Stop claiming you're the best! AI will mark you down for it | Tom Rudnai on winning influence in the age of Dark AI

Speaker A: Hello, welcome to the Trust and influence of B2B podcast. I'm your host, Joel Harrison. Now here's a counterintuitive thought to open with. The more confidently your brand insists that it's the best, the more likely AI is to quietly leave you out of the conversation. Decides who actually wins. This runs against almost everything we've been trained to do. Strong positioning means claiming the category, asserting superiority, being bold and ambiguous. When a buyer turns to an AI assistant early on still working, what their problem even is. The AI isn't looking for a winner. It's weighing options, comparing approaches, quietly building the criteria decision that later will be judged against content that simply declares itself to be the best, gives very little to reason with, and by the time the AI is raised to name a recommendation, the field is already narrowed. You're either on the shortlist or you're not. So the assertive, uh, sales led messaging that performs at the decision stage may be winning visibility at the precise moment the outcome is already settled, while costing a brand influencer at the earlier stages where it's genuinely up for grabs. The implications of how B2B brands build authority, credibility and trust are significant, and they're not always comfortable. My guess Quest has the data behind this. Tom Rudeney is behind book Demand Genius, an AI search intelligence platform for B2B brands and the author of the Dark AI study, which analyzed hundreds of prompt clusters around more than a dozen B2B categories to understand when, why, and how consistent AI services around few people have looked as closely as Tom has at where influencers genuinely won and lost an AI led buyer journey. So in this conversation, today's episode, you're going to learn why bold cattle claiming positions can actually quietly work against you in AI search, what dark AI actually is, and why the moment AI names you so for the M moment, it stops mattering. How AI narrows its options, convergence, and where the real window for influence actually sits, and what to build and measure instead to earn genuine influence and trust early in the journey. So don't forget, if you find this conversation valuable interesting, please take a moment to subscribe or to like or to follow, depending where you get your podcast from, whether you're watching your audio visually or whatever. Um, and if you can give us a rating or review, it really helps other B2B marketers find the show and helps me understand what people are liking and helps me build an audience to create better content in the future. So that's the pitch with Tom. Welcome. Great to see you today.

Speaker B: Thank you, Joel. Nice to be here now one thing

Speaker A: we haven't really prepped for. You and I know each other quite well, but give us a bit of board launching. The questions give us a bit about your background, uh, and where you've come from. You're in the media industry, is that right? Yes.

Speaker B: Or working in. Yeah, in the technology, but kind of focused on the media sector. So yeah, for folks who don't know, Joel was actually once on our podcast. So this is a nice little role reversal and he gets to ask the questions and put me on the spot this time. But yeah, so my background, I spent the last 10 years or so in various different technology AI companies across kind of operations, marketing, sales, um, I say my most kind of uh, formative point in that journey was as an enterprise sales rep for a company called Zephyr, um, which worked in the media industry helping large publishers, BBC, Forbes, people like that, um, monetize their content through dynamic paywalls. So paywall, figuring out when to cut each person off at just the right time based on data that they owe to cause you maximum annoyance and milk every last subscription dollar out of their, their content. But you can see how that kind of brings me into the world that we're in now around B2B content where we're not trying to necessarily monetize it by a subscription, but the job to be done is basically the same. How do we create a value around content and information that uh, allows us to drive a commercial outcome? A commercial outcome is just typically a lead or engagement rather than a, a subscription.

Speaker A: Fantastic. And you launched Demand Genius for how long the company been around for two

Speaker B: and a half years now, which is crazy. I remember talking to you relatively early in that journey and we went for a walk near London Bridge. But um, yeah, two and a half years.

Speaker A: Excellent. And so, and so it's one of those exciting kind of scale up companies in that, in that you know, burgeoning space around media and using AI. So very exciting organization to talk to about how you're developing because it changes so rapidly and it's very dynamic times. So let's dig into the questions. Let's start at the top then. So, uh, so tell us, how is the way that buyers use AI different from how they use traditional search? And why does that difference matter so much to B2B brands? I think it's something we think we know, but it's worth really unpicking what that actually looks like.

Speaker B: And I'm going to look at it from both angles actually. There's how does buyers use it and Then there's also just how does the AI behave, which is really different. So, first of all, how do buyers use it? Well, we know they trust it an awful lot, right? So, uh, depending on those different studies I've seen from AHREFS and a lot of other people, depending on what study you look at, anywhere from 5 to 23x is the conversion rate of a lead that comes from AI versus one that comes from regular search. And for anyone who spent any time talking to Claude or ChatGPT or whatever your LLM of choice is, you will know why. At this point, I turn to it as my therapist, so I'm going to trust it with what CRM to use. Um, so that I think that's the first thing. It is something that buyers use very, very regularly, and they don't use it at just a specific moment in their buyer journey. Right. When you use search, you use search to find someone who can give you the information you need, not for the information itself. Um, and that is a big shift in how buyers use it as well, because it means that you don't just talk to it at that one moment to find which vendor to talk to or to build a shortlist. You use it to frame your problem, to build your requirements, then to define the shortlist, then to do due diligence on that. So as a brand, the way that you have to communicate with your buyers is much more difficult because you've got this kind of wall or this middleman or middle person that's been put up in between. And we know the last study I saw was a Forrester study, I believe, which is 89% of B2B users now use Gen AI during their buying process. And I would bet that is only going to go up if it's not already completely out of date. Um, so that's one thing from, like the buyer behavior thing, right? The way that I behave with AI is very, very different to the way that I behave with search, but there's also the way that it behaves. And the most simple thing that gets lost in this conversation is AI simply does not search or very, very rarely searches. So we did this big study, which is what we're going to talk a lot more about today, called Dark AI. And what we were doing was running a whole load of buyer prompts across awareness, consideration and conversion stages because we wanted to unpick. Okay. In more complicated journeys that are just kind of transactional. Give me a recommendation, I'll buy it. Um, how do the responses vary? We noticed some really interesting things. But one of the most interesting thing was guess. We'll do it as a guessing game. Guess what the retrieval rate was. So the amount of times in awareness stage, the AI invokes retrieval and goes looking for an answer.

Speaker A: Not much. Yeah, that's zero, right?

Speaker B: Zero. Not once. Um, same in consideration. I'm sorry, I won't keep putting you on the spot. Same in consideration stage 0%. And then in conversion stage, 48%. So only at the very, very end of a journey. Um, now this is obviously very relevant for B2B, because we know that the more complex and messy the category, the less of the job is one in the conversion stage. Right. And if you're buying a, uh, the way I always think of it is if you're buying toothpaste, there's very little awareness and consideration. Right. You assume the most grown adults know they should be brushing their teeth. And so really the. The only awareness is, I've run out of toothpaste. As you get more and more complicated and you get up to a billing System, you've got 13 different people going through these messy processes of figuring out how the hell you take payments these days and all the stuff that finance are going to need, there's so much awareness and consideration that goes into it. And during that time, it is not searching. I'm going to go a step further, and I know I'm going off on a bit of a rant here, but I want to hammer this home. M. Even when it does search, I think it's really important for people to understand what AI does under the hood. When it does invoke retrieval, it doesn't search for the original prompt. So you might, as a brand, think, okay, we want to optimize for a prompt, which is, what CRM M should I buy? Uh, or I'm a midsize accounting firm M, what CRM should I buy? Um, and so you would think, okay, let's create content that kind of hits that keyword, because that's AEO is it's only one left away from SEO. So let's just do that. Um, but what AI will search for, if it's trying, if it does invoke retrieval, is not that it does something called query fan out, where it decomposes that term and it thinks about all of the different things that it needs to know in order to compile an answer to it. So it will break that original term down into about 20 different search terms, search for the answers for those, because it knows it needs that knowledge, and then it will Compile that into its eventual answer. So, again, hugely different behavior to search. And I think one of the things I always want people to come away from any unfortunate interaction they have with me is just the AI. It's not a search channel.

Speaker A: The consequent point about that is you've answered onto the next question I had, which is great, is don't think of it as the same. It's different. Something different going on here. We think it's just better. The same, but better. It's not. It's very different. And the. Consequently, you have to behave differently in order to get the best out of that as an advertiser and as a seller. Right?

Speaker B: Yeah. The buyer behavior is different. And the way that it functions under the hood just is very, very little of it is reminiscent of what you would, what anyone would call search. And so the way that we need to optimize for it, that's even before. One other thing I always think is fun to talk about on that is, um, there's a huge difference between, like, when you prompt and a keyword. So the keyword that you enter into search is so different to the prompt that you would enter into Claude. Right. A keyword, it forces a user, uh, to consolidate. They puts the burden on the user to consolidate their intent into three, five words. And there's data out there that I couldn't quote now. But as search has become more sophisticated, we've seen keywords lengthen, but still, it puts that burden on the user. Uh, the way I prompt, I love. I have whisper flow. I dictate to it now. And my God, is it rambling. As people probably listening to this are starting to build up an empathy for Claude. If I did this on a podcast, imagine how much I ramble to Claude. But it's long and it's filled with criteria. Um, and all of that criteria goes into the answer, which means that the answers are so much more bespoke and specific. So the idea of optimizing for all of the different keywords, I did the keyword maths, like all the napkin maths. When you add in the fact that every prompt is going to have unique criteria, and it's going to talk about your tech stack and what you need and the type of organization you are, and it's going to carry context on the kind of person you are, one keyword that is best CRM solution can kind of extrapolate out to 22 and a half thousand different variations of ways that people can say that. So, again, you can't optimize for just that entry.

Speaker A: Wow. So it's opening a box, isn't it? Right. There's so much inside to unpack m in terms of the complexity of this. So it's fascinating stuff. So we're talking. Earlier we teased this piece of research you did, which is kind of like what you're going to talk through in more detail there, which is the dark AI. Which is a dark AI Sounds very seductive and sexy, doesn't it? Or it's scary at the same time. It's a great title. Um, I think you've kind of alluded to this slightly previous answer, but one of the central findings of the research is that there's a difference between AI visibility and influence. Um, can you unpack that? What it means and why it matters?

Speaker B: Yeah. And we, we use this analogy to help me understand it that I think is useful for other people as well. Of the iceberg. Right. And it um, kind of, we're used to thinking of anything in B2B marketing as a funnel. Um, I kind of think the iceberg is it flips that on the head. But the reason is that the visibility or the um, results from AI come from the complete opposite end. So the way that in an SEO world the funnel works is the broader, the more top of funnel, the more traffic it is, the more people there are. And that's actually where you get your traffic. So you get traffic at the top of the funnel. AI is different because in our study, what we noticed when we went through that study that I was Talking about earlier, 0% retrieval rating awareness, 0% retrieval rate in consideration, 0% conversion. The same is true of citations. So what most people in the AI world treat as the kind of holy grail you actually only get in this very, very narrow slither of prompts, which is the conversion decision oriented prompts. Um, so that's what we call the tip of the iceberg. That's what's going to show up in your current metrics if you're using an AI tracking tool that probably counts citations or you're looking at your Google Analytics. Because in order to click through to your site you have to be cited and there has to be a link there in the first place. And that's very, very rare. So only in 16% of the prompts that we tracked did that kind of behavior exist and did that outcome exist that might send traffic in your direction. The other 84%, no brand is getting explicitly cited. But they're the messy conversations that are deciding how the problem gets framed. What I prioritize what requirements end up on an rfp. And that's the analogy that I like to use. Um, if you think anyone who's ever run an RFP process, there's a kind of common refrain that if you didn't help to write, help the customer write the rfp, then you've already lost it because there was a salesperson who was in there feeding them the criteria and helping them to write it. And that's inevitably going to be the one that ends up winning it because the questions are all shaped in their favor. Um, that's dark. AI really is all of those little conversations where you don't get mentioned the 84% that ultimately decide whether the category is who wins the category.

Speaker A: Well, fascinating stuff. So another thing we talked about, so you've got this layer, this iceberg concept, another concept introduces the idea of convergence and way of measuring it in plain terms. How does AI actually arrive at recommending one brand over another?

Speaker B: Yeah, and uh, I think I always need to caveat this slightly with I don't have a perfect answer to this and I don't think anyone, anyone does Wellington, that there is still an element that is a black box. Um, there was a trend that we noticed very clearly as we looked at the data see, which was as you go down the funnel, the language changes in a really interesting way. Um, so at the top of the funnel in those awareness kind of responses, the language is very, very exploratory. So the AI is in exploration mode, is kind of talking about trade offs and comparing things and there's also very, very little consistency in terms of the browns brands that it's citing and it's doing it all from memory as we've established. It doesn't invoke retrieval. Um, as you go down the funnel the language evolves so it gets very, very comparative. Um, in the middle of the funnel there's consideration ones. It starts kind of comparing in a very structured way all of the different solutions and then once you get down to conversion stage it becomes very, very directive and it comes very, very risk averse. Actually the variability between runs in terms of the brands that get surfaced goes next to zero. Um, which basically kind of. Yeah, that's where we get this trend of it comes much firmer and much more specific and clear in its recommendation and that's how it kind of over the course of a conversation arrives somewhere. Now again, there's a couple of implications there. It means that actually at the bottom of the funnel on those prompts you have very little opportunity to influence it. So in terms of once the problem is framed, it knows who it's going to recommend. There's not a lot of chance to go and create a piece of COVID listicle and change that because it doesn't vary much. Um, and the other one is what you touched on in the introduction to this, which is it just has a bit of a bearing on the language that we use. If we know that a lot of the where the opportunity lives is more top of funnel, that's where there's more of an opportunity for influence and that's where it's in exploratory mode. Well, there's a big contrast there between the language that most of us use in all of our marketing, which is extremely directive, it's extremely bold, as we're the best CRM for anyone with 10 quid for me. Um, and that's where I think there's also a little bit of a shift that has to come to us being a little bit more realistic and a little bit more specific. We're the best for this specific scenario and we're not in this scenario. And that exploration of trade offs helps give the AI a framework to understand how to recommend you to the right

Speaker A: people, which, which is, as you say, it's completely counterintuitive to what, you know, we've all believed and brought up to think is the right thing to do. Right. Um, and I think what you've done also, if you just answer that kind of the next question was around consistency. So if you, you might think that being consistently, consistently shown up in answers is great, but actually you're saying, no, that doesn't really help because it's actually, you know, that's, that's not, you've just explained how that process works at the bottom of the funnel. So, yeah, um, all of our, all of our expectations are being kind of upended in all of this, I think a little bit.

Speaker B: And I think that's the challenge is those expectations were set by, and I try to, I think I sometimes go a bit too far with this actually. But AEO as a problem was adopted by the SEO team, the SEO thought leaders, the SEO vendors, because it makes sense. Initially we thought it was a new channel. Um, it's just as when you actually unpick the way it works, it works in a totally different way. And I'm a big believer that if you're trying to solve a problem, you need to engage with it without any kind of preconceptions. And the reason we started off down this path actually Was we just had this massive sense that it's unbelievably convenient, that, uh, all of the advice looks just like SEO advice, despite the fact that it's a completely different technology that behaves in completely different ways and has completely different user behaviors, like, what are the chances? And that's what kind of set us off down this path. But that's, I think, where a lot of the missed expectations come from.

Speaker A: Is that cynicism or naivety?

Speaker B: I would argue naivety. Um, and that's where I think I sometimes have gone actually too far in making this point. There are definitely some snake oil salesmen out there, and there's some who, if it's not cynical, they certainly don't care. But in the majority of cases, no, I think it's just you have a framework of thinking about things, and it makes a lot of sense to go attack a new problem through your kind of existing framework of rankings and thinking about that. But that's one that's, that's the best example is rankings is I, I think, a ridiculous concept to ever use associated to AI because the answers are really verbose. And I've looked at the number of prompt responses I've seen where the response, like, I'm going to simplify it drastically, but if the response is, you know, don't use Salesforce, use HubSpot, it's like, well, congratulations, Salesforce, you ranked number one. Um, but what a lot of these. This advice leads you to is that. So it drops all of the context out of a response, or just all of the nuance out of a response, uh, or a channel where the input and the output are both so much more nuanced. Um, so, yeah, I would argue naivety and kind of unconscious bias rather than cynicism and evilness.

Speaker A: Okay. It's nice to know that there are some instances where cynicism and evilness aren't the root cause of what's going on in B2B marketing. But, you know, that's.

Speaker B: Maybe this is just my only naivety here.

Speaker A: Let's talk it on the naive bandwagon. I like that. It's good. That's good for me. Um, so another recurring thing that we're seeing in your report is a concept of information gain. Can you tell us what that is? And whether a marketer, uh, how would a marketer know whether the content actually has it?

Speaker B: Yeah, so, uh, both difficult questions. So information gain is the idea that basically in order to be useful in today's landscape, or in fact, I'm going to Take a step back. So let's think about in the old, old world, um, or actually I shouldn't even call it the old world because SEO is still perfectly relevant. But in that world, um, what made content good? The idea. Well, the strategy in its most kind of basic form was you find a keyword that is either high volume or high intent and you summarize or synthesize knowledge against that as best as possible. And the best summary wins. Right? That's what Google is, is a very complicated algorithm to think to find what the best summary is to that question or that intent. And that worked because Google was a directory that needed that summary. It needed to link off to that or to ingest it into the signal. AI reviews for just a minute and keep things a bit more kind of foundational. Um, AI is very different. It doesn't need that. It has the summary, it knows it. So it has no, no reason to go looking for your content, to cite your content. And actually the human on the other end of it has no reason to use your content if it doesn't teach it something new. So that's what we refer to as information gain, and I can't claim that we coined this term, but local attribute, who actually did. Um, it's basically does your content produce net new knowledge, net new information that AI doesn't know elsewhere and that the humans can't get elsewhere as well. And if it doesn't do that, it doesn't have value. I think this was in the realm of theory. Google's core update in March actually made it pretty explicit, like there was a 22% rankings boost even in search and AI overviews, um, for content that contained original research, which often is one of the best indicators of information gain. We think of it in two, two ways. Right? You can produce information gain through a really clearly defined perspective. Um, so like an opinion, do you say something? Right. And, and we have this kind of tiered framework and level one is what we call interpretive gain. So a new slant on an existing piece of information that has value. Right. It can add something new to an industry. You then have level two and three, which is empirical gain and conceptual gain. So empirical gain is new research data that either backs up existing knowledge in a new way or adds kind of incrementally onto it. And then U3 is where, with some foundation in that data, you have a genuinely like new mental model or framework. That's what I would like to think that dark AI is a good example of one that we were able to produce, obviously you can't always come up with a groundbreaking. We're lucky. In our category you can because it's like two days old. But that's how we think of that. And that's what I would argue should be the absolute number one north star of any content team or really marketing program. If you can't back up who you are, who you're for with some level of new information, then it's going to be difficult.

Speaker A: Well, it seems bizarre that we even have to make that point, but I think evidence suggests that we do. So. Um, so yeah, if not before, then now is the time we need to be saying something different, saying something new. Uh, because the game's changed. Um, and so just. I see back to the kind of central point of the top rich is that we're trained to think about things that are bold and definitive. Um, to claim a category. Um, how does that kind of confident positioning actually fare in AI led by a journey? So it's not clear.

Speaker B: Exactly. I think there's two things that I can, that I think are useful to understand here. One is what I kind of touched on earlier. Uh, and this is entirely theoretical. I can't prove this with data, but it stands to reason, right, where if your goal is to get snippets taken or is to get your content ingested by AI to answer a particular question, obviously biased information is much less likely to do that. So if you do a listicle and you're number one for everything to everyone, um, or your homepage is kind of the all in one solution for anyone who needs a CRM, then you're not giving it anything. And AI is smarter than that. And we see it's kind of search for legitimacy in how often it goes looking for reputational signals actually. So index is a lot more heavily on that than search did. So that's one thing. Um, and also just the language that you use actually. Again, if we know that it's exploratory mode at the top of the funnel, we'll give it content that helps it to explore, not and to understand and to compare trade offs. And that requires an element of kind of self evaluation and self awareness, um, which is often difficult to communicate to the C suite. I know. And that might be where a lot of people are shouting at their phones right now. Um, so that's one thing I think the other thing that is just in terms of one thing we know authority is incredibly important. Exactly what I said earlier, it becomes quite risk averse in terms of the recommendations that it gives. But what it does do is it takes on criteria. So it's very, very hard. I'm going to stick with the CRM category because I think it's very well known. It's very, very hard to become the best CRM over HubSpot Salesforce. But what we do know is that, uh, as it takes on criteria, I can converge in lots of different directions. We will start to understand I'm actually looking for a CRM for a small startup in this specific industry with these specialist needs. And so if you can position yourself as that, then that's where you do see people get faster results. Um, like you can be the best CRM in Milton Keynes for accountants. Um, and if, and if you're specific on that, then that does seem to allow you to. That does seem to bear fruit. Um, the way that I think of it, if anyone listening to this has ever read Crossing the Chasm, which I know in the SaaS and tech world is like a bit of a bible, but that's like the strategy of how to build a business. The idea all come. I'm going to go history note. It all comes from, um, landing on Normandy, right? So the Allies need to invade France. You have this big beach with a lot of troops on it. How do you do that? Well, you don't invade the whole beach. You concentrate. They have to defend the whole beach. You don't have to attack the whole beach. You concentrate your forces in one small area. You get a foothold and you expand out like that. And that's Crossing the Chasm is a book basically applying that to technology, which is you choose your very specific ICP and then you strategically expand out from that. I think you have to take a very similar approach actually to search now, which is focus in on, okay, this is who we are and this is who we're for. And then get what you can from that and kind of strategically expand that out as your TAM needs to expand in order to hit whatever goals you have.

Speaker A: Very interesting. Very, very interesting. I love the historical analogy as well. That's always what works for me, certainly. So, um, okay, so let's think a bit about measurement and trust and what to do next. And this show is really about trust and influence, as the title suggests. And when the aim becomes shaping how an AI frames a problem so your solution looks inevitable. Where's the line between earning genuine trust and simply gaming the machine?

Speaker B: Oh, that's a fantastic question. Um, I would say the line is, are you producing information gain? Um, one thing we know, volume is important and Volume helps to signal authority in depth, but volume can be a problem for AI because it becomes very hard to maintain quality, maintain consistency, clarity, the bigger your library is. Um, but equally you do kind of have to be everywhere these days. In our marketing strategy, that's just what has happened. Um, so I would say that where the line is, whether you're maintaining a high level of information gain. So if you're doing fantastic original research, you're really clear on who you are and the kind of unique perspective that shapes you, you can use AI to scale content across lots of different channels and you can actually use it to reproduce that information gain. If you have it sat there in a database somewhere, you can turn that into content across lots of channels and scale it. Um, so that to me is kind of the acceptable level of this. And that's just the world that we live in.

Speaker A: We would teach the lines very much the project that you and I worked on together, which is to map agency store leadership and we, we did a, we'll put a link to this in the, in the show notes, but we just done a project looking at B2B agencies and how, and they used to thought leadership and actually having uh, kind of thematic based content on a consistent theme and then done. Well, it's about, you know, with, based on something new to say rather than just spieling out thousands of articles that are random on a different topics. You know that, that's so, so that aligns very strongly. That's, that's, that's good to see that, that, that um, theme being picked up here as well. So I think one of the things you've mentioned before is that citations probably in the old world were a good thing to do, but these days are uh, the wrong North Star to be looking at. So what do you think that was something that people used to measure. What should B2B marketers be measuring instead

Speaker B: in terms of AI? I think at this stage, I think, um, unfortunately you just have to wrap your head around the fact there is no clear North Star that we have kind of settled on and understood well enough to know that that is what you should point out. Um, but we do have some ideas for additional indicators that you should be looking at. So citations I would say is one of them. Right. It is valuable if you get a citation that's a link that can be a source of traffic and direct engagement. Great. But you have to accept that's a metric for the tip of the iceberg, which is one part of the job. So I would absolutely be looking at that we do that within our own product. The other big thing that I propose you've talked about here is just from a content perspective, information gain. If you're providing that consistently, it is very kind of predictive of whether you're going to be serving up in AI responses. And I encourage people particularly in B2B to focus a lot more on a slightly woolly metric around fit. So does AI view you like what Does AI genuinely understand the strengths, weaknesses and trade offs of your brand to be and how does that fit the requirements of different groups of buyers? And you can quantify that and visually map that, right? So you can say okay, what are the key buying criteria in our space? Maybe it's ease of integration, quality of support, pricing and something else. Um, you can talk to AI, understand how good it thinks you are, all of those things and then map that to what your specific ICP and buyers need. And then you do get a kind. You can't have a single metric about that that seems to be much more predictive rather than tracking any one prompt and saying do we show up? How does it recommend us? That's much more predictive of course. All of those 22 and a half different variations, how are they going to, how is AI going to talk about us? Because you understand what it actually thinks. It doesn't give you this nice clear, this is the number. But it's a really good extra layer of detail to go into before I

Speaker A: guess knowing a little bit about how margin genius works. This is a lot of what you're talking about here is what you're building, what you built the platform to help people understand. I mean certainly information gain aspect of it. Am I right?

Speaker B: Yeah, exactly that. Analyze content to understand information gain. Like I think there's three things we do. We talk to AI a lot algorithmically, um, and analyze that to build these stakeholder maps and run some pretty sophisticated sentiment analysis on its responses to kind of get, get under the surface of what it thinks about you. We give you AI agents to analyze your content and turn like information gain into an actual KPI across all of your content or if you're an agency, across all of your clients content. Um, and then the third one, which ultimately is the answer really to the North Star thing is we connect all of that to revenue. So you integrate your CRM and we tell you okay, are you actually getting pipeline and revenue out of all of this? Um, we try not to be kind of jumping on the bandwagon of chasing hype with AI because I think it's a powerful channel, but there's a lot of people out there selling hype.

Speaker A: Yeah, but in the absence of any North Star emerging, which I got the impression you suspect that you're not saying it's impossible, it might happen, but in the absence of that, it's going to be nuanced and require. There's no perfect, easy way to do it.

Speaker B: Yeah, exactly.

Speaker A: So your wider research spans categories from fintech to HR tech. Tens of thousands of pages in some studies. How consistent are these patterns across different categories of models? And where do they start to break down?

Speaker B: Yeah, so I'll start with the models. One we've generally observed there are differences between models. I don't think it's personally too helpful to get caught up on them simply for the fact that you can't do anything about it. Um, so like, okay, Claude thinks this of us. ChatGPT thinks this of us. You can't create content for ChatGPT only. There are some nuances in terms of sources that they each look to. So Reddit is more prominent for some, LinkedIn for others. Um, so there's some nuance, but not drastic. The one exception to that that I would tell people is perplexity. If that's important to you, that behaves totally differently. Because perplexity actually isn't an LLM. It's a wrapper built, it's a research agent built on top of lots of other LLMs. So everything that I've told you today about AI, ignore me when it comes to perplexity, because that really is just the smartest search engine in the world. Um, in terms of differences across categories, we've noticed some subtle differences, not to the point that I can draw really, really clear conclusions yet. We've actually got some research coming out in about two weeks that I think will have a little bit more on that. Um, generally speaking, the more complex and messy the category, the less important citations and just visibility is, because the less important visibility at the point point of transaction becomes is much more about how you're influencing problem framing, thought leadership, all of that kind of stuff. And actually your AO strategy needs to be a lot more sophisticated. But I would add, I think the thing that people ultimately miss more than anything is that AI is such a wonderful opportunity for brands to own and influence an entire category. Because your, if you can influence the way that AI flows, frames a problem in your space through original research, your distribution on that research piece is like genuinely unlimited. It goes to everyone. It is almost like it gets ingrained into our collective Knowledge as fact, um, that is so powerful. If you're going down this kind of category creation play, I think not enough people are recognizing the opportunity to build a genuine moat there.

Speaker A: Uh, that's fascinating. And that's a real development on the theme of, you know, brands as publishers. You know, you really can, uh, leverage these engines at this point in time to be able to do something meaningful in a way that you could never do with traditional or even, or even SEO before then. So that's fascinating advice and insight. Um, so, I mean, Tom, we've covered a lot of ground in this conversation. I've always impressed by your knowledge and your thinking and all this. You're so deeply ingrained in it. And, um, it's such a fascinating area. But just to kind of wrap up the conversation, if you had asked to advise a marketer, a marketing leader or someone involved in executing B2B marketing, take just one thing from this conversation, what would it be?

Speaker B: AI doesn't search. Just remember that. I'm going to go, I'm this close to getting that tattooed. Uh, I think, uh, other than that, AI doesn't search and I think be very, very. It's a space and it's not for me to say whether I'm the person to listen to. But be very, very careful with. There's so much noise in this space. Push people on the methodology and the research that sits beneath what they're doing. If all they have for you is case studies, then that's very, very unreliable. Because often it's the AI visibility. AI search is so ingrained in the broader marketing of a company, they'll have all of these case studies where it's like, we drove this percent visibility. But it was a series A company who also started doing 1 million quid's worth of brand investment on the other side of the business. And so it's like there's so much flawed information out there from that. And that's again, not badly intentioned. Got a good case study, they talked about it. But push people on. Do they do actual original research in good faith and publish it openly, transparently to understand this new thing? And do they ever say they don't know? And if the answer to that is no, then run a mile.

Speaker A: Okay, so basically, beware of snake oil. And this is how, this is how to spot people because it's emerging stuff. And this is, and this is what you're talking about here. You're talking about people who are selling some kind of AI solution, either as a vendor or an agency. It's um, AI search orientated solution. So, um, yeah, be consumed with care is probably the overriding advice. Um, Tom, it's great to talk to you. Thanks so much for your time and being part of this today and your insights. I really appreciate it.

Speaker B: No worries. Thank you for having me do. It's been. It's been a pleasure.

Speaker A: Always good to talk to you. So I hope you enjoyed this conversation. If you have, uh, there's lots more content about AI, dare I say it, and lots of other aspects of, uh, trust and influence in B2B marketing, um, in the archive and coming up soon. So don't forget to subscribe to like or to follow whichever one is appropriate to you, given whichever platform you're consuming this content on. But from me and Tom Ford for today, thanks for listening and goodbye.