Why Aren't AI Engines Citing Your Content? (Hint: You're Missing Knowledge Graph Enrichment) — with Paul Rowe

📬 Love the podcast? You’ll love the newsletter.
 Get the weekly 3-2-1 on AI search + marketing: Subscribe

If you've been wondering how to actually get AI engines to start citing your brand, this episode is for you.

Cassie sits down with Paul Rowe, founder and Chief Generative Engine Officer at NeuralabX, who has been heads-down researching generative engine optimization since the Princeton GEO study was published in June 2024. Paul brings a rare combination of original research, live citation testing, and a refreshingly clear point of view on what's actually working inside ChatGPT, Copilot, Perplexity, and Google AI Mode right now.

Spoiler: the brands getting cited aren't winning because they nailed traditional SEO. They're winning because they understand something most marketers haven't caught onto yet.

In this episode, Cassie and Paul cover:

• What knowledge graph enrichment actually means in a GEO context (and how it's different from the SEO definition you already know)
• How Paul is running live screen-recorded citation tests across four AI engines—and why that kind of original content is catnip for AI engines
• The AI citation benchmark study Paul is running on his top competitors, and what the data is showing
• Why service-based businesses have a built-in GEO advantage through case studies
• The recency stat from Profound that should change how you think about content updates (spoiler: 56% of citations come from sources updated within the last month)
• What "updating a post" actually means, and why changing the date doesn't count
• Why the "SEO is GEO" narrative Google keeps pushing is only about half true

Resources:

Princeton's GEO study

What is the FSA Framework?

Let’s connect:

LinkedIn → Cassie Clark | Fractional Content Strategist
Website → https://cassieclarkmarketing.com

Download Freshness, Structure, Authority: The Framework for AI Search Visibility:

Amazon

P.S. Is your brand losing its "Answer Authority"?

Most series A/B and enterprise brands are being "nudged" out of AI search results because of entity gaps and "stale" content. I am opening a limited number of specialized audit slots to help you reclaim your Share of Voice using the FSA Framework (Freshness, Structure, Authority).

Request your 7-Day AI Search Visibility Audit: https://cassieclarkmarketing.com/ai-search-visibility-audit/

Transcript

SPEAKER_01 1:13

Hey, welcome back to Found in AI. I'm Cassie Clark, a fractional content strategist, AI search optimization expert, and the host of this show, Found in AI, where we figure out GEO, AEO, and AI search together so we don't get lost in this wild new wave of how people are actually finding information. Today's episode is one I have been excited to share since Paul and I sat down together. I sat down with Paul Rowe. He is the founder and chief generative engine officer at Neurolab X. That is a real title now. Paul has been studying GEO since June of 2024. This is right after the Princeton GEO study dropped. I will link to that in the show notes. If you have not read it, I encourage you to do so. It is a fascinating read. But in this episode, Paul and I get into knowledge graph enrichment and what that actually means in a GEO context, not the old SEO definition that you're used to. Now we also talk about the live AI citation benchmark test that Paul runs across Chat GPT, Copilot, Perplexity, and Google AI mode. And he tells us that doing this kind of original research is one of the fastest ways to get AI engines to start noticing you and treating your brand like a trusted source. We also talk about recency and we get into the GEO SEO debate that Google and half of LinkedIn keep discussing. Paul and I both think that's maybe half true, but grab your coffee and your notebook, whatever you need. You're gonna want to take notes on this one. Let's get into it.

SPEAKER_00 2:57

Okay, my name is Paul Rowe, and I am the founder and chief generative engine officer at uh NeuraladX, and I should say CEO with that as well, because I didn't get the actual title right. And uh we specialize obviously in generative engine optimization, which is transferring your website's content to have all the criteria necessary for AI engines to want to cite you as an answer to users' queries.

SPEAKER_01 3:29

Yeah, so you and I have been talking a little bit before we hit the record button. So, how long have you been working inside GEO, AI search optimization?

SPEAKER_00 3:39

Yeah, that's a good question. Um, for generative engine optimization, I started studying it around June 2024. That was when the actual Princeton study was fully published. Um, and so from that point, once I found out all of that information, it really ignited a light bulb in my mind. And I thought, wow, this is revolutionary, this is something that I need to look into and study and potentially move forward with in regards to a career, because the gigantic scale of what that represents shook me, and so that's what really sort of uh drew me to GEO.

SPEAKER_01 4:26

Yeah, so we were also talking about knowledge graph enrichment. Now, if anyone has been in SEO for a while, knowledge graph is not a new term, but when it comes to GEO, can you explain what that means?

SPEAKER_00 4:40

Yeah, yeah, sure. It's uh knowledge knowledge graph enrichment is effectively when you are putting new information into the knowledge graph. So, as from my side of things, trying to do it with my organization, I've done it by doing live screen recording testing. So you can imagine for a particular prompt, let's say what is the cost of generative engine optimization in the UK, I then did a live screen recording test of that in four AI engines, being Chat GPT, Microsoft Copilot, Perplexity, and Google AI mode. And I did them consecutively, and then I recorded the results. So then obviously, by doing this by video, I then put it on YouTube, I then put it on my website with transcripts. This is new information that Google's receiving of somebody actually doing live testing and recording the results. So you've now given a new piece of knowledge enrichment to the knowledge graph that Google has. So that's one kind of methodology that I've employed in order to achieve knowledge graph enrichment. And also by doing that, you are getting the attention of Google because it's looking for specific, unique, original content. And so if nobody else has done this before and you're now providing this type of content, Google goes, ah, this particular organization is providing knowledge graph enrichment. We want to crawl their site more often, and it helps give you a bit more authority as well. So it's a particular tactic that I've used in order to get my company more pronounced in the kind of uh understanding of Google, if you will, how they perceive me, how the algorithms look at our organization. And so that's the kind of uh strategy that I employed there. And then I thought to myself, well, what can I do next? Okay, I've done live screen recordings. How else can I contribute to knowledge graph enrichment? And so I did my research and I thought, well, is there any AI citation benchmark studies? And so I looked on the knowledge graph, i.e., all the information that's out there on Google, and there was nothing at all. So I then found an AI citation tracking software that had uh very strong credibility and respect in the industry, which was Otterly AI. I jumped on board with them, I then researched the top five organizations in my industry in the UK, and then I put them in the software and also added my organization, NeuralAdex Limited, so we could test that with All Encompass, the top six, these particular six organizations, who is having the most citation dominance? And so for the last two and a bit months, I've been running that AI citation uh benchmark testing, and I will continue to do so. But again, while this is important, this is another area of knowledge that hasn't been presented to Google before. So, once again, this particular source, Neuraladex Limited, is providing new information for Google. So it's like, wow, these guys, in terms of GEO, they're just you know, they're providing golden nuggets. We're gonna keep going to that site and seeing what's happening there with their evolutions, because that's constantly enriching our knowledge graph. So that's the kind of explanation, or the best way I see to put it in real-world terms of knowledge graph enrichment. And I can say from that definitely that it has proved dividends, because if you type in search queries like generative engine optimization, excuse me, generative engine optimizations companies in the UK that provide real-world proof, we will come up as the number one surfaced um citation because obviously we're providing factual evidence of delivering that. So it's a it's a very, very powerful technique to employ for clients, and that's what I do do with my clients, and that's why specifically I like dealing with organizations that have that are service-based industries, because then I can do case studies and present that information to Google and enrich the knowledge graph alignment and then get that company highlighted to the uh Google algorithms, or I should say the AI engine algorithms.

SPEAKER_01 10:38

So you you are doing your own research, and like I've seen a ton of other people who have been testing out GEO, SEO, even. Hey, if you want, you know, more recognition within these engines, you need your own research. Now, if I'm sitting here and I'm listening to this conversation and I'm in an industry where research is common. So, like a CRM industry or something where they're doing like state of marketing or state of customer service, and I'm thinking, oh gosh, what new can I give? Would what would you suggest? Do it anyway and just have that out there so that the knowledge graph at least sees you doing something?

SPEAKER_00 11:20

Well, well, I think in essence, that the simplest thing to do and that makes the most sense is ask the knowledge graph. Because if you think about it, the the Google algorithm itself, it obviously has all the information up to this point in time. So if you have a particular strategy where you're thinking, okay, maybe for this client, I want to, you know, have them participate in knowledge graph enrichment. They're in, say, for example, the plumbing industry. Has there been um any case studies and evidence of a video showing a live installation and showing problems that they've overcome and then completing the job and then the customer being happy and images to back it up? So you've got different evidence forms. You've got videos, you've got images, you've got maybe a customer review of all that one specific job. Ask Google, has that been presented before? If it has, obviously it will search the case and reference it to you. If it hasn't, it won't be there. So then you realize there's a gap for new knowledge graph enrichment to take place. So that's that's effectively what I did. And so that's how I learned. So you're using the existing knowledge graph to ascertain what knowledge is not in that knowledge graph.

SPEAKER_01 12:42

What other source it's kind of like teaching in public. That's how I think about it when you're like record everything, record everything, and then just post it for the world to see you. That's how I'm thinking about it when you say that.

SPEAKER_00 12:56

But it's also very clever because it aligns with how the AI engines like the content to be because they want unique, original, trustworthy content. So when you are doing a video, you've got the face of the person, you've got the organic conversation exchange, you've got the footage of what's happening, then you have the transcript explaining everything, then you've got images, so you've got multimodal content as well that AI engines love. So, in every facet that they're trained to look for to give content priority, you're providing it. Plus, you've got the rich new data. So it's everything that you could possibly do to win in generative engine optimization, you are doing by following that practice.

SPEAKER_01 13:45

And then you just repurpose it for the channels that matter. So the blog, Reddit, X, wherever those engines are looking, just stick it on there.

SPEAKER_00 13:57

And the important thing to do when you do that is make sure you always do reattribution, because in generative engine optimization, that's a very important thing, and people can leave that out. They'll just do a post on a social media site and they think that's enough. But then how does the AI engine know where that information of that post derive from? You have to provide a primary source link back to your website. So then the AI engine can trail it back to your website and go, ah, it's this website that's provided the source of this content. I'll remember this website. It's it's got authority. If that link back to the primary sources, how can the AI engine know where it comes from? And also, AI engines will only quote content that derives from a website page. So if you're doing posts on social media and you haven't got a link back to the primary website page, you'll never be cited from your social media posts because there's not a core base of origin website that the AI engine can draw from. That's in its um directives, in its AI algorithm, too. So that's another massive key bit of information that has to be known because you're losing attribution. So, you know, such a plethora of knowledge points like this that are so, so important that are known in generative engine optimization, but not so much talked about in SEO. So um, yeah, let's bring some more.

SPEAKER_01 15:27

I'm curious because I know that LinkedIn posts are now surfacing within those answers. Like I was geeking out about that last week, and I don't know how they choose the title for like if you have a post on, I don't know, GEO optimization, there's one that was like GEO Trump something else. I'm like, I would never use the word Trumps in a title like that. I would never do that. That's a different tangent. But I'm curious if those links and posts that are showing, if they link back to my website, I I've not tracked it because I've I've not had a minute to do that. But have you noticed that that the post with the link are the ones that show up in those answers?

SPEAKER_00 16:09

I I haven't chased that up specifically, just because as you can imagine, with what we do, there's so many other thousands of things.

SPEAKER_01 16:15

So many, yes.

SPEAKER_00 16:17

What I do know is that's the core understanding of the root directives in the AI algorithms that only cite content that is linked back to a web page. They won't just cite it because it's on um a social media platform, because they need to make sure it's from a trusted source. Because you see what I mean. Somebody can just set up a company website page and post things. How do they know that that company is actually registered, that it's authentic, that it's not a criminal entity? The only way they can know that is by a primary link going back to the website, which in the schema markup declares the company is registered in the in the UK or whichever country it might be, it's registered. This is the company number, this is the VAT number, etc. Now the AI knows I have proof that this source is legitimate, I can safely trust it to a customer. If it hasn't got that trust factor, it will not cite it. Because it's again, so you know, that's how it's designed.

SPEAKER_01 17:24

So through your testing, like what are other ways that a company can increase their trust signals?

SPEAKER_00 17:31

Um, there's there's loads, there's absolutely loads. So if well it's kind of feeding into what we were talking about earlier, is the only way to answer your question, is going into the geo factors, and we work with 11, but we were saying we'll do uh different podcasts on that particular topic. But it would be covering, I'll just uh you know, give a synopsis of the main ones, which would be citation edition, um, statistic edition, quotation edition, fluency edition, easy to understand, uh, schema markup, authority, recency, source diversity, and so on and so forth. All key factors that were proven in the Princeton GEO study that I have found when implemented definitely enhance your ability to become the cited answer in AI platforms.

SPEAKER_01 18:28

So I want to one question I did want to ask you because I was looking at your article, How to Get Cited in Chat GPT this morning. I think you posted it this week, week of February 6th, maybe. Um, but you mentioned that recently you need to update your post every month. Some current recommendations that I've heard is every three to six months, but you're saying every month.

SPEAKER_00 18:51

No, it's every month. I mean, funny is you even say that because um there was a study that came out recently. Typically, while I'm talking to you, my mind can't remember the source, but it was a higher source that was showing that. Ah, that was it. It was from profound. Uh, I think it's like Joss Bliskel, I think his name is, I think that's his right name. And it was something like 56% of all citations derive from sources that had the recency in the last month. Then 70 percent 76% of all sources cited um had the recency of three months. So you see three-quarters of all citations are happening within the first two months of recency, and 56% being in the first month. So basically, what it's showing you is recency is huge, and again, it makes sense because the AI engines want to give you the most accurate, up-to-date information at the time. So, if they're quoting a source from a year ago, imagine how much can have changed in the industry in a year. Just look at how quickly AI evolves. So, therefore, their train, the most recent information is priority because that's going to be the most accurate. So, recency is a massive factor. And there's even been a I think a study with um I think it was HubSpot, where they updated their past blog posts from like a year, two years ago. And um, when they did that, their blog posts' organic um views went up 50% just because of the fact that they updated it with current information. So, again, just because of the recency bias that the AI engines have, that content got a lift in its organic views. So it's like in in a lot of ways, with the AI algorithms, as the more time I spend with it, it just makes sense, it's common sense. Like, what is the most intelligent way to serve an answer to a customer? Well, let's get the most up-to-date information. So, obviously, we're going to look at the date. When was that published? Obviously, they're going to look at trustworthiness, site authority, but they also want to get the most current. So it's just very logical, which again makes all the sense in the world because AI is intelligence, is logic. So it's kind of you know self-explanatory, really.

SPEAKER_01 21:28

Okay, so that was Paul Rowe. Now, before I let you go, I want to give you a few things you can actually do this week. So you walk away from this episode feeling like you've made progress on your GEO strategy, not like you just listen to two people nerd out about it for 15 or 20 minutes, because the whole point of the show is to help you get your strategy ready to go so that we're not left behind in how users find information now. So here are four actual takeaways from this entire conversation. Number one, ask the knowledge graph what's missing. So before you go out and create anything new, prompt Chat GPT, perplexity, and Google AI mode about your topic or industry. See what's out there already, see who's being cited, make a note of it, and the gaps that you find, that's your knowledge graph enrichment opportunity. That's where you get to be the original source. Number two, run a piece of original research or live test once a month. Now, this doesn't have to be a massive benchmark study like Paula is running. It can be your screen recording of you testing the same prompt across four engines. It can be a small comparison, whatever it is, it could be something else completely. But whatever it is, document it, publish it on your website with a transcript, and then repurpose that everywhere. You could do this with customer data or, you know, something. It doesn't have to be testing a prompt if you're not in marketing. I mean, it could be something else. I mean, you are in marketing, but you get what I mean. It doesn't have to be AI source related. Number three, when you do repurpose, link back to the primary source on your website. Paul and I talked about this, and yes, social posts can absolutely show up and answers on their own. I've been running an experiment with LinkedIn. I am seeing my LinkedIn post surface and citations, and Paul has seen something similar. But linking back to your site reinforces that trust trail. It tells the engines, hey, this content has a verified home, it lives here, this one's mine, it's a signal strengthener, not a gatekeeper. So do that anyway, but even if you forget, it's fine because it will still reference it. It's just always good to link back. Number four, audit one ranking page and actually update it. I know I talk about this one all the time, but this one is so, so important. I'm not talking about just going in there and changing a date and not just changing one sentence and calling it done. I mean a real update. So new data, new examples, a new section, fresh content, you know, the FSA framework. Paul and I both agreed on this. Recency is doing a lot of heavy lifting in these AI citations. More than almost anything else right now, I think I saw a stat that content decay happens in it was two different ones. So like 47 days, and then it like disappears, and then I saw another one that was like half of that. So content is decaying pretty quickly. And a profile study that Palm referenced where 56% of citations come from sources update within the last month. That means something. So pick one page this week and update it like you mean it. That's it. Four things. And they all follow the FSA framework or freshness structure and authority. If you are new here and are unsure what the FSA framework is, I will drag a link to some resources below in the show notes. Do those four things and you are already ahead of probably 90%. That's a guesstimate. Probably 90% of the brands out there still wondering if GEO is real. They are out there. I'm seeing it every day. But if you want help running an AI visibility audit before you start making changes, you know where to find me. Head over to Cassie Clarkmarketing.com. The link is in the show notes. All right, that's it for me. Until next time, stay visible.