Measuring 3rd Party Influence on AI Answers: Citation Optimization and the Future of SEO

How do brands actually appear inside AI answers?

In this episode hosted by Search Engine Land, Garrett French and James Wirth from Citation Labs explore how generative search is changing the way marketers think about visibility, rankings, and attribution. Instead of focusing only on traditional SEO metrics, organizations must now understand how AI answers are generated and how third-party content influences which brands get cited and recommended.

The discussion introduces the concept of citation optimization, a strategy focused on ensuring that brands are present in the citation sets used to construct AI answers. Because generative systems retrieve information from multiple sources across the web, influencing AI answers requires both onsite and offsite visibility.

Garrett French explains that if a brand is not included in the citation set used during the retrieval phase, it is unlikely to appear in AI answers. This means marketers must think beyond rankings and begin measuring AI Answers directly by analyzing prompts, query fanouts, citations, and recommendation visibility.

Throughout the conversation, the Citation Labs team shares early experimental findings showing that third-party comparison content cited in AI answers correlates with higher recommendation visibility for brands. These findings suggest that citation optimization may play an increasingly important role in influencing AI answers as generative search continues to evolve.

The episode also introduces several strategic frameworks used by Citation Labs, including decision efficiency, comparative matrices, and structured content designed to reduce cognitive effort for users and AI systems alike. These approaches help brands present clear value propositions that are easier for AI systems to retrieve and reference when generating AI answers.

Listeners will also learn why measuring AI Answers requires a new set of metrics. Instead of relying solely on traditional rankings or click data, marketers must track prompt visibility, recommendation positioning, and citation patterns to understand how often their brand appears in generative responses.

Additional topics include:

• How AI retrieval systems gather information from multiple sources
 • Why third-party citations influence which brands appear in AI answers
• How citation optimization differs from traditional link building
• The importance of comparison content and decision matrices
• How marketers can begin measuring AI Answers and tracking visibility
• Why influencing AI answers requires both onsite and offsite content strategies
• The role of recommendation rank in generative search visibility
• How Citation Labs is testing new approaches to citation optimization

For SEO professionals navigating the rise of generative search, this conversation with Garrett French offers practical insight into how AI answers are formed and how brands can begin influencing AI answers through smarter content structures and strategic citations.

As generative search continues to evolve, understanding citation optimization, measuring AI Answers, and the role of third-party validation will become essential for maintaining visibility across the modern search landscape.

Featuring:
Garrett French – CEO, Citation Labs
James Wirth – Managing Director, Product & Growth, Citation Labs

Key Topics Discussed

citation optimization
 AI answers
 influencing AI answers
 measuring AI Answers
 Garrett French
 Citation Labs

Transcript

SPEAKER_00 0:02

Welcome and thank you for joining today's webinar, Measuring AI Answers, How Third Party Cites Influence Who Gets Cited, presented by Citation Labs and Search Engine Land. I'm Danny Goodwin, editorial director of Search Engine Land and SMX. Before we get started, I have a couple viewing tips for you. If you have any audio issues, you can click the audio icon on your screen. If you have any viewing issues, you can use the QA section to communicate with us. And you can send any questions or comments directly to the speakers about their presentation at any time. Now it's time to introduce the speakers. Joining us today from Citation Labs are CEO Garrett French and Managing Director, Product and Growth, James Worth. Welcome, Garrett and James. Thanks so much for joining us. I'm going to turn things over to you. Thank you, Danny. Thanks, Danny.

SPEAKER_02 0:56

Okay, welcome everybody. We're excited to be here today. We're going to do our best to earn the labs part of our name, citation labs, by sharing some things we've been experimenting experimenting deeply with and hopefully give you some guidance around all these crazy paradigm shifts that we're all experiencing together. So Garrett, do you want to say anything before we kind of jump into the presentation? Do you want to talk for a minute?

SPEAKER_01 1:19

I would be concerned about keeping it to a minute, but I am deeply excited to be presenting today. We are, you know, here at the beginning of really trying to understand how do we represent impact or change that we're seeing in a way that's meaningful in our in our at the end of a Q1 review, right? So what's been happening? It's at the end of the month here. So we're at the end of Q1. So what the heck has been going on when it comes to work in um in in AI and AI visibility? What are we doing? Uh so really that's been our lens all along is kind of having something to show, right? Not just a for us in link building, not just a link built, but what impact did we have? And as we veer into uh what we're calling citation optimization work, uh how do we show what has happened and you know claim some some you know uh relationship between our work and and and what we see uh occurring. So that's what we're here for today. We're we're grateful to have everyone here. We are you know in active practice, we're we're we're in in work, we're a work in progress, working in progress, um, and and we look forward to sharing what what we're seeing.

SPEAKER_02 2:55

Yeah. Okay, great. All right, so let's jump in. This is gonna be a great uh discussion today with Garrett and me, and we hope hope we get some great uh great uh questions coming through. So please send those questions. But let's just kind of you know level set here. This this may be, you know, call me Captain Obvious after this, but the old SEO playbook really is just gone. Um actually, I shouldn't say gone. It has shifted very dramatically. It almost feels like we're in the industry of realigning to the moving goalposts. That's our job in you know, so many different ways, because things do shift all the time. And uh I uh depending on how long we've been around, I think back to some of the really big what felt like earthquakes to me early on in my SEO career, you know, the the the penguin and the panda update, those earlier era um sort of big shifts that happened and all of the other ones that came along with it. And each one of those felt like, wow, this is a huge, huge shift. And then we would get another one and it would feel like this is a huge, huge shift. But this one, this shift, this feels this feels at a level that I just had not experienced at any point. It may be bigger than all of the other ones combined. So, you know, let's talk about that and just kind of dig into that a little bit and level set on some of those changes. And so they're listed right here. Um, you know, AI AI answers start with retrieval. So if you're it's just gotta be, if you got to be in it to win it, you know, from that perspective, if you're not in the citation set, you're very unlikely to get recommended, included, et cetera. So we really are talking about kind of that baseline of getting into the citation set. And um, you know, I promise that I wouldn't use the F word, but I'm gonna use it one time here, fan out. It's all about the query fanuts. So, you know, we're not gonna cover a lot of that really deeply in here, but I just wanted to get my one F-word in or two. I guess I snuck an extra one in. And then, you know, talk about how uh AI systems are retrieving that content are retrieving content not by just serving up a list of pages, of course, but by individual um contextual elements of a page, often referred to as chunks. We call them echo blocks because you want to echo that content across the web, especially in the AI answers. And if your um if your content is buried, uh your main selling points are buried in a mountain of content, then the the AI is likely to not retrieve it. Um and they're gonna be looking through a crow across a lot of different um uh sources of information as well, not just individual um uh provider websites. You're providing that product or that service if that's your offering. And uh and then of course we know, much to the chagrin of all of us, most of those, many of those uh those prompts, those, those, I'll say it again, fan outs, those uh you you know, those opportunities to earn a click are not involving a click. And so what do we do about that? How do we then even track things? Uh, you know, it's an it's a new era for sure. And we've been experimenting like madmen and women in the company. Um, it's been an amazing kind of experience of learning all this and really digging in. We've built tools, um, we've got all these initiatives going on. So I'm really excited to be able to share, you know, more of that with you. But the key is we've got to be able to, we still have to be able to track, even if we're not earning the clicks, we still have to be able to track. So the biggest part of our discussion here, the recurring theme, if you will, is what the heck do we do? As Garrett mentioned, we're gonna be at the end of Q1 at the end of this month. Um, how do we how do we report uh uh activity? How do we we report visibility? What does that look like now? And that really comes down to prompt level tracking and I'll sneak sneak it in one more time, fan-out level understanding of what's happening when a prompt is given and the AI system grounds itself in that uh in that prompt in order to generate an answer. So we'll be weaving that throughout the um the process, essentially. And a big reason for that is because you know, there's there's some skin in the game from our perspective. This is this is self-serving to some degree for us, but it I imagine it is all for you. It's we need to defend the spend, whether that's our job or our client relationship. We've got to be able to speak to the purse strings, the higher ups, the ones making the decision about whether we keep an investment or end it. Uh, how are we providing value? And that's shifted so much that it's challenging to provide that proof. But we're gonna do our best to give you some guidance on where we can go to deliver that, you know, essentially. Um things have just changed so much. And uh and so really there are kind of two components here that are important to double-click on. One is uh is how you're structuring your on-site and off-site uh assets. Uh if you can think about them as surfaces, you can think about them as as uh content, as pages, as documents, et cetera. Where content is being served, how is that content being structured? So that's sort of number one, and we're gonna talk a lot about that. Decision assets and citation optimization and a number of other things. Really excited to get to Garrett's slides. Um, but also that so that's one aspect of it is how is we're are we how are we structuring our content on and off the site? And then the other piece of that is how are we reporting on the visibility around that content? And you know, there are a lot of conversations. We're in them about whether rank is still something that should be tracked. That's certainly um an ongoing challenge, especially as the goalposts continue to be moved and are just very active right now in terms of where we have no idea where they're gonna land. So um, how do we how do we make uh good on all of this effort, the investment, our time at working for a company with these goalposts constantly moving? What does that look like? And how can we, to some degree at least future-proof that content and then, or that effort? And then how do we track it? How are we looking at competitors? How are we looking at our own visibility? What metrics do we use? That's a really important one. Um, how do we measure that over time? Are we even measuring the right things? Are we looking at things through the right lens? These are all questions that I'm suggesting as questions now because I'm constantly trying to answer them. And we're uh we're constantly looking at uh whether we can answer them or not. And then can we can we influence them? Can we have an impact on them? So clear reporting is critical, um, especially when you're having those conversations about defending the spend and certainly keeping our jobs and everyone's responsibilities are shifting, what does that look like? We've got some really interesting things to share. And so um so this is kind of the baseline for us is essentially this was our hypothesis. Uh, if we publish a comparison asset, and we'll kind of talk about what that means uh you know as we go, but if we publish a comparison asset that matches what these AI systems are looking for, are grounding themselves in, are citing in their answers, uh, can we influence that list of providers that they're gonna serve up at some point during that conversation? Might be a couple of threads down, but at some point they're gonna surface as that customer journey continues and gets closer to a decision, they're gonna surface a list of providers that could be that can sell you the product or provide you the service or have that offering. Can we influence that? Can we one get a brand into that uh that list? And two, can we move them up in that list? Can we do that? That that was the hypothesis was that if we publish content that aligns with what AI systems are looking to ground themselves in, mean they're gonna find it, retrieve it, and use it in generating their answer, then we can influence uh that list. And so we have been building these off-site comparison pages with uh a number of features and elements that have been surfaced through our own research and through data that's been released, case studies that have been released that will influence or are like at least correlate to a higher level of citations in AI answers. And so we've been working on what those look like, how can we improve the uh the frequency of being cited, the recurrence of being cited, as well as uh can we influence where a subject page or domain uh or offering is showing up in that list. And the and the not to bury the lead here, so I don't bury the lead, the the result here is that when these assets are cited, um the subject page or the subject client's performance has outperformed when they're not cited. So we've got some early but substantial, statistically significant, um high confidence, reasonable sample size results sort of thing. And you can kind of see the stats up here on the screen in terms of what's behind the graph that we're seeing right here. And so we're not gonna just throw a bunch of graphs. You never know if they're actually accurate or not, or real. They are. This is this is actual work that we've done and exciting work because while we can't necessarily claim causation, we at the very least see a very relevant, very direct, very high correlation between having comparison content cited in results and a brand increasing their rank in the list of offered uh services that the AI answer provides. And so essentially just to just to unpack this a little bit, what you're seeing on the screen is this is rank. You can see the vertical axis on the left, that is that is the position of a brand is showing up in that list of uh results. And the two lines represent the orange being the instances when our comparison content was cited in that AI answer versus the blue line, which is when they were not and while the numbers are relatively close together at the volume that this is at and at the competition level that this is at, this is uh this is highly relevant to the brand, and they were over the moon about the results. And uh, this is just one study. So we've done a number of these for a number of clients already, and we're seeing similar results. And we thought it would be worth sharing. You know, essentially, we've got a tool that helps to do this, uh, to be able to track this, um, and we've got an approach that is is working. And uh so if you have questions about that, please please put them in the chat. Uh, I don't expect this to necessarily be earth-shattering to everyone, but it certainly is to our clients, and we're very excited to be able to see those sort of results, be uh be able to acknowledge that our hypothesis was correct, at least from a corollary perspective, if not a causal uh relationship there, and we're just over the moon about it. So at this point, I'm gonna uh pull Garrett in so that we can talk more about you know essentially what this means when brands are trying to stay visible in AI answers, and he's got some uh some great content to share. Do you want to say anything before we jump into the slides, Garrett?

SPEAKER_01 14:32

For sure. Um we're these are early stages, these are you know, wouldn't I wouldn't say that we've um changed the training data. Uh this is not like we've we've affected the core beliefs or core entity scores or what have you uh for these. This is when search is invoked. Yeah, yeah. So this is just just to again level set. And so but but but from this from this, you know, this perspective of having positive impact to be able to talk about uh in in any quarter, uh it's it's definitely been meaningful for us to uh be able to bring this as an example of of of impact. Um we we do come from link building, right? Like I've you know, we did uh citation labs was is, was, will probably always be a link building agency. Um and and we really uh learn we we we found our way forward with with enterprise as we were demonstrating uh link impact and being able to say, hey, when we build these links, here's the impact that we see. And so that kind of uh ethos really has been guiding us is how do we what kinds of constraints can we put around our what we're measuring, what we're seeing, so that we're able to have a small enough um uh kind of sample, not sample size, but a small enough pot of water to get it boiling, I guess, so to speak. Really be able to show what you know how how hot we need to get it to get it boiling. But yeah, it's been absolutely um thrilling to be able to see um uh make you know thrilling to see the early evidence of of impact. Um so we I I I kind of teased earlier, we're we're trying to brand, just walk with me here, folks. We're trying to brand this notion of citation optimization. Um and so there's a lot of you know what well what what what are we talking about? What citation? Um do we mean on-site, off-site, uh cited domains, cited pages? And so the answer is yes, all of those things. We're we're we we've we've uh expanded our kind of our umbrella of of examination and kind of what we feel like we need to have a say in and uh and be a party to impacting um based on uh this notion of well what is going to be uh cited. Now there doesn't I there there's a lot of this I'm gonna I'm going sideways just slightly, James. If I go too long, pull me back. But um there is thank you. Um there is discussion and I think important discussion around uh a lot of times a LLM any given surf uh any given like kind of model will have already made up its mind and it's looking for something off page or out of its current index to kind of justify its recommendation or or what it's saying.

SPEAKER_02 17:49

And so like you know, there is there is question, there is there is uncertainty here, but we're but it wants to validate its its descriptors, it wants to have a high confidence in what it already uh assumes or has inferred or implied, and it's gonna look for those signals off off site.

SPEAKER_01 18:07

The other key piece that we've been dealing with, and this is a significant place, is the notion of like because because we'll say you know, citation optimization, and people will think, oh, if I get my my domain cited more, I'm gonna do better, I'll I'll I might appear better. And that's not necessarily the case either, because a mention of so so we're we're we're differentiating between a citation, which is you know, an actual like link or which we're not expecting anyone to click, no, but uh you know, an actual link within the responses, and then like a brand mention within the responses as well, but then more to the point, an actual recommendation, or at least not a don't use this one, or a a warning. So we we've we've experienced and ex and and seen cases where uh being mentioned is is is um is unpleasant. Um so so we're really trying to shift into I don't like to say James said the F word, I'm about to say the F word of sentiment, okay? Because it really is like a little bit of a uh it's it's very slippery to say another S-word, but um so but anyhow, uh there's there's a there's a huge difference, and I think a really important distinction to be made between um a citation, whether it's your domain, it's a competitor domain, a third-party domain, and then which page specifically are we talking about being cited, and then the brand mention within the actual response, and then the sentiment around the mention. And I'm preaching to the choir here. Thank you uh for being with me uh through through that, through my sermon. But the the the point is that we are really working hard to advance this language around what are we what are we talking about, what are we doing, what are we trying to impact. So the real goal for us is like how do we get those those recommendations and those meant uh recommended uh mentions. Um so uh to do this, we're looking at um on on site, you know, we are looking at on client site, you know, what pages are getting cited. Uh is there more evidence we can provide here that's going to um be in a format that's uh that first of all have information that's consistent across um across pages, across the the the brand or the um the web's understanding of of what this entity is or what this this offering is. Um then off site uh and then you know we're really looking at at trying to get in volume. We're looking at volume. I'm just gonna say it. Uh now we we believe, but haven't proven yet that this is going to be a pathway to um uh impacting uh the the the training data as well. That's uh that's our hope um ultimately. But we do we have been working a lot at what are we talking about, what is citation optimization, and we uh uh invite you into that that that quest with us, that search. We're we're we're defining it, we're we're doing the work or that you know, we're we're evolving link building for our in our own practice into this this place of uh what we're calling citation optimization. James, if you had advanced us, sir. Thank you. Uh one of my favorite topics is decision efficiency. How much time do we have left for this conversation? Uh 30 minutes. Okay, okay. Oh my gosh. Stay on this slide for 40 minutes. No, um this this place is is quintessential from we've been for years uh you know telling people hey, you need to have uh help. Helpful high quality content on your website. And we've we've been pushing people to have helpful, high quality content on their sales pages, right? That really genuinely helps people inform the decision. And what we're also seeing is that, you know, very particularly to the slides that James shared earlier, when we're when information is provided in a comparative matrix, you know, uh think affiliate site, right? Uh we have distilled a number of decisioning factors down uh to a point where it's it's more it's faster to make a decision, to try and select the right one. Now we know if we've been in the space for any, or I mean we all know what what affiliate pages can be and and uh that that data can mislead and and structures are can can uh inform a choice that might not be optimal for the human making the choice. But we uh fundamentally believe and lean into the belief that uh enabling the best possible decisions will always result in the best possible content. So really trying to enable a human in that moment of decision to make the best possible decision for themselves and putting that on your website, off your website, is particular with particular to the decis the purchase decision, right? You are going to worst case scenario win because you've established more trust with the person visiting your page. Um so decision efficiency is something I've worked a long time around and built up formulas that I you know built around how do you genuinely enable a human and sometimes a group of humans, right? Like we're gonna have uh different stakeholders making a large B2B uh decision. How do we enable these humans to to understand each other's perspectives? What are the grids and and for um the graphs that they're using to uh decide, and then how where's the overlap, and then how are they arriving at the best possible decision for themselves? Um so we believe that's possible and that the uh really standing up content that allows particularly allows people, you know, the decision makers to see into kind of the potentially negative ramifications of doing something, buying something, who who is this not a fit for, uh is is uncomfortable for us as marketers to say, but it ultimately uh assists the decision process. It makes it a faster no, you're enabling a faster no. And anytime you're you're trying to enable someone, enable a person to do that, um I think you're helping LLMs as well, right? Like so you're helping them inform them uh or their models uh for for uh around decision efficiency because what are they trying to do, right? Like what they're they're trying to help humans make faster, smarter choices. I think, I hope. I very much hope. I I let's just say I believe. Um so so decision efficiency though has been at the at the core of of the the the kind of the ethos or how how we're trying to approach content uh on behalf of our clients and and um and certainly when you know we we we we see it we see that um helping make fast decisions with with comparative structures is rewarded by LLMs, and we saw that uh in the in the slides that James already shared with us. James, am I leaving anything out there? Or did I is there anything I could like?

SPEAKER_02 26:14

I think one element yeah, we might want to include here is just a little bit more about the about one of the levers that AI systems have, which are these descriptors. And so, you know, they over the course of prompts and uh what they're getting from humans and how they're grounding themselves in search results when they run these queries, you know, first of all, it's the intent classification and then it's the query transformation. They decide, okay, what are we gonna do with this? We're gonna are we gonna decompose it down to these, you know, strip out all the stop words and anything extraneous and just run these five or ten queries and get results back? That's the fan out. And then, you know, what do we do with that information? Does it confirm what we already know that's in the training set? Is there net new information that we want to factor in? And can we get signals about this? Can we establish or challenge the descriptors we've already assigned to that entity or that brand? And that whole process is designed, at least in part, to around this decision efficiency. Can I make this easier for this human to make the decision that they're discussing with me and how do I do that? And there these descriptors come into play there. And so one of the things that we're influencing when we talk about the, you know, the essentially the narrative out there around a brand, especially off the page, off of our perfectly sculpted uh sales landing page where we have no negative information, we don't want to send them anywhere else on the site, we just want them to convert. You know, the challenge with that is that the internet out there is already influencing perception. And so what Garrett's talking about here is essentially participating in that conversation instead of reacting to it. And that's where the real magic of this decision in efficiency is. And it does involve those descriptor descriptors and and um influencing the narrative around them.

SPEAKER_01 28:03

Thank you, James. And just to belabor one last time and we'll move on to the next slide. Uh this is that this is literally one of my favorite things to talk about. But um, we really are trying to ground this in like uh from from a cognitive effort perspective. We're trying to reduce the cognitive effort of the person in the moment deciding kind of what to do. How do I how do I make the best possible choice for myself? So we're de-risking, how do we de-risk their their whether they yes or whether they no? Um and then and then the in concert with a larger with with a decisioning committee, how do we de-risk the whole uh committee's kind of decision, whether that benefits our our brand or not, or whether we get a sale or not necessarily. Um so we we do really uh I and and it really this has come about from I got sick of hearing make useful content, right? Make helpful content. Like, well, what do you mean? What does that mean? And when we when we really aim at this place, I I I feel like we can take a clearer, uh, more certain shot at enabling uh or content that's genuinely helpful in in the in that particular moment. We maybe should have said at the very beginning, James, we have said that we hyper focused on the bottom of the funnel.

SPEAKER_02 29:23

I mean, should that have been I don't think that we said that, but we probably should say that we work from the bottom up, yes. Oh gosh. Yes, that's that's true. And if you followed, if you follow Garrett for any period of time, you you'll recognize some themes in here. You know, what came before before he was talking about he kind of boiled it down to decision efficiency. It was, you know, a number of other things that tie right into this. The purchase decision committee, we've written volume Garrett's written volumes of that on the website and other sites. It's the, you know, it's the flux, the frequent, the the excuse me, the friction-inducing latent unasked questions, which evolved out of the frequently unasked questions or the fuchs, which didn't quite paint quite enough of a picture. And so this is this uh evolutionary process, as Garrett has iterated on this, to get to the point where we're really talking about reducing those frictions from the perspective of making a decision easier. And that's how we're that's how we're shifting the narrative inside the AI systems responses. Thank you, James.

SPEAKER_01 30:18

Um we're still, I think, okay, two points. So so this is this slide is great. This is kind of illustrating or or helping see into the the kinds of the areas that can for our audience um cause a friction or a slowdown or a freeze in decision and consternation. Yeah, yeah, for sure. So so we do so so one of the places where we've landed, especially in in the work with in comparative assets, are the axis of decisions, like what are the specific columns in the in the in the table, in the comparative table. Because these are the axis of decision, right? It from from in our model. And so audience is going to align or um see or or appreciate these these various uh um kinds of uh table headers, so to speak. And then very specifically, we're we're looking to one of one of the places we've we've been leaning uh more recently is is a place we call the axis of advantage, which is just hey, if we add, like we take we look at all the all the comparative tables that are currently being used at the bottom of the funnel, we look at what are the what are they mostly comparing one to each other, and then uh uh what is a unique table or unique rather uh column or set of columns that uh the client or is is uniquely uh valuable. Does that mean we're shifting the audience, we're shrinking the audience size? Uh, but when we when we you know, where do they really shine and how do we make sure that that is a a column within the comparative uh um kind of universe, right? So so then it becomes an issue not just of like, hey, how do we improve our visibility or how do we improve our rank? We're like, okay, well, how do we continue to assert that this unique value uh proposition of ours is genuinely valuable, right? Like, how do we put data into that column that proves that this is a valuable column to include in your decisioning matrix, right? So this is you know, this is what we this is part of our work. This is part of the work we do, is is arriving at that place. And I my my humble submission to this, you know, everybody listening today is hey, this is the new, this is the anchor text. This is what we're really trying to uh say is important and say we want to be visible for. Now, does that mean there's demand around it? No, there's not people searching for your unique value proposition necessarily, but you certainly need to be able to express uh what your unique your your unique value in a in a in a data supported manner, and then continue to assert why that column should be included in everyone's uh comparative matrix. Again, we're we're bofu centric. You know, we have some we have some tofu stuff though. Is that our next slide? Jeez, what's our next slide? Yes, tofu. Tofu, we got some tofu for you guys. So this came about because we were asked to do some high-level um like what do we call um they we had a client that wanted to know what should they write about on their blog that their competitors are not writing about, right? And so for us, like our focus has been like how do we, you know, we we really focus in that place of how do we show that our work is having impact in visibility, right? Like in in um at the prompt uh tracking level or the the prompt evaluation level. Uh when we're looking at 400 prompts that are you know kind of isolated on this on this one audience set or this one problem space. Um so the O20 model um really uh we we arrived at this place because we needed a way to connect higher level topics with the granularity of our this is a a client that has has software uh SAS product, uh higher level topic with the granularity value granular value of let's just say a single button within the the the product itself, right? Like uh yeah, you can, you know, topically you can achieve X with this. You can uh improve your customer response rates, but and and y and it's a it's a blank claim. And so we can have a topic that's you know, uh we could write, we could say, hey, you know, competitor X has written about this, you guys haven't, so how do we um here here here's what you need to say. Here's kind of a outline for the for the copy, but to really ground and bind this topic to your actual offering, the thing you sell, it was uh we we said we we we arrived at a place well first of all we took all of their help documentation so we kind of map out the the the UI itself, right? And then be able to uh show from this larger topic perspective the granular buttons and how they connect to the the the broader uh outcomes that people were trying to achieve in the in the topic. Um we haven't shown that this we don't have any results from this. This is this is uh CBD in terms of what impact have we seen from this. But this was a very exciting place to arrive because it it gave us, it forced us into the top of the funnel, first of all. Um not you know, not normal uh place for us to to really have to work and and solve for, right? But it gave us something we could then track, right, from a prompt perspective. Are we beginning to see breadcrumbs for how to achieve things with our uh with our with our client's software in LLMs? Are LLMs able to give clear instructions to people on what to do, which buttons to press, where to find them as it relates to topic, right? Like um, so very exciting, but also again, this is the labs, this is the labs talking here. We haven't we haven't uh uh built into that or or or um our instrumentation hasn't shined its light there yet. Uh what else do we have, James? Yeah is that our is that our last one?

SPEAKER_02 37:05

No, no, no, no.

SPEAKER_01 37:06

Oh no. This this looked like a James slide to me. I don't this is like guys, this is like I'm uh I love the the theory, I love testing, but then really being able to prove something, I'm I'm always gonna default to to James.

SPEAKER_02 37:23

Well, you mentioned it already though. You you teased this topic out about you know recu recommendation rank when you were talking before about the different the different ways that we can gain visibility inside an AI an AI answer. And so you can have, of course, the citation, um uh that you can be included, you can get a mention, you can inform the answer itself because of content. And and that's that's where what we're really looking at with that test that that Garrett was talking about, where it's higher in the funnel. Can we inform the conversation um in a way that that you know essentially um shifts the shifts the narrative toward the offering, tying back even earlier in the presentation, to the unique value points of that particular brand. And so that's that key aspect of this is where higher level content can come into play. One challenge is you know that that when the query fan out runs, I I said I was only gonna say it one time, but that's like my mystery.

SPEAKER_01 38:24

Well, you said it you said query fan out, not just fan out.

SPEAKER_02 38:26

So you I changed it, yeah. Um so when that what you know that when that runs, the the AI system is gonna look for a variety of responses. And so this is where those multiple surfaces come in as well, and how they can influence recommendation rank, circling back to that graph that I showed where we had the two different lines, and that was the time series graph, and the the line that was performing better was the one that had third, this is third-party content that we had placed, not content that we built on the brand's website, but it being present, it being cited, it's informing the discussion and it's increasing the recommendation rank of the client, of the subject uh page or or domain or brand. And so that's a critical aspect of all of this is you know, we can't we can't just go for rank anymore. First of all, just kind of let to level set for a minute. Um, you know, of course we want to see how we're doing in a prompt. Are we getting visibility in a prompt? But in terms of ranking, we don't, it's not so much about ranking for the prompts. I hope everybody is you know up to speed on this, but if not, I just want to mention it just in case. It's not so much about ranking for the prompt itself. It's more like we want to track visibility in that prompt, but in terms of, with the exception of recommend recommendation rank, but in terms of thinking about it from a traditional SEO perspective, what we want to rank for are those query fan outs. And it may or may not be on our website because as I mentioned, the AI system is going to look for a variety of sources and surfaces. And so it's about ranking for those query fan outs, not necessarily for the prompt itself. In other words, the we get into the query fan out, that gets us hopefully cited, included, mentioned, recommendated recommended in the prompt. But what we're doing to get in there is to rank for in traditional SEO, we're good to rank for that query fanatic because that's what it does. It try it classifies it, transforms it, decides what's it's gonna do with that, decomposes or infers or adds additional information. Um we it looks at all the signals in that prompt and it runs those query fan outs that you can look in your Google Search Console. It's the ones where you rank well for their super long tail, they're very, very well structured and and uh and organized, and you get zero clicks for them. Those are likely your query fan outs. And so dig into Search Console because you can find those. You gotta do some digging. Um but you can find those, and that will give you a sense for where your visibility is coming from now. Does that align with your offering? If it does, great, do more of it. If it doesn't, now you're that's your opportunity for net new content on your surface as well as third-party surfaces that will in that will inform and contribute to that narrative. You've got to be in it to win it. So this is how you get in it is you rank for those query fan-outs. That's what gets you in the grounding layer that the AI system does in order to include you in the generative layer when it builds that AI answer.

SPEAKER_01 41:31

And I'm pretty sure we're only pulling those right now from ChatGPT, but those are I my understanding from my dev team is those are implicit in or explicit in the uh what we get back from. I don't I'm already out of school here.

SPEAKER_02 41:48

But we're we're not necessarily servicing imminent tool. Right. Yeah, yeah, by the way. It's on the it's on the slide, but if you want to go you know look now, it's XOFU.com, Zofu.com. Um you can run a snapshot. Um, it will simulate the queries for you to give you a sense for where your visibility is based on the content on the page that you put in. So it's just a good starting point, you know, there to start to dig into this and better understand it. There are a lot of tools out there, AI measurement tools. As Garrett mentioned, we're from the ground up. So we're starting with you know, put your sales landing page in there, see what kind of visibility it has. If you have content that's parked in different places on your site and it's not on your sales landing page, uh, if you've been following Garrett for any period of time, you're he'll you'll know the phrase money page. If it's not on your money page, it's likely to not be included in the AI answer. So we can think about content from that perspective as well, because they're not gonna crawl your entire site just because you showed up for one query fan out. Um, we got to think about it from that perspective too. The AI system wants to be efficient also. And so, where is it going to go for the grounding layer in its answers? Um, so uh just think about it from that perspective. But this is what we're talking about, some of the elements that we're using to be able to track AI recommendations.

SPEAKER_01 43:03

And I would, I would, I would have to also just completely be transparent here. And we I don't we really in in our in our tooling, we we're manually identifying recommendation, right? We don't have that baked fully into the tool itself yet. And but so so but we're we are able to express it and see it uh and and and show it in a graph form, but that still is a moving target for it's evolving, yeah, yeah, yeah. So yeah, that's the the recommendation, it's just a the a reminder for us as as marketers um that it you know being mentioned in an AI response is isn't always great, isn't the best thing that can happen. Um, so uh like we don't want to be number one for worst citation optimization, right? You don't want to show up for that, but there, you know, what if we

SPEAKER_02 44:00

were so that that's a that's a place that you know you you genuinely are not winning even if you get mentioned right so should you count it well we want uh we want to be able to earn the recommendation and that isn't just from off-site like things you can do off site or because obviously like if you've got you know thousands of comments and Reddit and like whatever G2 or all the different types of surfaces where uh kind of general opinions are formed uh you're you're going you you you you're gonna have a lot of work on your hands um for for for for optimization purposes but yeah can I can I say one more thing right here Garrett please um one one thing that I also hopefully everybody is is uh has this understanding but just in case um just in case this hasn't kind of clicked there is a lot of discussion on LinkedIn and elsewhere about Google versus AI and I just want to plant the seed or maybe water the seed if it's already been if it's already germinating in there that Google is AI. So it's a it's a it's a it's it's a it's a strange comparison to think about. And and this is important when you're when you're thinking about you know laddering up these conversations to the higher ups um from the perspective of um uh when when your CMO for example or whoever you report into says oh well you know chat GPT has such a low response um uh or it's uh it's such a low amount of our visibility we don't need to think about we don't need to think about AI. Well we want to you know just approach that from the perspective of Google is also AI too. So um just keep that in mind as we're as we're kind of wrapping things up here and I want to make sure we have plenty of time for questions because we've had some questions come in Garrett. So do we want to um just we got like two slides left and this is kind of the main last uh meaty slide here do we want to talk about uh talk about this? Yeah uh um this is this is our our denouement you know if you will uh if I can borrow one of my favorite LLM words uh but this is um uh this is our this is what happens when you uh publish comparison assets and they get cited and the comparison asset uh favors your brands um your brand recommendation uh increases you are recommended uh higher than your competitors are i is that rank I mean I don't know what it's like I don't know another thing to call it I don't know another thing to call it no we have to tra we have to track rank it's yeah you cannot rely on running what am I supposed to do like you're number you you were number two and uh now you're number one or whatever you're moving up in visibility percentage or you know okay but it's it's it how at what point are you recommended across you know a hundred queries that are uh not overlapping and from different kinds of perspectives and informing different aspects of the decision uh different aspects of the problem uh of that that people are likely to be solving at the bottom of the funnel as they're trying to assemble enough information to make a great decision for themselves uh uh how how often are you recommended there?

SPEAKER_00 47:17

I mean it's uh well we don't have to call it rank we don't tell us what you think we should call it obviously yeah and do you do you have the visibility you want whatever that new metric is which still the you know the the industry is the jury is still out for sure but everybody's got an opinion on it um whatever that is that's what we're trying to improve let's get to questions we're gonna talk like okay i i can tell i'm i'm getting i'm ready to rip now but we better we better move on to questions so i don't just keep talking yes let's let's let's make sure we get them answered all right uh great presentation as always from you guys uh we will dive right into the questions uh there was one from may here uh about off site or off domain what domains are you referring to others you own or sites like forbes yeah those are yes both of those so forbs is like usually um the front and center most frequently cited domain at the bottom of the funnel for large um you know consumer categories insurance uh loans uh business related things uh oftentimes um but i would also call out that when you are uh uh tracking prompts that are more granular or from an expert perspective uh you will find that there are different um cited sources so it isn't gonna be enough to simply call it if you're if you're not in Forbes though and and your competitors are get in Forbes yes by all means so we're not gonna help you with that you're gonna have to go knock on those doors that's not a usually a citation labs type of problem we're really saying what can we as an agency address um at at some larger scale and you're gonna still need to uh you know knock on your regular doors and go because that's an affiliate play I mean that's that you're gonna have to hand that over to your affiliate bring your book yeah get your bring your bring your money with you but yeah we're absolutely talking about modeling content in in certain verticals we're not saying this is a panacea um hope I pronounced that right never heard it said uh but but panacea I think yeah I have heard it said it's motto it doesn't matter it doesn't matter um but but but my point is uh you you you you will need to publish uh we we we have published content that uh is in the format of an affiliate site and we owned it it was it was our our we we we owned that um so uh that that's what we saw get get uh drawn into the answers yeah it's kind of like an all of the above too start with the prompts get your prompts dialed in figure out who has the visibility look at the citations that are helping them get there at least a correlation to the citations that they have and the visibility that they have and then see if you can also get included.

SPEAKER_02 50:15

So it's it's definitely an all of the above for sure. Build your trust build your authority get cited on content that's already being cited in the prompt in the AI answers and then fill in the gaps around that. All right uh we'll move on to another question from Emily have you both found if there is a specific number of citations that is needed for your brand to be mentioned in a fan out oh no I can't answer no you know like the trillion dollar question yeah no there's no like fundamental like number we James has yeah don't say it James I could see you the well the one consideration is let's use the data that we have we can't there there's not a there's not a uh I wish we could say exactly you need X number of links but you know all of the factors that we've always dealt with around qualitatively around content or backlinks et cetera et cetera all that is in in play all the all the many algorithmic factors and subfactors all that you know factors into some degree haha factors into some degree um and so but what we do have is a platform that we've built to help us dig into that primarily built for ourselves and so that we can start with the prompts be able to understand the fan outs that are being run are we showing up in them yes or no who is showing up in them who's getting the visibility in these in these prompts and start to look at their citation sets. You know who else is showing up in those same prompts that is benefiting in terms of recommendation rank like we were showing that is benefiting those other brands that we're competing with and want to um to outpace. So we can get a sense quantitatively at least to start looking through that lens that you're mentioning there in terms of who is showing up at the top of that visibility list and what does their citation set look like. That will give you at least a benchmark and we have another tool that we that's built for traditional SEO um link backlink graph comparisons that helps us do that as well. So yes you can kind of get there you can look at who's getting the visibility now and and that's your that's your North Star.

SPEAKER_01 52:25

And I think I I'll also say that we we hear a lot like hey you guys don't know what people are actually prompting or how people are actually using um AI right now and what kind of multi-hops are so so it a lot of this is like not moot but and then personalization you know we hear a lot about well if with personalization you can't really know anything. And and I still believe that when we construct prompts and we've we've worked at this a lot and thought about this a lot is when you're constructing prompts to represent uh the the different sort of commonalities about how people make purchase decisions particularly uh you're able to at least model the types of concerns that are likely not definitely but likely to be represented or kind of um uh solved or trying to be solved for with personalization. Am I gonna can we see into that place?

SPEAKER_02 53:23

Of course we can't nobody can but but that's a I just want to call that out like it really like but but the prompts that you're tracking are 100,000 percent like as an industry we're under uh indexing on that right now like really we there haven't been enough discussions about what prompts are we actually tracking well you'll hear a lot of people tell you what not to do right like it's what LinkedIn is full of right now don't do this out don't do that you can't do this you can't do that well what works in your industry isn't going to necessarily make sense in somebody else's so proceed slowly conscientiously and really build out a basis of uh uh the types of prompts you're tracking so that you could be more likely to answer questions like well how many citations do we need well are we even tracking the right prompts do we even know are we and again another just a reminder to everyone on this call bottom of the funnel is where we have hyper focus so when I say prompts I'm I you I mean like bottom of the funnel types of decisioning prompts what else we got Dan uh we had a question from Yazid why is decision efficiency often overlooked in content strategy um it might just be semantics like my own because I you know do people try to help others make great decisions with their content I mean theoretically maybe it's so persuasive like our as you know the marketing side of content marketing is is going to be kind of trying to get people down a funnel and to you know make a decision that benefits the company and and I still want that to happen too but I know that that's not necessarily giving our humans the people the audience the the community the information they need to make the best possible decision for themselves and their their organization and so there that that's where you know things like Reddit come into play and the reviews and the review sections and and so we really have pushed a lot for people to lean into uh customer service logs where are we seeing frictions emerge what should we have said on the sales page before we took their money right so that we wouldn't have this issue we're having now uh how could we have better informed our our uh our buyers it's a scary place we're we're marketers we we get we get paid you know we get paid if we if the company gets paid right like this isn't a it's it's it's a it's a yeah tough spot that's the that's the thing Garrett is you know I think one of the challenges I'm slightly more jaded on this I think I'm I think one of the challenges is that because we're not our job as marketers is not to help people make the best most informed decision in other words remove that friction make it more efficient our job is to sell them on our offering our product where it's always always with that bias. There's this book called Answering the Central Question and the central question is what decision can I make or action can can I take in this moment to create the greatest net value. So creating the greatest net value or reducing the the friction or increasing the efficiency might be you know what this product isn't right for you and we don't want to say that. So I think that's just certainly plays into why um decision efficiency isn't always present in our sales landing pages.

SPEAKER_00 56:44

But it's present in your humans that you're marketing to and and like they really do want to make the best possible decision for themselves and uh you know and and so will their their LLM uh and AI uh associates right the the in we hope we hope that's where we're where they're gonna lead us uh what else uh we had a question from Kelly how important is video in AI used to be considered good for SEO and what is the status with AI?

SPEAKER_02 57:11

Oh um okay so we are absolutely seeing YouTube videos show up I I you know in what vertical for what types of prompts well not every you know we're we're looking in the the pharmaceutical consulting space there are actually YouTube videos there but not as many as in like the um some different small business SaaS type spaces right like invoices or you know these types of of areas but so so some of it will depend on you know the the what what's being published here but we did see this is crazy but somebody had taken had published to LinkedIn the recommendations from a video and that what got cited right was the LinkedIn post but what the post was was a top 10 list that had been pulled from a video from the video from the video and I don't remember if the video was actually cited as well in addition but my point is is video important I I I think absolutely I mean let's talk about our the the platforms that you know everybody's vying for uh Google and who owns YouTube uh so uh I how do you not try to it depends you can look at the data I mean you can look at the data you can see if video is getting is getting included in the in the um AI answer you can also look at the citations that inform that AI answer to see if video is in there and that then you can answer that question with a relatively low lift and for many of the prompts that I've dug into yes video is important.

SPEAKER_00 58:52

Okay just looking through these questions of a few here I think we'll just do one more and call it a day uh there was one from ARR. Do you think citation dependent AI mentions will stay or be replaced by more bigger brands dominated just not here okay we just are getting comfortable wow wow okay yeah I mean moving target like how long okay ask the question again Danny I might not I might have just been uh responding emotionally to what I thought I was hearing let me hear one more time okay do you think citation dependent AI mentions will stay or be replaced by more bigger brands dominance in the next one to two years I think it's gonna depend on the space I think it's gonna depend on the types of questions that are being asked I think if you have a product or you're in a space that uh things are shifting quickly if you sell hammers maybe your your your it's game times up.

SPEAKER_01 59:53

But if you're selling like instruments that uh are part of I don't know an evolving market space where you have to kind of stay up to date on things that are shifting and changing uh then you're going to continue to see uh emphasis on citations.

SPEAKER_02 1:00:08

Um I would I would guess so if I understood the question correctly uh that that's my answer if I didn't you know just email me like yeah I think I would confidently say we don't know at this point. We know that we've got to throw out the the playbook it's such a different calculation to serve up that AI answer than it was to serve up to serve up a single page or really 10 pages. Just 10 pages or more go it go into a single AI answer now. Inform that AI answer up to 2,000 words maybe more inform that AI answer. So I I think we don't know at this stage but we're definitely trying to figure out.

SPEAKER_00 1:00:48

All right let's sneak one more in really quick if you guys can stick around for just one minute. Brenda was asking small business here how do we find where competitors what citations they have and mentions on a budget friendly site a high AI hasn't specifically shown me.

SPEAKER_01 1:01:03

So this is we might be conflating and we we we realize we would be leaning into this space but there there are you know small business citations and um our our colleague Darren Shaw made a tool for uh citation building right which is NAP data name address phone number and trying to get that on as many different websites as possible. And so Darren go to Darren like he if you if you're talking about like that type of small business citation which isn't like it's kind of I guess the same relative kind of idea for um you know entity citations uh but so so it's not too far removed but I think that's the I think that's the answer that would is is go go check out LightSpark. He's he's built out a site you know that kind of citation uh analysis. We did release a um citation optimization or you know kind of comparative uh tool within it's in beta still but within the Zofu uh tool set so it I don't think that's quite what you're asking though. But it's more about where you know at the bottom of the funnel uh when someone is asking about what's the best um umbrella for hiking uh well we're gonna be able to kind of show all these different surfaces that are being that that you're sighted on but uh or that you're not sighted on but your competitors are and kind of do comparisons there so you can see uh which which um domains and publishers are uh contributing in in in a in you know where where where are they but you aren't uh sort of sort of play the answer's a hat by the way Garrett the best umbrella for hiking is a hat wow I love that all right I'm gonna put that I'm gonna put that to the test okay like I I'm really I'm about to go hiking in the rain actually we're in Louisville and it's it's it's hiking time for me it's walking time and we get out and walk um but Danny you wrap I was about to wrap I'm wrapping like we're on a like like a a client call. No we're good I'm in I'm in Danny's house.

SPEAKER_00 1:03:15

Wrapped in all right we'll wrap it up that's all the time we have for today if we didn't get to your question we'll be sure to pass it along to Garrett and Citation Labs on behalf of Search England I want to thank Garrett and James for their engaging presentation and thank you all for attending this webinar. We hope you will join us again soon. Bye everybody thank you by guys