What the Data Really Says About AI Influence in B2B Buying Decisions

Tom Rudeneye: AI does really boring, repetitive tasks really, really well and it can use judgment. That job four years ago would have involved a poor human sitting down and literally every month reading every piece of content on your website and flagging. Okay, this product description doesn't match. This claim is out of date. This is no longer quite as relevant.

Bailey Gunnell: Welcome back to Content Logistics. I'm Bailey Gunnell and today we're tackling one of the most prominen freshers in B2B marketing. We need to show up in AI. A lot of teams are treating AI like the new SEO, counting mentions, chasing citations, trying to rank in ChatGPT. But the data suggests that's not where influence is actually one. In fact, a large set of B2B prompts. Most AI prompts produce no brand citations at all and those citations are only showing up in bottom of funnel queries where recommendations are required. So if citations are just the tip of the iceberg, what's actually happening underneath the surface when buyers are learning, comparing and building the criteria that eventually determines who makes it to the shortlist? To unpack this, I'm joined by Tom Rudeneye, founder and CEO at Demand Genius, who spent a lot of time studying how LLMs surface, um, brands across the B2B buying journey and what marketers should measure and do differently if they want real influence. Let's get into it. Well, Tom, so excited to chat here. I love the research report you put together and just I'm really excited to dive into it a little bit deeper. I feel like every day I am hearing people say, how do we show up in AI? Like, this is the conversation I'm hearing from all my clients from all over LinkedIn. I feel like it's a really hot topic and I'm sure you're seeing the same.

Tom Rudeneye: Yeah, absolutely. Thank you for having me. Um, definitely the short answer is that I think never trust anyone who says they know the answer to that question because it's so new. Like, I think one of the things that I've fallen in love with is trying to figure the answer to that question out. And it's super fun. Like, I don't think there's many people that have got to build businesses in such a new, exciting, dynamic, fast moving space where single hand, at the same time you've got loads of people being able to build loads of technology really fast and the technology that that's all trying to, um, unpick moving on really fast. So it's like a. Feels a little bit like a game of Whack a mole sometimes, which is an analogy that I find myself using a lot. But it's great fun.

Bailey Gunnell: Yeah. Well, that's a big reason why I loved reading through your report because I haven't seen anything like it yet. Like that's so detailed in the research when like so much of this is new and I feel like we're of uh, like the analogy of like you're building the plane while you're flying a little bit. Like we're trying to figure things out as like someone. So much stuff is changing and it's difficult.

Tom Rudeneye: I feel for like it is literally my job and we as a business dedicate, I say 30 to 40% of our time to research into this area and I cannot keep up. So it's like I really feel for marketers who are left with that alongside this other job of trying to unpick this brand new thing that uh, I certainly believe is completely transformational in terms of how they're going to reach their customers, how they're going to talk to their customers. Because you've got this intermediary placed in between now. Uh, and you need to learn to work through that. Yeah, I have a lot of respect for people who are able to do a full time job and keep on top of that. And that's one of the, one of our big goals at Demand Genius Beyond. Obviously we want to build a great business, but I also, I am really enjoying building out the research arm, um, of what we do. I want that to be what we're known for and partly because it's fun and partly because I think it's just a useful thing for people.

Bailey Gunnell: Yeah, it does seem like a lot of marketers are just getting this like added on to their job description. They're like, okay, you're already doing content, you're already doing SEO. Like can you add AEO to that? Can you figure out how to show up in LLMs, um, on top of the other stuff you're doing?

Tom Rudeneye: Yeah, well, and I think that's a, I think that's one of the challenges with it frankly, because it doesn't create any space for people to really think deeply about it. Like one of my maybe slightly more controversial opinions. So I certainly don't. There's a. You'll find different camps on the AAO question, but one of the popular things that people will say is AEO is just SEO with a bit of a new rebrand, bit of a fresh lick of pain. I could not disagree with that more strongly. Or at least certainly there are huge components to it that are A very different skill set to SEO. But I think one of the challenges is it's the SEO vendors and it's the SEO kind of thought leaders and the SEO practitioners who have adopted aeo, and they've very understandably applied an SEO lens to it. SEO best practices help with aeo. Uh, they are kind of fundamental, but there's a lot more that goes into it.

Bailey Gunnell: Yeah, that's really interesting. That was one thing I saw, was pretty prominent throughout you, your research. What would you say are, and this might be a hard thing to answer, but, like, what are some of the big differences you're noticing and like, how people are approaching SEO versus aeo?

Tom Rudeneye: So I think, well, one of the challenges is I don't think there are enough differences in how people are approaching it. But I think maybe, maybe an easier question to start with is what do I think the differences should be? Because, and certainly I always think, uh, sometimes come across way too disparaging. Certainly there's a lot of people out there that are doing some fantastic research into this topic. There's just also a lot of people who are spouting myths. That's the kind of noisy world that we live in, unfortunately. But I think there's a few fatal flaws that people make at the moment. One is they equate a prompt and a keyword as if they're the same thing. They're really, really different. So a keyword forces a user to consolidate their intent or their question into a pretty sure kind of string of words. There are only so many variations, therefore, of each question that you can produce. A prompt is totally different. It's long, it's personal, it adds criteria, it assigns Personas, it's verbose. Everyone does it differently. I do. I just discovered speech to text. So I have, if I press the function key on my, on my laptop now, it starts recording. So my Claude. Claude could look at when I started doing that and be like, what is happening now? It's just like talk. Yeah. Like, whereas I used to, because I had to type stuff out, I was pretty concise because it was boring and long. I lost sight of my point a little bit, got started going on that tangent. But yeah, they're very long, very personal. So there's so many different permutations that you have to optimize for. So it changes the way that you approach optimizing. Because one of the challenges, I think the, the biggest trap people are falling into is they are using the old role of content. You create a piece of content and you match it to where you keyword and you keep doing that one to one and kind of knocking off keywords that are relevant. The temptation at the moment is to say, okay, well there's loads more prompts, but we also have this thing called AI that can produce pretty okay content at scale. So where you've got more prompts, we're going to produce more content. And that's where everyone is just creating these huge libraries of content, attempting to do that. But the challenge is, I think you and I equate it to the black hat days of SEO. It's like ao, uh, you create this mess of a library and within that are all sorts of contradictions, confusions as to who you are, who you're for. Like the fundamentals of marketing suffer. And as search algorithms, if we want to call it search, get more and more intelligent as we move from Google to AI, the fundamentals of marketing become more important because they're more able to pick up those things, pick up who you are, who you're for, where you're strong, where you're weak. And as Trump's assigned criteria, that's actually what they're looking for more than this one piece of content that might be able to be cited for this very specific thing.

Bailey Gunnell: So I loved the way you coined like the age of black hat ao. So it sounds like you're strongly discouraging people from just like mass producing content, using AI engines to try and basically like hack the system, um, try and show up for all these different unique prompts and use cases.

Tom Rudeneye: Yeah, basically, I think, I think, I would say you have to understand the trade offs. So there are short term visibility gains to be had from doing that. And that's what you see around at the moment. Right. It's one of the things that can be a little bit wary of with, with, with some agencies that are out there that profess to do this because they will be able to show you a case study that shows that you're a predetermined set of problems that helped. But you, you take on when you publish a piece of content now because m far more so than search did, AI looks at your entire library and all of the information that is out there to form a picture of you. When you publish a piece of content, you take on a maintenance burden, a maintenance debt. Right. That's what we call content.

Bailey Gunnell: Yeah.

Tom Rudeneye: Over, uh, time that expands and becomes unmanageable and will start to have a cost. So I'm sure you can make a case for go and publish a load of content over the Course of six months, get the benefit. But you just need to know that you're building up that debt and you will have to pay it off at some point. I think most brands are doing this without any plan for how they're going to pay off that debt and maintain the consistency, clarity and quality of all of that content they've produced over time.

Bailey Gunnell: Mhm. So depending on the size of your team, you're probably better off having a smaller number of really solid pieces of content that maybe can help with multiple prompts or just help AI get a bigger picture of what you're doing than to mass produce a lot of pieces.

Tom Rudeneye: Absolutely. I think there's always a balance to be had in a certain amount of volume. Does help with authority. The way that I like to think of it as a mental model I have, which I find really useful, which is that you have three levers to influence AI. Right. You have positioning, content and reputation. So your positioning is who you are, content is how you communicate that and reputation is whether other people enforce that. The challenge is. And most, most brands find that over time those three things creep apart. Because what you do is you decide on day one, this is who we are. You produce content, you capture reviews to that effect, and then a year later you reposition and this is now who you are. That content stays and now you do the same again. And over time this gap opens up. That didn't really used to matter because that old content just sat there. Now that content does still matter. So brands have done absolutely what they should do. You should be repositioning and tweaking that stuff over time. But you need to kind of clean up after yourself a little bit more. So I think the risk is you over index on the content side. And what suffers then is your positioning because it now is confused because you've got so much content saying all of these different things.

Bailey Gunnell: Wow. That, that is a really great like lens to view it through that. I love that. I'm going to use that.

Tom Rudeneye: You can have that.

Bailey Gunnell: I really like that. I need a visual. That's a great way to, to view it through things because I can see where, yeah, as, as one area starts to change a little bit, the other areas just need to adapt to keep up. So I'd love to dive in a little bit into your research now. One of my big takeaways from it is where citations actually show up. And I feel like, uh, that's the thing I hear from so many people is like, well, how many AI citations do you have? And that I Feel like is a lot of the metric a lot of people are using to show if they're winning or they're doing well. But I'd love to hear your thoughts on, um, why is that maybe misleading and where we can maybe view it in a different way.

Tom Rudeneye: I think that's another hangover from an SEO lens a little bit, is focusing on citations. It's like thinking about rankings and visibility rather than kind of deeper influence. So what we found as we went through the study and for context, maybe for people who listen to this, what we did is we did a fairly deep study where we. And when I say we, I, uh, do of course mean my co founder, who's far better at this kind of stuff than me. We simulated prompts across complex buyer journeys in loads of different categories. The idea was to try. And one of the hypotheses that I had was that one of the errors that we make at the moment is we treat each prompt as an isolated event, whereas in fact it's influenced by all of the stuff that came before it because that conversational context is maintained. So what we wanted to understand was okay, if we simulate prompts across awareness, consideration conversion stages in particularly complex categories. And we did it across 14 different B2B categories until we were kind of satisfied that what we're seeing is pretty persistent and reliable and we publish all of this, by all means, go recreate the study. What we found is that over the course of that journey from awareness through consideration to conversion, a few things change. So the behavior of the models in terms of how they create their answer changes a lot in awareness and consideration. We, uh, very rarely saw retrieval invoked. So it's working mostly from its internal knowledge. Right? It's like me answering off the cuff instead of going and looking up. Produces much worse answers. But that's the medium we're in. Very rarely invoke retrieval. Brand mentions were rare, but did happen. Brand citations never happened. Right. They never linked off to a page. As we got to conversion, then brands are starting to be cited a reasonable amount. Mentions and retrievals were all pretty common. What we also noticed though, uh, which was particularly interesting, and that's all quite known, particularly interesting though was the language completely changes. So as we go from awareness through to conversion, the language goes from um, exploratory, comparing options, comparing trade offs to much, much more decision oriented. And what we realized is that's there's a big danger that brands can have, which is you track BofU prompts, bottom of funnel prompts, and you get back decision Oriented language that is just like it's, it's recommending things.

Bailey Gunnell: Yeah.

Tom Rudeneye: You think that's because it really, really likes you? It's not. It might really like you, but that's not why it's doing it, because of what we call intent matching. So it matches the intent of the user. So the user AI is smart. The user's query was helping. They wanted you to help them make a decision. So it helped them make a decision.

Bailey Gunnell: Yeah.

Tom Rudeneye: So what you want to measure is like a cross reference of top, middle, bottom of funnel, adjust for the likelihood of, or the natural positivity that AI, uh, has at that stage and then understand. Okay, how is it talking to you relative to that? So that was one of the big takeaways that we had, as well as just the fact that citations in B2B journeys, 16% of prompts that we ran across them produced a citation. So it's a really bad kind of North Star metric. It is a useful thing to track, but if that's your headline goal, then you're building your whole case around 16% of AI's impact and that that's underselling what you're doing and it's also going to miss a big chunk of the opportunity.

Bailey Gunnell: Yeah. So much to unpack there. Yeah, there's a lot of good stuff, but I feel like you nailed it when you're talking about like, it seems like we have such a focus on this bottom of funnel, like, are we getting citations? Are we showing up positively? Like are we showing up when people are asking these conversion questions. But you have the analogy of the Seisberg and I'd love for you just to explain a little bit of what that means because I think it's a great way to kind of view this.

Tom Rudeneye: Yeah, well, it's my way of visualizing what I was talking about at the end there, I guess, which is what you're seeing in. So if you're looking at this at, ah, through the traditional metrics that we as content marketers, as SEOs are uh, trained to look at, which are what things like ranking citations and traffic. Then what you are seeing in your traffic, what you're going to see is a fraction of the 16% that I just described. Right. 16% of prompts produce a citation. So already that is showing you a tiny fraction, the tip of the iceberg. And it's the stuff that's above the surface. Then how many people actually clicked on that citation? It's one of the things that we want to study next. Actually, I would Guess much? Let's say conservatively half of them. So if you're looking at traffic, that's 8%. So there's a lot of mistakes that people make there. Sometimes people say, yeah, we don't actually think that AI ah is having that big an impact. We looked at our traffic, the numbers are pretty low. Like, okay, we'll times it by 10. And then you're starting to see the actual impact that it has the same processions except times it by five. A lot of the mass sits beneath the surface. And that's where the iceberg analogy, I think is something that really helps me picture that. That's the kind of messy stuff where problems are framed, requirements are built. No brand ever gets cite. All of the cited, all of your metrics are missing that. But that's where the meat of the challenge and the opportunity exists. It's harder.

Bailey Gunnell: Yeah, I want to make sure I'm understanding that. So you're saying if, if you're, you're only seeing like a couple clicks coming from Chat GPT or whatever, like it's probably actually more than that. Like is uh, that like an attribution challenge or, or what would you say is going on there?

Tom Rudeneye: The challenge is that, so people only click through to you from chat GPT. When Chat GPT shows them a link, that allows them to click through through to you from it. Okay, so if we know that tip of the iceberg is someone clicked through to you, they only see a link 16% of the time. That would allow them to do that, uh, how many times out of that do they click on that link? So that's all the stuff that your measurement is capturing because you're able to track citations against a flawed set of prompts. But we'll leave that one for another day. And then at the tip of the iceberg, you see the traffic. So it's very tempting to say, no, I can see the traffic, I have visibility on that. It's not that high. But when you work through that pain, because it's another thing, like we're trained to think of all of this stuff in a funnel, right? Because that's how search was. It was a directory that was designed to link to you. So it was, it was a funnel. You were capturing people early in their journey who had a question. And your content job was to help them take that question and lead them from problem aware to solution aware and buying something. Right? That was the job. The content performed. I, this is all kind of led us to the analogy. I was like, actually it's completely flipped because all of the traffic now comes from bottom of funnel. So people are arriving at you solution aware they've been on a journey to that point that they're talking to AI to. That's the kind of dark AI piece. But so yeah, what you're seeing is no longer the broad part of the funnel, it's the narrow tip.

Bailey Gunnell: Okay. Um, I think there was one line from the report that said something along the lines of like AI visibility and AI influence aren't the same thing. So is that kind of getting at the idea of you're really only seeing the part when they're clicking on your link but you're not seeing all the stuff that's going on beneath the surface?

Tom Rudeneye: Exactly that, yeah. Like visibility is the stuff that's above the surface. Right. That's what you're currently seeing. And you, you want to be visible there. Right. So it's not that visibility isn't one of the outcomes that we want, but you also want to influence in a deeper way how the model thinks about you and your brand. There's a couple of things that I can go into here. Like one of the trends that we really, that we really noticed as we did this was this trend called convergence. So that's the idea that over the course of a conversation, what's tempting to think of from a search mindset is that ah, we ask it a question, it goes out, it finds an answer and that's kind of how we're thinking about this, not how AI works. Right. We take it a problem, it talks to us, it applies criteria, understands who we are, may already have some of that knowledge and then it gives us the answer. So over the course of that process, what it's doing is it's applying criteria and narrowing down its option pool. One thing that we noticed, awareness stage prompts, the option pool is super, super broad. So there's very little variability between if you run the same prompt ten times, it'll mention different brands each time because it's like thinking it's like it's in crazy thought partner mode. And then over the course of that process, as you, the user, uh, with it hone in on your own on a decision, it gets much more conservative and it starts drawing from a much more narrow option pool. So what that means is there's very low variability by the end.

Bailey Gunnell: Yeah.

Tom Rudeneye: So basically if you're showing up as highly visible, it doesn't mean that you've done some brilliant optimization. It means that you're just one of those few people that's in an option pool. Now the big, big challenge that a lot of brands have is I've signed up for lots of different of these tools. How do they go? How do they produce the prompt that you want to track? That's one of the big challenges at the moment. How do we know which prompts to track loads of them? Well, they scrape your website. So what we do is we scrape your website, find prompts to track and then we tell you, great, you're tracking against them. There's a clear like flawed, self fulfilling prophecy there. And when you understand how it works in terms of the narrowed option pool, you're picking on prompts where you explicitly fit the option pool and then saying, great, we're showing up. What you really want to know is as the option pool gets a little bit broader and the LLM essentially has more of a choice as to whether to suggest us or not, then are we starting to show up? And a lot of that might not be showing up via explicit citation. But what gets really interesting is, okay, is our brand terminology showing up. How aligned is our brand narrative with the narrative that the AI has for our uh, problem space? The analogy that I really like here, I go back, if you can't tell by the amount that I'm talking. I used to be a sales rep and one thing that we always said when we were doing that, you know an RFP is like a request for proposal.

Bailey Gunnell: Yeah.

Tom Rudeneye: If you get an rfp, I could look at that rfp, read it and I knew exactly who's going to win it. I knew whether, and I knew why we would lose, why we would win. Because typically some, some, one of the vendors has helped the prospect write the RFP and I could read it and I would see certain requirements in there. Like those guys helped write it. I could tell you who did because. And that's what happens when you start to really understand a sector. You know it to that level of depth. I equate that to uh, uh, this. If you're just thinking about citations, you're receiving an rfp, filling it in, send it back. What you actually want to be doing is be the one that helps them understand how to frame their problem, what requirements they should have. And then winning the RFPs becomes automatic.

Bailey Gunnell: Yeah. When you're, when you're the one setting

Tom Rudeneye: the requirements, you're setting the parameters, you're setting. What if we're going to like, okay, let's, let's peel back the veil a little bit. That's what I'm doing right Here right now.

Bailey Gunnell: Right.

Tom Rudeneye: I'm going on these things and I'm trying to explain to people how, how we've understood AO to work. And obviously we've gone and built a solution around that. So, yeah, there we go. There's a, there's a horrible dirty window into the unspoken context of this conversation.

Bailey Gunnell: Of course. Okay, I would. I was really glad you got to talk about convergence, because I felt like that was a really interesting part. And I think in the, the report, like you mentioned, it starts really broad and then it starts to get more specific as you're. You're going through the funnel. And then I think it had also mentioned that it starts recommending like pretty consistently as, as you get lower.

Tom Rudeneye: So I would say two things. There is an opportunity for like, optimization in the way that we traditionally think of it will produce some results because there is some variability. And so you can influence that. And you can also clarify exactly who you are and make it easier for the AI to understand that. So things like making like a lot of the hacks around data schemas and stuff like that. I've not seen a lot of good evidence that works. I'm not saying it doesn't, but I've not seen it really proven. AI is. LLMs are designed to take unstructured text and turn it into structured data. I've not seen much evidence that kind of shortcutting that for them helps. But, uh, that's not an opinion. It's just that I'm not. So a lot of those hacks, I'm not fully on board with certainly making your content direct, giving concise answers to questions that come up a lot will help that even if it doesn't produce direct citations, it's good for you. So produce quality content that answers questions people have and you're going to help yourself. So that's one thing. And I do think that there's certainly value to that. And that's where a lot of the current AO strategy, that's the part of it that is a lot more like SEO. And that's what I think many people do a good job at and actually probably is much better people than me to tell you about how to do that, because it's, uh, it's not the part that I'm most interested in. What I think the really big kind of untapped opportunity is by learning to shape which option pool it goes to. So if we're saying that, okay, once it's, if this is the option pool, there's very little variability. Your question is Kind of like, well then what's the point? Do we all just throw up our hands and say we can't? This is, this is impossible. Then what you can do is be really specific with how you win within that option pool and you can help push more and more people in that direction. So if you can influence further upstream, if you think you've got five pools down here that people could possibly go into up here, there's a great big sea m mess of people. This little visual diagram is terrible for podcast, but what you want to do is push more people into your pool. And the way that you do that is through really good in depth research that actually moves the industry's understanding forward. That will influence humans and that's what will get LLMs to pick you up whether they cite you or don't sight you. It will help them frame problems in a way that will inevitably lead to you. There's a, there's a framework that we have for that which I think everyone should make like a content KPI at the moment which is, and it's not our, we've kind of built a model to measure it, but not, it's not our concept or anything but information gain. So like I said, like LLMs are fantastic at summarizing information for people. They do it really tailored for every individual. Traditionally content marketing summarizes knowledge for people, right? That's what we do. We find a high intent query, we summarize knowledge against that query and Google ranks whose summary was the best. That that is how the Internet has always worked. That's pretty pointless now because I can get a personalized summary from Claude chatgpt. So why do I need to go and read all of these ones to find out which one can give me the information. What isn't pointless is net new information that can allow Claude to include you in its summary. So the way that we think about it is we have a framework. It's like three levels.01,2 3 each one is a different level of information gain. So there's basically no information gain. You've just summarized interpretive information gain, empirical information gain and then conceptual. So interpretive is here's existing knowledge, but a new way of thinking about it or a new take on it. So that still has value and that will seep in and can. Can help move knowledge forward. Right? Empirical gain. Data. Research loves data partly because it's that we love data and research and then conceptual information gain is kind of data and research that leads to almost a little bit of a eureka moment right? So leads to something, a completely new concept. I like to think that time will tell, but I think that what we've uncovered in Dark AI is a good example of that. Now not every topic allows that. Right. It's very hard to predictably produce conceptual information gain because it's hard. We understand a lot of things. But I tend to think that or the way I approach it for Demand Genius is that level one is the baseline content doesn't go out of this business unless it's level one. Because what's the point? Uh, like uh, there's no point in anyone reading it and we want to consistently on a monthly quarterly basis produce twos and threes.

Bailey Gunnell: Yeah, that's a great way to frame it because I think pretty much anyone can create the level one content. That is the stuff that is the easiest to create. But I mean because it's easy, everyone's doing it and that's the stuff that people can easily find the answers on AI and through Google Search. But yeah, that unique research that you have, which is, I feel like probably part of why you did this, that this report is, is what's gonna um, it's like citation worthy content.

Tom Rudeneye: Yeah, exactly that. It's, it's. What's the reason for AI to cite you? If you, if all you're doing is putting out content without giving it a clear, clear reason, you should cite me. Because this is some knowledge that you need and people need, then you're not going to do that very predictably. So yeah, absolutely. I think that's where there's still an important role for humans. I think it's one of the challenges that a lot of marketers are going to have is there's a lot of headcount pressure at the moment. And that makes doing. Because there's such an obsession with volume and such an obsession with efficiency that it's going to make doing that kind of deep experimental work a lot harder. So the temptation is to cut headcount and go and produce crap. But I think that's actually going to be very counterproductive. I would encourage marketers to think about how they can find a way to build that in. And it's foundational stuff, right. AI. I tend to think AI can produce level one information gain. It's kind of capable if you use it well and you feed it with your original perspective. You can point to the new topics and it can kind of bring that perspective in. It can do level one. I've not really seen when we've, we've built a free calculator on our website that uh, takes a content brief or a piece of content and scores it out of three and gives you how it could, if it could be, be elevated. I've not seen that. I can predictably produce level twos and threes and unless you use it well, it'll produce banal. Yeah, zeros. Right. Because your content is doing nothing that I couldn't just go and ask ChatGPT and have it spit out to me personally.

Bailey Gunnell: Yeah, I'll include the link to your, your tool in the show notes too so people can test it out. One thing I'd like to dive into a little bit more is prompt tracking. You had mentioned previously how people will like scrape their website, see like what prompts they're maybe already showing up for and then it's that self fulfilling prophecy of like well yeah, you're like already showing up for those things. So that's not really the most helpful. Like what would be the alternative approach to that when they're trying to track some of these prompts?

Tom Rudeneye: Yeah, so there's a few things that we recommend tracking. I do think there is value to just tracking blum visibility in that way. I think it's useful. We have this kind of adjusted metric which basically weights the visibility based on how hard it is to get. So we're going to weight it differently based on whether it's top, middle, bottom of funnel and so you can kind of see the weighted and the raw score. And that's quite useful. It just gives you m. It makes it gives it a bit more context. The other thing is just to think really carefully about how you hone in on that original list in the first place. So the temptation is to. It's a really annoying job and I've spoken to lots of marketers who are like how do I come up with this list? Am I going to actually learn about my buyers and write that down? Heaven forbid. But like being pedantic aside, like yeah, it's, it's, it's a hard job to go and fill out a list of 200 things. The temptation is you get AI to do it for you. And what does it do? It uses your website or that's what a lot of the like the free tools out there will do. There's some really good sources. So one thing we do is we use performance data to constantly improve that list. So we can integrate with your CRM and then say okay based on what people are reading in each stage. We can actually work back from that and say okay, what's likely leading them here. What questions are they having at each stage? So that's one good thing that you don't necessarily need us to do. Call transcripts. Great. Get all of your sales. Call transcript. I'm actually, I was saying you at the uh, the start of this conversation. I'm having a lot of fun vibe coding at the moment. I'm coding another tool for our website which will be that you can upload these transcripts and it will go and shoot out your shootout and cluster your prompts for you. But that's a really good one. Take, take that. That's like good like what questions that Buy is actually asking. But I think it's that list of prompts. While it seems like an annoying admin job, I think increasingly it should be something that you view as quite foundational. Right. Like that's understanding your buyers and understanding your buyer journey. If that's not worth thinking a bit about, then what is?

Bailey Gunnell: Yeah, that's uh, a bit of a

Tom Rudeneye: mic chop, isn't it?

Bailey Gunnell: Yeah. I gotta sit with that for a minute. Yeah. I love using sales transcripts always. There's gold in those. Really, really helpful to, to hear how customers are, are talking about things and, and if I'm understanding right, it sounds like prompts across all uh, the entire iceberg. So not just the, the citation worthy type stuff but uh, where are they starting and the journey to asking what kind of features do they need in their financial tools and stuff like that all the way down to the bottom.

Tom Rudeneye: Yeah. I think you want it to be a kind of a representative mix that allows you to understand how you're performing at every stage. It changes the way that you have to analyze that as well. So that's kind of the first step is get a track the right things, ask the right questions that give you good answers. You also need to do some deeper analysis. So it's not enough just to count citations because particularly at the top of it will be zero what you like. There's a few things that we do. One is looking at uh, sentiment. Sentiment is quite a tricky thing to track. You can get very like sentiment analysis is very easy to do nowadays. So you can kind of vibe code. Yeah. Into your, into a product. The challenges LLMs are crazy, crazy positive. They are. I always think that they're very, they're very American because they've been built largely in Silicon Valley and you need to apply some healthy British skepticism and grumpiness to try and like they're going to say you're great and if you Ask it about a brand, it's going to give you quite a positive reflection. So you need to kind of learn its language and learn to treat lack of endorsement actually is a bit of a negative. It's a sign that it's not very okay. So there's some complexity to understanding sentiment and I think it's very tempting to look at two products tell you they do sentiment analysis. One tells you you're great, one tells you you're not. And you know, we'll just use that one. So that, that's one thing that can be very useful. The other thing that I think is really important for understanding is actual perception of you is your alignment with different criteria. So one thing I mentioned is LLMs. They tend to be assigned Personas and assigned criteria. They understand a lot more context. So what you want to understand is how, how does it perceive our strengths and weaknesses and how does that map to the criteria that the people that we want to reach will apply? So what are the different kind of lines of that industry segment? Use cases, stakeholders? What you want to understand is, okay, who do we want it to think we're for from an industry perspective? Or what use cases are, uh, we really strong for? Do we know that when a deal comes and they're saying this is what matters to us, we'll win that deal, what do we want that to be? What does it think that is? And you can kind of, again, you can do math and measure, um, the alignment between those two things. So I would make that a KPI. If those two things are in lockstep, visibility will follow.

Bailey Gunnell: Okay. So rather than just tracking citations when they're only showing up 16% of the time, track sentiment and other, uh, other things that I missed something there sounds

Tom Rudeneye: like brand sentiment, maybe mentions brand sentiment and alignment. So how does, how does, how does its perception of, uh, what you're good at align with what you're actually good at? If those two things are really good, it's going to surface you to the right people and you're going to survive that process of convergence.

Bailey Gunnell: Love it. So now for the million dollar question. How do you help people do this? What does Demand Genius do to help solve this problem?

Tom Rudeneye: Yeah, so we've kind of recently launched the product. Basically the first step is measuring the right things. So we make it really easy to hone in on a list of prompts, use real performance data to evolve that and improve that over time, make sure that it reflects changes in the market and what you're actually seeing and kind of plugs into all your data sources and then ask LLMs the right questions and do much, much deeper analysis on the outputs. So not just counting the citations, but looking at the outputs to try and understand, okay, how does it align with you? How does it align with your competitors? Importantly, how does it view your strengths and weaknesses? How positively does it talk about those strengths and weaknesses? And how does your problem framing and the way that you think about your market show up in the LLM dancers? Or does it at all? So that's one step. And then the next step is like, how do we act on that?

Bailey Gunnell: Yeah.

Tom Rudeneye: So what should we publish? What's like across content and reputation? What should we be trying to improve off the back of this? Don't go into the kind of generating content part of it. There are tools that do that, there are people that do that. You need to decide for yourself that. And then the final part is the ongoing maintenance of content is one thing that we do. So we have AI agents that allow you to basically constantly evaluate your entire content library across literally anything that you want. So it's really good for doing deep qualitative analysis. So a good example is information gain. You can set up an agent that tracks information gain, goes reads every piece of content across your whole site, once a month, once a week, whatever the right thing is, scores it out of three. Suddenly that's a content KPI that you can actually track and report on and measure and you can flag whenever a piece dips or if you're using a different one, whenever, chat currency, whenever there's an out of date claim in a piece of content. So to that point of keeping your positioning content reputation all aligned. It's the monitoring that you need to do to keep bringing content back to positioning.

Bailey Gunnell: Yeah. Wow, that is really cool. That is the time suck there is keeping track of all the content and making sure it's aligned as things change.

Tom Rudeneye: So yeah, and it's something that I think is a lot more important now than it used to be, which is why it's a bit of an unsolved problem.

Bailey Gunnell: Yeah, yeah. And goes into like what you're talking about with alignment, you might have little pieces that just, just need a little tweak there to make sure they're aligned.

Tom Rudeneye: Yeah, well, and it's AI does really boring, repetitive tasks really, really well. And it can, and it can use judgment. So that job 40, like what four years ago would have involved a poor human sitting down and literally every month reading every piece of content on your website and flagging. Okay, this Product Description doesn't match. This claim is out of date. This is no longer quite as relevant because this person has published this or all of that kind of stuff. You just couldn't do that across 500,000 pieces of content. AI can. It's quite patient.

Bailey Gunnell: Yeah. And that's where, like, where humans can spend more time working on that. Level three and level two content. That's the stuff that's.

Tom Rudeneye: Yeah. If you're listening to this as a marketing leader, don't fire those humans. Get them doing, get them thinking.

Bailey Gunnell: Yeah. Yep. I got one last fun, fun question for you, but what is your biggest content marketing? Or it could be related to AI. Like, what are some of the things you're seeing right now that you're like,

Tom Rudeneye: oh, the one that has really started to wind me up is they're like, it's not this, it's that. Yeah. Do you know what I mean? It's like, it's not. And it's like, ah, uh, what I hate most is that it's learned that from us.

Bailey Gunnell: Yep.

Tom Rudeneye: It, like. So I, I think in the early days that was something that would have persuaded me because I think that kind of comparison, it. It was. I feel like it sometimes reflects back the, the most annoying parts of ourselves.

Bailey Gunnell: Yeah. Oh, yeah. No, that comparison gets me too. I'm like, oh, yeah, just delete it. Just say it is this.

Tom Rudeneye: Don't. Yeah, yeah, yes.

Bailey Gunnell: Delete the comparison, please.

Tom Rudeneye: Yeah. And the ick is when you go, are you going to. Like, I've gone to big, big businesses that have like whole product marketing teams and on their homepage, I'm like, this is just so obviously AI, like, and then you can go on their on like sales app and see, like this. There's eight product marketers there. What are they doing?

Bailey Gunnell: Yeah, yeah. I think. Well, I think not everyone maybe sees that, but anyone who's using AI pretty regularly, that's like a red flag. You're like, okay, delete.

Tom Rudeneye: Yeah, yeah. The EM dash is the other obvious one. Although I feel people who just used that before, it's like, okay, uh, I now just can't.

Bailey Gunnell: Yeah, no, that one, that one's pretty bad. Delete all the EM dashes, please.

Tom Rudeneye: Just, just go through it. Turn it into a regular dash. Yeah. That's what ordinary people use.

Bailey Gunnell: Yeah. And I, I do feel bad for the, the writers who were doing that well beforehand. I'm like, I'm sorry,

Tom Rudeneye: I heard a funny story. I can't remember who it was now, but I was on a call with someone and it was a content marketer, and they were like, there was someone who interviewed and they got them to do a task and they sent it back and they suspected it was just AI. And it's like, I don't mind using AI, but it's not really the point of this. Right. Like this to test your. And the guy was insisting that it wasn't, and so he was like, I tell you what, just do an M desk for me right now and put it in the chat.

Bailey Gunnell: Oh, no.

Tom Rudeneye: And it's like, he does. He doesn't know how to do it on the keyboard.

Bailey Gunnell: Just calling him out live, man. That's a good test. Got to see if you know how to use them. Dash. Love it. Well, Tom, where can people go to learn more about you? Where can they connect with you?

Tom Rudeneye: Sure. Head over to my LinkedIn. So, Tom Rudnai. Rudny is incredibly hard to spell. R, U, D, N, A, I. Good luck. Or demand-gienius.com slightly easier to spell, so maybe go for that one.

Bailey Gunnell: Yeah, I'll make sure to attach all those links in the show notes so people can find you easily and not struggle with spelling. Thank you so much for letting us chat about this today. I learned a lot, so I'm excited to get started on implementing some of this stuff.

Tom Rudeneye: Good. Well, uh, lovely to talk to you. Thank you for having me on.

Bailey Gunnell: Thanks for tuning in to another episode of Content Logistics. Make sure to subscribe to the show so you don't miss the next episode. Bye.