early advertiser to kind of not game the system, but take advantage of inefficiencies in their pricing in your favor. Or do you think that it's going to be a little Wild West?
- [20:15]
A
No, I think that it's going to be a bit of a, I think there's going to be a little bit of a wild west. And the way that it's going to manifest itself is kind of the same thing that you saw a few, a few decades ago is because what happened early on is the people that, you know, again, you know, we go back in the Wayback Machine, right, and it's 30 years ago, people are asking the same question like, well, does SEO really matter? You know, does. And the people who adopted SEO early had a nice durable advantage over the other people that didn't. And the, the quote, unquote good news if you were behind was, is that at least with Google Ads you could start to buy your way out of the hole. But with AI, if you are not early on in making your, your pro, you know, making your content easily and substantially understood by these large language models, you're probably going to fall behind in the same way that, you know, the other brands fell behind 30 years ago. And, and you're going to have a harder time buying your way out of this when you know, consumers are going to have a nice competitive environment where they're like, I don't want this Gemini. You know, I don't want Gemini telling me what Procter and Gamble or whatever other large corporation is advertising. I want true and legitimate answers. So I don't know if you can buy your way out of being behind on this one. And that's why we've, you know, that's why we've designed our, Our Reach product, which is Retrieval Evaluation and Agentic Commerce Health where we can actually measure those layers of AI. Because this is a, there is a, is a much more transparent environment if you can understand how the large language models work. Because underneath the hood, you know, because Google, right with Google's Google before we
- [21:53]
B
go down that rabbit hole Let me ask you which listeners need to know this? Do you think that even at the CEO level that they should understand what reach is doing? Even if they're like, I got a marketer, I don't deal with that.
- [22:08]
A
Yes, I think so too. Because at the end of the day, what this is, is this is how your company is having a conversation with these large language models about what it, what problem you're solving and why the problem you're solving is meant, is best for this customer. And so it starts. You know, some of the smartest CEOs we've talked to, we talked to the CEO of a 10 figure brand and he was like, sweet. He's like, talk to him. Yeah, he's like, my marketing guy is going to call you. But first, do you mind having a meeting? You know, can we just show my product development team this really quick? Because I want them thinking about. And it kind of makes sense, right? We've kind of grown up in this, this, you know, world of SEO. And it's like at the end of the day, the product development team that is designing your product or your service to solve the problem, they're the ones who should be talking about it. Yeah, it's just the marketers who are taking it downstream. And so, so, yeah, so the SEO is the, you know, he's the unifying force between product development and the marketers. So. Yeah, absolutely. Yeah. And some of the best questions, yeah, some of the best questions I've had about this have come from CEOs who kind of understand that it's like the whole thing, like between product development, product launches and then the tip of the spear marketing.
- [23:27]
B
So for the listeners, like, here's my takeaway from that. Everybody in your organization, maybe they don't need to be AI pros, but they need to understand how this stuff is working. Specifically, how does what I do and as a cog in this machine impact our business through the lens of that SEO doesn't have to be their primary consideration. But if your people aren't aware of this, they're not able to make that judgment. Call and go, oh, maybe I'll do it this way because it's better. So I think that's a great tip, Cole. So Reach, tell me more about this. So what are you guys looking for with this Reach protocol that you're talking about?
- [24:03]
A
Yeah, so what we do is actually is that when AI. So when somebody goes out and asks AI a question, Right? Yeah. Important to understand that the AI has a lot of data at the inference layer. Which is basically within its training model, right? So if I go to chat GPT and I say, hey, tell me about Napoleon, right? It's just going to be like, cool. It's already read every book on Napoleon. It's going to give me a nice little rundown if I say, what's the temperature in Palo Alto today? It doesn't have that information, so it has to go out and ask the Internet what that is. And that's called retrieval. And so in case you've ever heard the term like rag systems, that's retrieval, augmented generation. And so when you're asking an LLM a question and it's saying, like, you know, hey, what's the best bike I can buy today? Right? Or what's the, you know, what's the best fishing reel for me to buy my nephew, Whatever it might be, it's doing retrieval. It's, you know, it's. It's sending out a request across the Internet and saying, all right, this, we need this information. And the way that it gathers that information and rates the information and then source the information as part of that generative answer is. Is understood mathematics. Basically, it's an understood process, and it's very transparent. And everybody kind of does it the same behind the scenes. This is basically. It's the, it's the deterministic part of the modeling. Right. And, and so basically, if you're. Right, if you're, if you're designing an LLM. Chris. Right. Is that what you want to do is you want to say, hey, I want to make everything as repeatable as possible.
- [25:28]
B
Yeah.
- [25:29]
A
And then when it comes to the final. What we're putting on the screen, that's going to be a little bit stochastic. Right. There's a little bit of flavor to that. But I want everything ahead of that to be as deterministic as possible. And so what we're able to do with our Reach program is we actually, we follow that deterministic pipeline. And so where everybody else out there is saying, we've, you know, we've had this Google, you know, we're used to this Google world where we're just measuring symptoms, right? It's like, how many times am I getting cited in AI? How, you know, how many times am I getting cited for these questions? What we can do with Reach is we actually say, this is why you're getting cited, or this is why you're not getting cited. Because if you go to a ChatGPT right now and you click on the sources, it's going to show you the sources on the right hand side and then you're going to see who's in the sources. What we're able to do is, I'm saying like I know exactly what information you know, was retrieved as from that source, how that information was scored at every single layer of these AI retrievals and how to change your content so that it scores higher at these deterministic layers to get you retrieved more. So if you're a service provider and
- [26:38]
B
it's like how guys do. Yeah, it's not a monitoring, it's not a passive action, it's an active participation from you guys to make sure that my website is optimized for considered resource for the models.
- [26:57]
A
Yeah, no, I mean, yeah, because if you think about it this way, it's like any question you ask at LLM, Chris, is there's basically 30 pages on the Internet that are going to be used to answer that question outside of the inference layer. And we're able to show you, well, these are the 30 pages and this is the order in which they are and this is how they worked. And so if you're not one of those 30 pages for the question, but your competitor is, this is their content and their content scored higher at some of these layers and there's entire industries. Go ahead.
- [27:30]
B
So it's not so much. And I say gaming, I don't want people to think that we're like doing anything gray market or anything, but like, like making an active effort to influence the outcome. Let's say when I say gaming, the. So I'm not gaming the large language model, I'm gaming the retrieval activity, correct?
- [27:51]
A
Yeah. The cool thing is like we don't even follow your brand. What we, when we are engaging with the client is we say what is the question? Right?
- [28:01]
B
Yeah.
- [28:02]
A
It's like what this product that you have here, what is it meant to solve and tell us forthrightly what that problem is meant to say. And then we build questions around the problem and then we take that question and we send that question out and we emulate exactly how they do it. And we use the same embedding models that these ChatGPT uses. We use the same retrieval, all sorts of different things that we just match to what's going on there. It's not, I mean I could actually do better retrieval than AI does, but all we do is we do exactly what they do. Right. Because they have latency parameters. They have, they got to get an answer quick. And so we follow the question and at every kink in that Retrieval pipeline. We can show you who's winning at that kink and why. And so, and we see it all the time where it's like your biggest competitor is showing up behind you in search and then ahead of you on the middle layers and then you're coming back to the top on the final layers. So it's like, all right, those middle layers represent a buy encoding model. What they are, that's a single vectorization that the LLM is doing to measure.
- [29:08]
B
So this is geo, what we're doing right now. This is geo.
- [29:12]
A
Okay, exactly. And so, yeah, this is basically generative engines. So it's like what goes on in the generative and we're showing you exactly what happens in generative. And then what we can do is we can actually test the content because it's like, no, yeah, keep your brand voice, don't just try to game it for the bots. But it's like you can walk and chew gum at the same time. Like, let's talk about it here. And all of these LLMs, they have what we call training data priors or biases. There's actually certain features that matter a lot more to the LLMs than other features. Like you may want to talk about, you know, how well your umbrella works with your ear pads. Right. But LLMs don't have any training data to talk about how that matters. Here's what matters.
- [29:52]
B
Hey, so let me ask you, Cole, like, I'm following what you're putting down for sure, but your mechanical understanding of what's happening with the engines, did a lot of that come from that MIT class?
- [30:06]
A
Yeah, yeah, I mean, it was. It actually came from Amazon as well, because MIT and Amazon have relationship going back quite a ways on some of the early, like A9 modeling and stuff like that. So, yeah, I mean, that's really. I mean, Amazon uses a multi node inference model model. But yeah, I mean, it's basically like. So let me just put it in. Is that so? Yeah, Amazon's had these machine learning driven retrievals since like the early 2010s.
- [30:33]
B
Yeah.
- [30:33]
A
And then, you know, they had a 9 and then the a 10 used some, you know, mixed in off platform signals. And then Cosmo and Rufus. And so what you were learning, if you were an aggressive Amazon seller, and of course we have a significant Amazon portfolio as our company is that like the Amazon brands that were winning got there because they understood the retrieval architecture early, or at least they were playing along nicely with that retrieval architecture. And so it's just a matter of saying, like, well, this is what brands had to think about in Amazon, where it was a very straightforward competition, but that competition is now emerging onto the open web as AI is coming there. So the brands that think about this, that to think about this, they're going to discover the traffic they thought was organic is actually retrieval mediated when you're taking these right steps. And that's one of the big things that we see is that if you're doing AI retrieval, right, and you want, you come back and you're like, hey, Cole, what is six months of this meant for us? You're going to see higher organic and you're going to see higher brand search. Because we talked earlier about how people are doing all of this research and they come back and then they search for, you know, as six or Brooks, right. And all right, they're seeing a lot more brand searches.
- [31:47]
B
Random question. What A. A lot of the clients that we work with, as at Chief AI Officer, our main company, where we, we do the thing, like we go in and train and then deploy and all that, most of them are using Chat GPT in the past few months. You know, I'm surprised. But new companies new to AI are choosing Claude. So that, you know, that's some shift in the market for sure. But for the ones that we mostly deal with ChatGPT, what is the search engine that ChatGPT is using for its retrieval?
- [32:18]
A
So what they do is, well, it depends on the kind of retrieval that's going on is that ChatGPT of course had a long standing relationship with Microsoft. So they do rent the Bing index from Microsoft and no bueno.
- [32:33]
B
Right.
- [32:33]
A
Like it's, you know, they've actually done some pretty good indexing in terms of like, just like raw. Yeah, I mean, yeah, no, it's like, it's not like it doesn't seem nearly like a volumetrics like Google Search does, but in terms of like what indexing does and how indexing cues up the data, I think Perplexity still uses the Bing index as well.
- [32:51]
B
Really?
- [32:52]
A
Yeah. And then of course, Gemini, you know, in the Google family is happily using that Claude uses Brave search and they, you know, the good thing about Brave search is that Brave tends to try to be unaffected by, you know, advertising and stuff as well. So.
- [33:09]
B
Okay.
- [33:10]
A
But you know, the behind the scenes is that like when you go into. The big difference is like when you go into like research mode, like you know, when you're Claude, right. Or ChatGPT and use research mode rather than search, that's a different. It's typically A different retrieval architecture. Because what they're doing is, as we talked earlier about how what AI is doing is, it's what we call the candidate generation list. It's like, bring me back the web pages that relate to this, and then we're going to start to analyze their content with deep research. It's like, hey, if there's a page, like, you know, if, if your collection page, right, or your, your main services page on your websites, you know, get you in the door, the deep research actually crawls your whole site and pulls data from, you know, the rest of your schema markup or, you know, the rest of, you know, your associated web pages. So. Yeah, and the other thing too, though, is that what really surprises a lot of people is that when we talk about how retrieval, you know, this retrieval and this new AI landscape affects them, is that everybody, like, already knows about the conversational chatbots. And. Yeah, and you should totally do that. And it's the fun part of AI, but there are AI shopping agents out there with various layers of handoffs and chaining that go all the way to checking out for the consumer. And there's autonomous AI browsers that'll surf the web on someone's behalf the way a person would, and then these vertical special AI tools that are embedded inside other apps and platforms. Like if you're on Home Depot, right, when you're at Home Depot this weekend, they've got a thing called Magic Apron that's an AI vertical inside of there, helping to describe these things. If you're using Kayak or Expedia, any of these other things, there's so much retrieval. I saw a stat, I couldn't source it, but I remembered it was like 60% of the time that you're being part of a retrieval As a consumer, 60% of the time that the answer is part of a retrieval request, you don't even know it. You're just engaging all the levels. Because if you think about it, if you're a service provider, if you're somebody who's trying to give a consumer an answer, this AI mediation is so useful that you're using it not to try to pull the wool over anybody's eyes. It's just, I want to give you the right thing. So, yeah, so it's not just about ChatGPT. If you just think about all of these different ways that consumers can find their way to you and all of these ways that AI exists before the consumer relationship. It's, it's a. I, I mean, I think we tried to Count the touch points once and it's over 300 different AI mediated touch points at this point out there. And they're inventing ones right now that we don't even know about. So you know, it's, yeah, it's pretty cool. And that's why we've built Reach the way that we did as a deterministic modeling is because if you had built like if you had optimized your site for Claude 4.6, right. Let's just say like, hey, my genius marketing guy has figured out how to make us just for whatever reason be the best in 4.6. When Claude came out with 4.7, there was some significant adjustments that would have messed with a lot of people's. So you can't optimize to a single model and you can't optimize to a single version. But what you can optimize for and what is the majority of what's going on behind the scenes? Or is all this determinism that's occurring behind the scenes and, and it's all open source. It's not, it's not a black box. Like you know, like PageRank is with Google or Hummingbird.
- [36:42]
B
So yeah, you know, so if, if you're listening to this, right, you're, you're thinking, oh, I just go to chat GPT, I like a drive through window. I ask for something, it gives it to me, I leave. But the reality is there is so much more happening behind the scenes. Like, like I would consider myself an expert level user of generative AI and business operations, right? But I don't necessarily like I can drive a car fast, but I don't know anything about the engine. And that's kind of how I am with generative AI. And I don't say I don't know anything about the engine, but like my question about the mit, like did you learn some of this mechanical stuff which was I find fascinating, but also all of these like, you know, I just assumed it was on site optimization and when they went out and when chat GPT went out and stole all the information from the Internet one more time that I would somehow like pop up to the top of the list just like an SEO. But what you've shared with me today is like SEO is child's play compared to what we're doing with the, the Geo effort. I mean there's, it's not just you're not playing in one field, you're playing on multiple levels in multiple fields because there still is some SEO prowess that would be expected from this. But there's also. If I didn't understand the, the retrieval and at each kink and why it's like if I didn't understand that that was even happening, like there's no way that I can, I can again game the system quote unquote. There's no way that I could have my site optimized so that it would be a result that showed up in the retrieval. This has been eye opening for me for sure. Now as a business operator, do I. Am I going to use this? No. But going back to this reference with the CEO earlier about, hey, the CEO wants us to explain this to their product team so that product can now think about it through a completely different lens. That's probably going to be more relevant in five years than searches, right? Yeah, like that's the way any listener here who's like this fascinating stuff, but I don't do SEO. I'm not going to go and optimize for geo. No, but you should understand it so that you walk into the room to have the discussions with the rest of your team more informed. Right. Like just like huge, huge gain. Thank you for this.
- [38:59]
A
Yeah, no, I mean, and the interesting thing.
- [39:01]
B
Go ahead.
- [39:02]
A
The takeaway for the CEOs is that like, yes, you're, you know, and particularly the ones in the service industries and stuff is like, yeah, your consumers are still relationship driven, like totally. But the way that your consumers or new consumers are building that short list of people that are consuming considering before the relationship starts has completely changed. Like that's what has really changed. And so they're not asking their network as much as they used to. They're asking an AI, you know, and sometimes they may not even know it. And so AI is now sitting upstream of that relationship, not replacing it. So the brands that figure out how to manage that visibility are the ones that are going to get the call.
- [39:43]
B
Interesting. Now I'd like to kind of shift gears a little bit. I'm always curious about what the, the AI journey has been like for your company for Crank Tank. And I know that when we got on for the listeners, what we always do, we never bring anybody on cold. We, we want to do a pre interview and make sure that like we. Oh, okay. I get what you guys do. Oh, that's great. Oh, that's fascinating. So that we come to the podcast ready to record with a little bit of rapport and you know, like some clarity on what we're going to talk about. About Scott, one of the co founders was also on that original Call. But he said you were the, really the, the one that led the push inside crank tank.
- [40:19]
A
Yeah.
- [40:20]
B
What were they doing? Like, what was there prior to your pushing? What was the discussion about AI like in, in the business? Because you guys have, I mean like, you're not a new company. You guys have been successful for a long time. You got a great catalog of like trophy clients and all that sort of thing. So what was that? When was that? And what was the catalyst for Scott to say, hey, who's going to lead AI internally?
- [40:41]
A
Yeah, no, I mean, because obviously AI has been out there in multiple forms and in our case, you know, so many of our clients are using Google PMAX campaigns. Right? So Google of course is using AI as part of how it syndicates those performance max campaigns. Google has their advantage. Sorry, Ameta has their advantage plus your retention marketing. If you're using a brand like Klaviyo, right. They've got AI inside of it. So we were seeing AI emerge as this big part of how a responsible agency would say, like, you know, the end of the month, it's like, did we deliver value for our customers? Did we make sure their Google Ads ran as well as possible? It was like this pressure for AI started to build and to build and we'd really seen a lot of growth, growth with our Amazon clients. Like our Amazon clients were succeeding beyond everybody else. And that was really the big signal that said like, I'm like, well, I obviously, you know, Rufus, right, Is like, I have a copy of Rufus's patent documents in my office actually. It's like we know exactly how Rufus thinks and how. And so we're optimizing for, you know, we have a running joke where it's like, if Rufus wants to know if this product is bigger than a cantaloupe, like, I don't care, you know, if it's $5,000 product. I'm going to tell Rufus exactly what Rufus wants to hear. And so we started to triangulate these things and say like, okay, well these Amazon, the way Amazon's working and the way that Amazon has grown to 41% of the U.S. e commerce market is, you know, is something that works. And this is now moving to the open web. So how do we, how do we help take the same success we've seen on Amazon and ensure that our webs, you know, that our web clients are seeing the same success and that we're, that we can know that we're doing is the best possible job for them on their Google P Max Meta Advantage plus You know, all of these different AI touchpoints that are now that drive their classic programs. And so between all of things it's like, well, you know, what we can actually do is we can actually see inside the beast. And yeah, and that's been the one thing is, I mean a lot of, you know, a lot of people, I mean Google's been really good in a lot of things but I think a lot of CEOs are a bit frustrated after years of saying like we keep changing our site or they keep saying to spend more. And it's like, you know, what with this generative engine stuff you can, you know, one of the things we've seen, right, AI overviews good in a lot of ways, but it's really driven down organic clicks. Right Is that as AI had been growing and growing it was taking. Making organic smaller and smaller. But once, but now AI is available for brands to grab from themselves to put organic as a bigger part of their, of their platform. So yeah, and organic is, it's not free. Right. But it doesn't show up on your P L. So yeah, which you know, was good. So that was really the, it was just. I think the same reason that so many other people are asking AI questions is it's just coming from every single direction.
- [43:37]
B
Yeah, no doubt. And like especially now. So for the listener, I want to clarify we the way that I explain quote unquote AI to non technical business professionals as I tell them there's kind of two paths. There's analytical AI which is more like machine learning and data science. And that's what we're talking about here. For most of you, you know, if you're not familiar with that side of it, the other side of it is what we call applied or operational AI. How do I use generative AI and the operations of my business? And that's what most of you are probably where you're at. I'm sure some of you are. Maybe you've done the mit, you know, AI program and that sort of thing. Maybe you've done some stuff like that. But so if you were a little confused about well, how is he using chat GPT for that? It's what he's talking about is more like the, the machine learning, the, the sophisticated, the stuff that's not for the layman like, like me and some of our other listeners.
- [44:30]
A
So the good news Chris is, is that oh, one of the, one of the really good things is that yeah like this, the stuff that chief AI officer is really good at doing, helping to basically take Your existing infrastructure. Right. And then start to harness that with the power of agents to run your agency better. Yeah. Is if you just imagine it's like if you take your website and your services offering or the way that you're describing how your warehouse works, right. And you're cleaning that all up and you're doing that optimized for the open web technical, you know, that we talk about. If that's, you know, if, if it's, if it's semantically optimized for that end of things, that same semantic optimization is going to be fantastic for all of the stuff that the chief AI officer does. Right. Is that your, the way that your product speaks into the rest of what's going on here. The same way that you syndicate inventory signals your chat bot. Right. If you know chief A officer is like, hey, let's help you build a better chatbot, that chatbot is going to run smoother when your products are designed from the bottom up with this semantic matching in mind so all those downstream effects work better. So it is, I guess, definitely layered.
- [45:39]
B
Yeah. So at this point I would, based on the conversation, you guys are offering geo services for your clients, correct?
- [45:47]
A
Yes. Yeah. And part of it is, is that, yeah, we get, we have different packages where you can do as many queries as you would, you know, as many queries as you, as you want to pay for. We do not sell it as SaaS. It is not just a dashboard where you're like, okay, got to help you. Yeah, good luck everybody. Every member of my Reach team is certified in at least one major online platform. So they understand when we say, hey, let's change your content in this manner or let's address this semantic depth with this. They understand the trade offs of saying like, okay, we're going to put an FAQ on your page, we're going to do a product comparison on each and every one of your collection pages. We understand what that means from a depth perspective and how to do that and make it look good for humans.
- [46:32]
B
So what about the companies that are on here? They're not selling on Amazon, they're not selling E Commerce, quote, unquote. Does GEO matter to them?
- [46:42]
A
Yeah, totally. I mean it matters in a couple of different ways. Is, is that on the one hand, we talked about those customer, the relationship management. Right. I mean AI discovery isn't just a retail problem. Right. It becomes really easy to solve retail problems, but it's a buyer behavior problem. AI retrieval is that and buyers in every category. You know, whether you're buying an E bike or you're just buying, you know, steel, you know, steel ingots or a 3 PL contract, a managed services agreement, they're all converging on the same behavior, right? And so, and so, and they're asking AI before they ask the vendors. And so it's just basically those kind of customers, you're probably going to tend to exist a little bit further up the funnel. But you know, again, that, that's all moved uphill upstream of the relationship. So again, whatever service you're offering, right, if you're putting that service out on the Internet and you want attention, this is how you can maximize the attention that you're getting under this new protocol where people are wanting to navigate based on meaning rather than just off of a three word sentence.
- [47:50]
B
Now, fascinating stuff.
- [47:51]
A
And then the other one too is that is, you know, the marketing teams, if you're a CEO, right, whatever you're doing, is that your marketing, a lot of your brand story gets told by other people, right, is that retailers or journalists, customers on forums are doing that are also giving signals about your brand story. And so when a ChatGPT or Plexity or AI answer a question, those systems are literally saying, what source is it pulled from? And so marketers forever have been having that question of saying, like, who's actually telling my story right now? And that's why we don't put any intent filters on Reach. What Reach does is again, we're just asking the question and we're exposing the entire pipeline so we can show exactly what who is telling your story. So if you are a, so if you are a, you know, like a logistics company CEO listening to this right now and you're saying, how come we're not getting these kind of contracts, we're not getting these kind of calls, you know, following the question, saying like, hey, I have, you know, 25,000 packages a month and I'm looking for it to ship out of the mid, you know, Midwest, you know, we can, we can show and say like these are the people that are telling the story because at the end of the day there's 30, you know, again, this is very simplistic, but there's 30 pages on the Internet that are contributing to the generative answer that has been put in front of your customers. Are you one of them? Are the ones that are on there talking about you? If so, what are they saying? If not, why? And then we come in and say, well, this is how you do it. And then you can imagine that this level of insight that we're able to have has a tremendously positive impact on how we can run your meta campaigns or your Google campaigns or your retention marketing. So, you know, so some brands come in and they're like, dude, this is awesome. This is great. You know, we did, we want the whole package, but let's just focus on one query at a time. Let's focus on this product launch that we're doing and then take everything you're doing and help make our Google Ads better. And that's one thing that we really like to do is we will scale it around the needs that you have. Because, you know, navigating based on meaning isn't just easier for consumers. It's easier for an agency that is, you know, has the whole picture in mind.
- [49:55]
B
Sure. No, I love it now. I've learned a lot today, Cole. So for the individuals who are interested in either following what you're up to, because again, we've had, you know, we've had Bruce, we've had people like that on the show, great episodes. But I think I've learned more about the the how today than I have on on previous. So if other individuals that want to dive in with, with either your approach or to find out more about what Crank Tank's up to, best places for them to go and either follow you as a thought leader or, you know, see, like use cases.
- [50:29]
A
Yeah. Come come visit reach.cranktank.net and you can sign up for your own free query. So if you're a business owner out there, a CEO or a marketer, come check out reach.cranktank.net, go down to the bottom of the page, put out your information, tell us the question that you want. Like, what is the question? We've been talking this whole episode, Chris.
- [50:50]
B
Right.
- [50:50]
A
And I'm sure these, your the people who've been listening have been like, oh yeah, what is the single most important question that matters to this product or service that I'm thinking of now? Give us that question. We'll give you a free demo and show you, show you what it looks like.
- [51:03]
B
Very cool.
- [51:03]
A
And some brands are pleasantly surprised. Others are like horror stricken. So, yeah, it'll be. But no matter what's happened, you know, at any stage of the thing and any size of the company, the one unifying thing is that people like, wow, this was, this was totally worth my time. Thanks for showing me what it looked like, you know, and then from there it's like we talk about next steps and yeah, it's just fun to end the rewards. I Mean companies. The thing I would really like to stress that companies that want to talk conversationally and authentically about their products, a, those companies are fun to help. So, yeah, so please reach out. And then two is like, those get rewarded by LLMs because LLMs have been trained on human knowledge, and so authenticity gets rewarded. So that same hard work that you were doing. Yeah, exactly. So that same hard work you were doing on SEO, now you get to kind of have more fun with it and talk authentically about your product.
- [51:58]
B
So. And if I'm listening to this and I want to send somebody to go and just out of curiosity, like, what are we showing up as in the search term? What role in my company should I say, hey, take this link, go sign up for this and tell me, tell me what you find out. Who is that role?
- [52:14]
A
I would say if the ones that get the best out of that is the head of marketing, because the head of marketing sees the whole picture. And one thing that gets pulled up a lot is you can imagine so many things. It's like, what's the best fertilizer for my lawn? Right? Is that you're gonna get a whole bunch of home and garden articles, New York Times articles. Like, so, you know, they're gonna. They see those people see the whole picture. Like, oh, my gosh, this is a great PR target list. Like, this is the tool that helps me justify the PR budget that I wanted. In addition to me giving me the leverage to. To wrangle the dev team. Right? Is like, that's the. That's the other thing is, is that what Reach does? And as a tech guy, you know, I did build it from a tech perspective. I'm more tech than marketer. At some point, at some times is like, it's a wonderful unifying process. If your marketing team and your tech team, you know, are, you know, it's just like the dev guys love it because they. They get tasked to do so many changes on the website and they don't know if it's going to make a difference. All of a sudden, they understand this stuff and they know they're making a difference. So, yeah, if you're whoever's high enough on the marketing chain to be able to tell the dev guys what to do, that's. That seems to be the magic junction point.
- [53:33]
B
Awesome. Hey, so we'll actually have the link to reach.cranktank.net it'll be in the show Notes. Pass this on to your marketing team. But Cole, thanks again, and I'm glad you guys reached out this has been a fantastic episode. And for those listening, look, if you're getting something out of the podcast, help a brother out, like, let other people know that you're finding value in the conversations, you're learning things and you're better understanding how to think about using AI at work, right? For your team, for your, for you personally and that sort of thing. We'd love it if you just help us get the word out. Leave a review, forward a link. So with that, everybody, we will be back next week with actually, we've got an amazing episode coming up for you next week. Another one. I'm looking forward to bringing that to you. But again, Cole, thank you so much. Cole from cranktank.net was our guest today and everybody will go out and use AI.
- [54:24]
A
Awesome. Thanks for having me, Chris. It was a great conversation.
- [54:27]
B
Thanks, Cole. Thanks for tuning in to Using AI at Work. Don't forget to subscribe for more conversations about how to use AI at work. And a special thank you to our sponsor, Chief AI Officer for Empowering Businesses with AI Education and Training. Visit their website for a free AI Readiness Assessment and AI Strategy Guide to help you get started using AI at Work. That's www.chiefai officer.com. follow us on Twitter at the handle Using AI at Work and visit www.usingaiatwork.com for free resources to help you harness AI in your role.