Name: Yext Research Live: What the Data Proves About Brand Visibility Today
Uploaded: 2026-03-12T18:11:21.786Z
Duration: 53 min 34 s
Description: Yext Research Live: What the Data Proves About Brand Visibility Today

Transcript for "Yext Research Live: What the Data Proves About Brand Visibility Today": Good morning and good afternoon, everyone. Welcome to Yext Research Live. I'm Javier Fabrega, VP of data strategy here at Yext. And I'm excited about what we're sharing with you today because this is original research that I think is going to change how you think about the visibility for your brand. The premise here is simple. Customers are no longer find you by scrolling through blue links. They ask AI a question, and AI and AI either knows the answer, about your brand or it doesn't. More visibility metrics have have most visibility metrics haven't caught up to this, reality just yet, but today, we're going to fix that. This slide says it all. Search started fragmented, converged at Google, and AI is fragmenting it all over again. ChatGPT, Perplexity, Gemini, Claude. Your customers are getting answers from all of them, but the question is, are you showing up? That's exactly where our research team set out to measure. On Elo ranking, we looked at twenty one point point six million search results, 87 geographic coordinates, 30 industry categories. Anthony Rinaldi, along with their data science team, built a brand new way to measure competitive visibility using an Elo ranking system. We borrowed that from CHESS. That actually reflects how search competition works in the real world. On AI citation analysis, we studied 17,200,000 distinct AI citations, four AI models, seven sectors. This is the largest study of its kind on which brands get cited in AI and which ones get left behind. This is Adam Abernathy's work. So with that, I'd like to introduce to you, Anthony Rinaldi. He is our senior director of insights and analytics. He's gonna walk you through the Elo rankings side of the presentation and also Adam Abernathy, our senior director of VX Research, who's gonna then walk you through the citation research that we've conducted over the last few months. Anthony, over to you. Cool. Thanks, Avi, and thanks everybody for joining. I am super excited to talk to all of you about the same type of thing I've been talking to clients about for a long time. So I would love to sit here and wax poetic, but I do have, I have a handful of slides that I'm gonna walk through. So I'll try to get them pretty through them pretty quickly. But what I found with clients is sometimes when I'm talking about Elov research, it it feels like I'm got a a whiteboard and a lot of markers, and and I the slides will just help here. So I I labeled this as part of the presentation finding signal in the noise, the case against Google rank. And I think what you'll learn is I'm not saying not to use Google rank. I'm just saying there are probably better ways to think about traditional search moving forward. So I have been been with Yext for seven years now, and we have always had tools, whether it's Google Search Console or whether it's our competitive intelligence and search tracker tool, we have always had tools that could tell clients how they rank. Right? And Google rank has always been the holy grail of KPI within traditional search, within listings, within within pages, within the organic part of search or the paid part of search. And I, for seven years, have had really frustrating conversations where us and clients are trying to come up what it means to rank first. Right? Because what I end up seeing is that when I search, it's different than when you search. And when we searched on Tuesday, it might be different than when you searched on Wednesday. And when you search from home, it might be different than when you searched at the store or at your location or at your hospital. Or when you search from your desk, it might be different than when you search for somebody else's desk. And what we were starting to find out was that, like, search ranking is is very, very tempting of a metric to to report, and to report to your leaders. And it's it's, you know, again, the holy grail because ranking is is all that anybody has ever wanted to do. But what we found out is there there's just shortcomings on saying what your rank is. Right? So we I I heavy we here. The research team did a lot of the analysis, but our data research team did a lot of the the heavy lifting when it comes to the data, science part of this, have started to use EloRank. And I'm gonna use the term EloRank here just a handful of times, and and then I'm gonna try to never use it again. Because for those of you who have heard of Elo scores before, it's not something that we came up with. Elo is something that the sports world has used for a long time. Elo does not stand for anything. It's actually the last name of the scientist that came up with this ranking system. But, essentially, EloRank is a scoring system used in chess, but it's also used in college football or college basketball so that we can try to put data to not just did you rank first or did you rank second, but maybe should you rank first or should you rank second. And if you ranked first today, are you likely to rank first tomorrow? And the way that EloRank does this or EloScores do this is they pit all of the competition up against each other in matches, and then they play these matches over and over and over again. So if you think about college basketball, when the number one team beats the number two team, that's a really big deal. But when the number one team loses to the unranked team, that's an even bigger deal. So this is how we, at YAX, within the research team and the data science team, have started to make traditional Google rank a bit more contextualized because we don't wanna know how did you rank on Tuesday at home. We wanna rank we wanna know how might you rank tomorrow no matter where you search. So here's where it helps to use my slides. So think about, think about a traditional search engine results page or a SERP as a matchup. Right? It's not college football or college basketball, and it's not chess. But, essentially, every time somebody uses, Google, they get results. And coming in first is essentially winning your matchup. Coming in second is good, but it's not as good as coming in first. So whether we're talking about sports or we're talking about, Google rankings, this is your first matchup. Before your first ever matchup, Starbucks and Dunkin' and Pete's and Local Brew and Java House all started even. Right? They get 1,000 points each. And the first Google searches run, Starbucks came in first, Pete's came in second, local brew sorry, Java House came in third followed by local brew and Dunkin'. And now we have our first matchup. Right? So this answers the question, how did I rank on Tuesday at noon? It doesn't answer the question, how might I rank next Tuesday at 5PM? Because maybe my coffee shop is closed at 5PM, so I'm not expecting to rank first. So then what we do is we run another search. Right? And we have another matchup, and we have another piece of data. So now local brew won this one. Right? They didn't jump to first, but, boy, did they come up a lot. And Starbucks came down a little bit because in the second matchup, we have a whole new group of results. And we do this and we do this and we do this until this starts to converge over many, many, matches. Right? Eventually, Starbucks has a certain Elo score, and Pete's has a certain Elo score, and Java House has a certain Elo score. And if we wanna continue this analogy to college basketball or college football, think about somebody like Pete's who fell off altogether. Right? They're no longer even ranked. They're unranked. So we know that there is a Pete's in this neighborhood. We once saw Pete's in this neighborhood. So we expect to see Pete's. If we never see Pete's again, they're unranked. And they're unranked in the next search, and they're unranked in the following search. And every time somebody new comes above them, they're getting bumped down just a little bit. So this helps us get, again, not a point in time ranking, but something that's a little bit more useful. So at Yext, we have like I said, for all seven of my years, we've had different products. We've had search tracker and competitive intelligence that could tell you about your traditional rankings. But now we have Yext Scout. And Yext Scout is doing these matchups for over 4,300 keywords in over 200,000 different regions monthly at a minimum. Right? So at the research, team here at Yext, we can run this and run it again and run it again and run it again. And we got to 4,300 keywords because, essentially, that is how many primary categories that locations we power use. Right? So we have things like coffee shops and banks and financial advisers. We have cardiologists and emergency rooms. We've got Italian restaurants, and we've got fast food restaurants. Essentially, 4,300 different types of businesses, and we're running them all across The US. So the research team before we started really ramping up YextScout clients was just doing this over and over around the country. But now that we have clients launched on YextScout, you know, the data comes by itself. Right? It just comes and it comes and it comes. So what we have now is over 21,000,000 search results by the time that we ran this analysis. And because of that, what's very exciting is that this research through YextScout, finally, is not just of Yext clients. Because every time we run one of these searches, we get the client that we were running it for or the business that we were running it for. We also get all 20 or 40 other results on the page or the second page. So what's nice about this is we get 21,000,000 businesses. Right? And what's very, very powerful about this dataset is it does not just tell us who ranked one and who ranked two and who ranked three, but it tells us everything about them. Right? Because when we get this data, we we scrape it from the Google My Business. We scrape this from Google Search Console so that we have right? For the person who ranked one, what was their name? Did they have a description? What were their hours? How many attributes did they have? How many photos did they have? When was the last time that they uploaded a photo? How many reviews did they have? Are they good? Did they respond to them? So we get about a 100 to 200 different facts about every one of these results. So if you think about this dataset, not only do we have for each keyword in each region over time who ranked one, two, three, four, five, etcetera, But we also have for the people who ranked one, how many photos did they have? For the people who ranked off the page, how many reviews did they have? And that becomes an incredibly valuable dataset. So what did we do with it? We saw, essentially, when you marry the results in the ELO rankings and you measure the pieces of data that they're managing, essentially, the conclusion here is the folks that do a really good job managing those 200 fields rank about three spots better than those that are unmanaging that information. So across all results, we find that, you know, intuitively, we see folks, putting in the effort, get the results. What does that mean? Well, in the data, when we scrape it, we get everything, like I said, not just the operational data, but we get data about the search. So we can tell how far were you from where we searched, what was the area like that we searched in, and how big were the businesses that came back. And these were three ways that I was very interested in searching or or doing analysis on. Because the first one, incredibly intuitively, you would expect businesses that happen to be right next to where we search should rank better than businesses that are 10 miles away from where we searched. The search competition, that was a tough one. So, I like to use the example when it comes to competitive markets. I like to use the example about New York City because you think that New York City is a very, very competitive market, and, generally, it is. New York City has, you know, thousands of coffee shops, and it has thousands of restaurants, and it has thousands of ATMs. But New York City is not a very competitive market in every keyword. So we see in certain markets, certain keywords are more and less competitive. The example I like to give, unless you're out on Long Island or unless you're out in the Hamptons, New York is not very competitive for the search term surfboards or surf shops. So what we did in this analysis was we looked at not just on the first scan for the first match, right, not just on the second match or the third match, but over the course of all of your matches for your keyword in your location, how many distinct businesses ever came up? Because if I search for the search term surf shop near me in New York City and I search that 10 different times and I'm always getting the same 10 businesses back, that means, yeah, New York might be a competitive market generally, but New York is not a competitive market for surf shops. So we wanted to look at this analysis to find out, do we get different results in very competitive markets, and do we get different results in in not so competitive markets? And then brand size and brand equity. We also wanted to look at, you know, when we see a result, it's very specific. It is not PNC Bank, but it's PNC Bank at 5th or Lexington. Right? So what we wanted to do is try to relate those actual results to the larger parent brand. Because, you know, for years, we've had inklings that big businesses will have a harder time ranking first than small businesses. Or Google's happy to put enterprise size businesses in the top five, but they'd rather put a mom and pop in one, two, or three local pack. So these were the types of questions that we wanted to answer. And what we ended up finding was that intuitively the results are very different by distance from the search. So we have the within a mile, We have one to three miles. We have three to five miles. We have five plus miles. I'm sorry, it looks like that first one should say within a mile, not one to three. But what we saw was that affect that two and a half, 2.7 position bump in the ELO rankings was for within a mile. Right? So essentially, when people are actually competing, when they are the ones that are most likely to show up, it matters most to have your data really under control. When you get a little further out, it still matters. And when you get very far out, it's anybody's guess. Right? We're we're doing more analysis and what it means to rank top one, top two, top three when you're more than five miles away. But to be honest, as you get further away, they're just filling the search results page sometimes. Right? You're gonna have a really hard time showing up in a search five plus miles away, but we can still try to find out who showed up five plus miles away and maybe why did they. So most of the analysis that we do in the ELO ranking research that we've done was focused specifically on that within a mile. Because to me, that was if you're within a mile of where we searched, you had a chance. Right? So if we wanted to honestly, truly say, did you do better than somebody else? You wanna make sure that you're talking about a fair fight. Because if you happen to be right next to where we searched, of course, you're gonna have a better chance of showing up higher than somebody who's five plus miles away. The second thing I wanted to talk about was the market. Right? I have tagged these as we've run a bunch of matches searches. And if you've never even had 20 unique businesses, that's a very uncompetitive market. If you've had somewhere between twenty and fifty different businesses show up on your first two pages, that's a very normal market. Right? Because the first two pages should be somewhere between twenty and fifty different businesses. 50 to 75 unique businesses and a handful of matches, that's a pretty competitive market. That means that for your keyword in your neighborhood, there were more businesses than there are spots. So you're not just competing to be in the top 10 or the top five or the top three. That means you actually won. Just being on that page was a win because there's so many options that could come out. And then ultra competitive, these are the markets and keyword combinations where over the course of a handful of matches, we've seen almost a 100 different businesses. So again, think about New York City coffee shops. In New York City, from any given point, you could probably come up with 50 to a 100 different coffee shops that are worth ranking. So we think about these in uncompetitive, in standard competitive, in pretty competitive, and ultracompetitive. So what we see here is that going from left to right, not so surprisingly, there was a two and a half bump, two and a half position bump from managing your data within a mile generally. Right? There's almost a six position bump in this long term ranking or this contextualized ranking or whatever term you wanna use for our ELO ranking. There's about a six and a half point bump in the standings when you're in an ultra competitive market and you're putting in the work to manage this data. And then what was maybe most telling was that that goes to eight positions when you're a larger enterprise brand. So this is the fair fights. They're within a mile. This is ultra competitive. So there was over 75 different businesses that Google has historically ranked for this keyword in this area. And you are part of a parent company that has hundreds of locations. Right? So I simplified this slide because I think the point is a very important one. The businesses that are biggest in the areas that are most competitive when the fight was fair, so when it when it was within one mile, you get, on average, an eight position bump from managing your data better than others. What does it mean to manage your data better than others? So I said we have 100 to 200 different metrics. Right? Is your profile claimed? How complete is your profile? Does your profile have business hours? Does your profile have a cover photo? Does it have reviews? How many reviews does it have? Does it have photos? How many photos does it have? When's the last time you responded to a review? I put a handful of them on the screen to show you the difference from being in the top three to being off the first page altogether was being in the top three, businesses often claim their profile at an 86% rate. Those that were off the page, 7% less often. And I think that one's pretty intuitive. As we go down here, you'll see maybe some surprising things and some not so surprising things. Right? Having social links. Not even doing anything with social, but if a listing has social links, they're more likely to be in the top three than somebody who doesn't have social links. Attributes and secondary URLs, I think is that definitely market specific. So this was all 21,000,000 results that we've looked at. It's tough to compare a hotel to a restaurant, and it's tough to compare a restaurant to a hospital. So generally, a lot of these things hold true, but then you start to see differences when it comes to things like hotels. So ultra competitive and big brands, that was the slice that I was most excited about having the big differences. And those differences, again, kind of get bigger. Being in the top three versus not being in the top three or sorry, not even being in the top page. In a not so competitive market, you only needed an 86% profile claimed rate. If you're in a very competitive market and you're a big brand, the bar's higher. 95%, 95%, 93%, 99%, 95%. So the bar, like I said, is just is that much higher if you're in a very competitive market and you're for a very big brand. I mentioned a minute ago that not all markets are sorry, are created equally. One example that I love to point to is hospitality. Right? Because we've been doing these studies. I think Javi showed on the slide, we called it seven industries, and those industries go into tens or hundreds or thousands of sub industries. So an industry could be hospitality. A sub industry could be bed and breakfast. An industry could be food and services, but a sub industry could be, Italian restaurants. So if we look specifically at hotels, one thing that kept jumping out, and I think this is a good example about how you need to treat your market uniquely, is that, generally, I would sit down at any any meeting with any client and tell them, you need to make sure you have core information. You need to make sure you have your name, your address, your phone number, your hours, your services, all of that stuff. And what we started to find out in, you know, millions of search results for hotels, the hotels that are more often in the top three less often have business hours. Right? That is so counterintuitive, but we've seen this over and over again. So, you know, you can hypothesize that maybe if you're a hotel and you're sending business hours to Google and you're syncing your business hours, then that's a sign that you're not twenty four hours. Because if you were twenty four hours, maybe you'd fill out the attribute that you had twenty four hour check-in and that the hours there are just redundant. Another thing here that you might see as a callout is photo count. Being on the top three actually has fewer photos than being off the first page. Almost 200 almost 250 fewer photos. This is another thing where I would say, think about the market you're in. Because overall, think two slides ago, it said something like you need a couple 100 photos. In hotels and hospitality, you need a couple thousand photos. So what's nice about this is you can see once you get to something like a thousand photos, every photo thereafter is maybe not so useful. Right? And this isn't every metric that we see. This is just a handful of metrics because what we do see is that for hospitality, there's only so many photos that are necessary to upload as the owner, whereas in certain businesses, you would prefer owner photos. But when it comes to a hotel, they would rather see nonowner updated photos because that's probably a more realistic representation of the hotel. Same thing with, like, an emergency room. Same thing with, like, you know, all of these service based businesses that would like to see real pictures instead of your, you know, photoshopped beautiful pictures. So just an example of when I tell you some big numbers overall, it it it varies by industry. It varies by sector. It varies by region. Because you could see here, the recommendation that I'm gonna give to a fast food restaurant is not the same recommendation that I'm gonna give to a hospitality brand. The last thing I'll give you with is an example that that Christian Ward and myself talk about a lot, and it's we did research about not just having photos, but having headshots for financial advisers. And, of course, I would tell the financial adviser to get your headshot. And what we found is that, generally, folks in the Northeast matter less to have a headshot for their financial advisers than folks in the Southeast. And maybe, again, we don't have these answers directly from Google, But is it that in the Northeast, maybe in New York City, they're they're more focused on your credentials? And maybe in the Southeast, maybe in Florida, they're more focused on, you know, where did your adviser go to school? Do we have anything in common? Do I wanna go play golf with him? Something like that. And it's just it's so important to know what you need to do and where you need to do it. So, finally, what I'll leave you with is that what we end up seeing is twofold. One, and I think this would be a really good transition to Adam Abernathy's research, is that whether it's traditional search or or, AI search, it's it's the same as it's always been. Having more information in more places seems to make you more relevant, whether that's Google trying to find out if it's important to show you for a search about having twenty four hour check-in, right, where you need to tell them that, or whether it's AI search trying to tell you whether or not this hotel is relevant for my my questions because, you know, you needed to tell them that. So it's the same as it's always been. I don't think about AI or traditional search any differently than I think of you or I because I think the results are supposed to act in your best interest. Right? So what would you look for? Well, you'd look for more information. What would you wanna know? Well, did more places say the same thing? And finally, I'll say this whole research around around Elo ranking and Elo scores and, like I said, finding signal in the noise with with Google rankings, it's it's just about measuring outcomes instead of numbers. Because, historically, people wanna know, did I rank one or two? But I think moving forward, you know, people vote with their clicks. Right? Yeah. Did I rank one or two as nice, but did I actually get more clicks than I used to get? Did I actually get more impressions than I used to get? Was my long term sustained ranking a little higher than it used to be? These are the things that, you know, I think that folks need to focus on a bit more than just ranking one, two, or three. And I I see Javi on the screen now, so I think I might be getting the the hook. How you doing, Javi? Thanks, Anthony. I'm doing great. That was awesome. I do have a couple of follow-up questions for you because that was that was a lot of data, as you know. So let me ask you. What surprised you the most when you actually went through your study? I don't, maybe I maybe I gave away some some of the cow without buying the milk. But I think that the differences in industries have been so interesting. So finding out those those nuances of, not that this industry, but this industry was really fun And, also, finding out how much higher the bar is for big brands was really interesting. Because, again, I I kinda mentioned off the cuff a while ago that that we've had hypotheses about about how hard it is for big brands to show up. And I think what we've seen in the data is it's really hard for big brands to show up. Right? Not all not all businesses are created equally. So a 100 photos for you is not the same as a 100 photos for them, and a 100 reviews from you is by no means the same as a 100 reviews for them. So I think the bar is just it's it's a moving target, which is you you need this data, and you need to be able to see this constantly to know that. Absolutely. So let's talk about AI. So how does AI generated search change the competitive landscape? One. And two, is a brand ranking number one in traditional search also winning in AI results? Sure. I think because of search fragmentation, you mentioned it earlier, because of search fragmentation, it is, more important than ever. I mentioned it at the end. Maybe I I said too many of these answers before this part, but I think it's more important than ever to to focus on outcomes. Because I've been having these conversations with clients for the last year and a half, maybe two years now about some impressions going down and some clicks going down and things like that, and they're not going down. People are still buying things. People are still going places. People are still shopping. They're not going down. They're just going to different places. So I think, ever since COVID, I've been mentioning things like being the best in a market that doesn't have a lot of engagement is not as good as being, you know, mediocre in a market that does. So can we focus on the things that matter? Right? Like I said, I don't wanna beat this into the ground. Ranking is not maybe impressions have not always been the best thing to look at because they're going to places like AI, and ranking's not maybe, like, the best thing to look at because it it could tell a different story than the amount of demand. But I think what I would say is that clicks have always been true. Right? So if you're getting as many clicks or you're you're looking for as many clicks or clicks are coming from somewhere or if you're an e com business or a hospitality brand who's looking for online engagement, you know, those clicks to your website might be going from different places moving forward. Got it. ranking is rank and I'll your follow-up there, Havi. Is ranking number in traditional search in AI? Are those the same companies? They have a lot in common. Right? And I think Adam's gonna go over this, but I I mentioned it before, having as much data as you possibly can seems to be the key to rank in either place. Right? So think about AI similarly to thinking about how you you rank on Google. I gotta tell them as much as I can about me so they're prepared to answer the questions about me. And that's that seems to be true with AI as well. Yep. So lots of structured data and optimized frequently. Alright. Last question for you is if a CMO or marketing leader is watching right now, and this has only been and and and they have been only been looking at rank tracking. Right? What's one thing they should do tomorrow morning? I'll give you two. One one way to think, one way to act. First of all, the world's not on fire. Right? Well, it might be. But in in search, the world's not on fire. Just because, you have fewer impressions does not mean things are going poorly. Just because your rank is not what you thought it is does not mean things are going poorly. So, again, think about think about outcomes. Think about bookings. Think about revenue. Think about actions because they're coming from a lot of different places, and not all those places have really good data behind them. So first, before you do anything, just make sure that you're not overreacting. And then, if I were going to tell you to do something, it's just, you know, double down on data. Double down on on on more pieces of data in more places, and and we can make sure that these AI models and these these traditional models can act more, more intelligently. Because, again, I think of them acting in your best interest, hopefully. And the more information you can feed them, the more confidently they can give the answers that you want them to give about you. Awesome. Well, Anthony, thank you so much for all the work you've done on the Elo side and continuing to do. And let's hand it over now to Adam Abernathy to learn all things citations. Yeah. So thanks for this today. You know, also wanna give a shout out to the, St. Louis folks. I've got a slight hometown connection to St. Louis, so, it's good seeing some folks from, like, my part of, you know, town here. So, yeah, Harvey. Had to hit the mute button there. So, Adam, you've been studying AI citations at scale for I don't know how long now. Right? So I guess for the crowd, what is an AI citation, and and why does it matter more than traditional search ranking? Yeah. So to start with, like, the citation, it's it's just a reference to to to some sort of, like, web resource that a model points to. You know, it's just a URL that that the model said, this is where I got that from. You know, in, traditional search, brands, you know, compete for rankings on, like, a results page, like the SERP, like Rinali just talked about. But with the, you know, AI powered search, the these models take a bunch of like, they go out to the Internet, and they search a bunch of stuff, and then they it's part of what's called the RAC process. And then they take all those resources, and they cite them as they're kinda synthesizing like an answer. And then, you know, they bring the like, the citations is kinda like the receipts as to where they got the information. And so with with these things, what's important to know about them is you're either in the answer or you're not. Right? And so unlike rank where there's kinda like a presence, you know, like, kind of, like, a positional thing, with the AI, you know, thing, you're either part of that conversation or you're not part of that conversation. There's not really, like, a page to to go look at more unless you ask, like, specific questions. And so that's why we're, that's kinda why we're spending quite a bit of time focusing on, you know, you know, what are the citations, what are the behaviors, different models, different questions, different industries, stuff like that. Very cool. so yeah. So from the market's perspective, keep in mind that 75% of your users are using and your just consumers in general are using AI tools more than than they did a year ago. We did a pretty large scale consumer study at the end of last year across The US and parts of Europe, and we found that, you know, that 75% are adopting these, like, these tools more and more. And so these search things are really more and more part of people's daily lives. And that's and so there's what the like, the research that we just published and we're gonna talk a little bit today, it's it's really like there's no single a optimization strategy. Businesses, you know, you kinda have to optimize for you know? Like, you can't just optimize for one thing. You have to because that that one thing doesn't exist. It it it it's a really fluid, you know, surface. Awesome. And to that point, we are now in the second research paper when it comes to citations. What were the updated queries, or I guess I should say, query categories you looked at? And what drove that evolution from your previous research paper? Yeah. So the first so when we we did the last paper last year, at the end of last year, believe it was October's when we published. It's we looked at like, we wanted to we we wanted to push a problem because a a lot of AI, you know, surveys and kind of brands, you know, stuff is done at at the brand level. So you're asking, you know, just say ChatGPT, you know, what do you think about insert brand here? And you're kinda getting this kinda high level kind of, you know, kind of in the stratosphere sort of type response. And so that's why but we wanted to understand how are consumers using these things. And so we developed what's called, like, the intent quadrants, which is that where we look at branded versus unbranded queries and branded versus unbranded questions and subjective versus objective questions. And we did this beca you know, because that kinda covers kinda every kinda question. And I I I think you'd be hard pressed to craft a question that doesn't fall in one of those buckets. And if you can, please, like, tell me because we're curious too. So but what's new about this one and we did we and we did that based on the sectors. Right? So we looked at very high level sectors. And so when you think of sectors, think of health care and finance, retail, hospitality, you know you know, food service. And so very kinda high level. And, but now that so that was really focused on q three of last year is what that dataset focused on. So then when we looked at q four of last year, we not only went from 7,000,000 citations in that dataset to about seven to just over 17,000,000 citations in this q four dataset, is that we we built out a taxonomy that models really closely to how kinda Wall Street and investment banks kinda look at or investment firms look at kinda how they sort businesses into different groups. And so we kinda filed a really kinda industry standard taxonomy there. And then what we found was that in the sector level is that there's a lot of industry and so we looked at the industry. So instead of so for, like, hospitality, there is everything from, like, recreation, or there's golf courses, there's hotels, there's all these things. And we started looking at the interplay kind of volatility in those individuals' industries, and we actually went down to the sub industry, which would be, like, very specific. Like, instead of saying outdoor recreation, you're saying golf courses versus tennis facilities. We found there's a lot of, like, volatility and activity there. And so what we're seeing we're able to see now is how these citations even move over time because we're we're we're doing this in time series because we're scanning stuff in real time. Like, all you know, kinda do this. And so we're we're starting to to like, to, like, to really see how these things kinda evolve and change over time, there's a kind of a different kind of a different formula for every one of these. Right. Like, all citations are not equal here. Correct. And so these help you you know, so I understand this helps the content strategist and stuff kinda understand where you need to target and stuff like that. No. And that's that's absolutely interesting because when it comes to also the different sectors that we looked at. Right? And so, I guess, for industries, which industries are getting cited the most, and which ones were, I guess, for better lack of better words, being left behind when it came to citations? Yeah. Let's not think of it like who's getting inside of the most kinda thing, but let's think of it through kind of, like, this industry kinda competition and, like, variability. Right? What we see and so, like, what we see and I'll show you a slide here in a second. But what we see in these in, like, your money, your life industries is a strong alliance on kind of first party sources, whereas in food, there's a stronger tilt towards, third party directories and reviews. In industries that depend on third party, you know, kind of pressing community coverage, like, they're like, they're more exposed. And so let me is is that okay. Cool. So what we see here is, like you know, when we look at kinda at the high level, there's, you know this is kinda like the a very generalized formula. This is across, like, the four major models. So, so, ChatGPT, Anthropic, Gemini, and and a a perplexity. And so we we kinda we we we can just kinda see in different, you know, sectors, there's kind of a different formula. So the the dark blue is first party website. So this is your brand's website. These are things that directly, directly under your remit. And then when you get into some control, these are getting into, like, directory services. So these are the Yelps, the MapQuest, all that kind of stuff. All those kinda things like those wagon like, spokes off the wagon wheel sort of thing. And then limited control is getting into social and reviews, and the no control would be stuff like like like like bloggers that, you know, comment on the on the industry or, you know, news, you know, you know, news outlets, regulatory type stuff. Reddit and Wikipedia are actually in that, like, the in that red column there. And so what you kinda see here is, like, just by sector, there's lot of like, there's, like, differences. But if you go to the next slide really quick. Never mind. Okay. You see so now and and this is actually new research. We haven't published this yet, but we will, like, shortly, is we started looking at the volatility between each individual sector that we're or each individual industry that we're looking at. So you can definitely see it. So, like, hospitality, there's quite a bit of, like it's, like, relatively stable. As as we look at some of these that are in, like, the more blue, there's a lot more volatility. And so there's just, like, those individual kind of, you know you know, sub industries, like, there's a lot of space. And I bring all this up to say that, you can't just look at kinda, like, the top formula and go, that's what we need to target. It's you really kinda have to understand what's going on in your specific, you know, area and then your specific geography. We also look at the at at this via geography. And similar to what Renaldi was talking about, Elo and all the you know, Google is basically a signal. It's all based off, like, a signal clustering is how, their search thing works. We we did some we we published some work on this a few weeks back, and, it it it does vary by, like, region, by, like, keywords, by those kind stuff. It's really veritable, and so you kinda have to really know and observe what your particular market's doing. No. That's awesome. And you you you mentioned signals, and signals, obviously, when it comes to AI structured data, are all signals of trust, right, that the AI is looking for. Question when it comes to well, I guess, let's just dive into AI. So legacy rank tracking gave marketers something to optimize for. Right? And now what's what. what would be the equivalent lever in AI citation visibility when it comes to that? It really gets back to, like, to, like, diverse diversification in, like, your footprint. And so when we look at who's rising to the top and who's consistently being cited when we ask these questions because remember, we're asking we're asking branded questions and unbranded questions. We're asking. objective questions and subjective questions. And so we're not just asking about a specific brand over and over. We're asking really about, you know, like, these kinda generic questions that, you you know, would really open up the field. And when we see who's right on top, the ones that continually come up to the like, the top here, it's they have a really deep content infrastructure. I really can't stress this enough. They have a lot of relevant and salient information to whatever topics are relevant to that industry and that subindustry. And so they they're they're conveying this through really strong websites, so, like, structured data, all the FAQs, really, well thought out kinda content strategy that answers questions, less kind of, like, the kinda less on the sales y part, but, like, more on the, like, we're answering questions that people have. Right. And and then they're also really strong through directory ecosystems and then active review platforms. Reviews have a smaller portion of the control, but they do have but they do have control over there's this kinda middle child theory that we're working on on right now where because we know all these AI models, they use what's, they use a rag process. When they when you ask it a question, it goes out to the Internet and it searches for stuff. They use indices to do that. Right? Gemini uses Google. ChatGPT uses, like, a series of, like, different ones and stuff like that. And so they have to go out and look for stuff. And they look for stuff often from classic search. And so there's kinda pull there's a bit of a circular kinda pattern there. And so to kinda bring this back to the AI piece, it's really strong first party websites. AI models return first party content about 4.3 times more, like, than all the others and two and half times more for listings than other URLs. And so it like and then just having, like, a really good healthy social footprint. We find very interesting for Claude and Anthropic because they actually lean harder on social. They're the one model that really outliers on that allies on this. They've been really hard on social in reviews. And I think and and so, like, we we we talk about that in-depth in in that paper. And so to be online and kinda neglect that just because it's a smaller percentage of that available kinda real estate that you can go grab kinda makes you invisible to them, actually. Yeah. And so, and then kind of the inverse of that is is Gemini leans on Google. And so if you're not doing well on on Google, don't have a good listings presence there, you kinda become invisible to them too. And so I think that the takeaway is, you know, brands that are winning in this space are really you know, like, they're spread across. They're really like, there's a a lot of diversity in their footprint, and they have a lot you know, like, they're covering quite a bit of ground, basically. Right. Because, I mean, at the end the day, the AIs all take location into consideration. So location and query context is absolutely you know, you. know, it's it's gonna be part of most, you know, queries anyways within the AI. So having hyper local structured data also matters. Now let me ask you, I guess, in in preparation for the q and a. For. a digital performance leader in the audience, what's the most actionable thing they can take away to get their brand cited more? The honest answer is it really depends on the model, and I would not, and I I've been a marketer, and so I I I I understand kind of what, you know, it's, like, really easy to kind to to try and tune just for, like, a specific thing, but you have to kinda look at this really holistically. And it can seem kinda crazy, but if you focus on just kind of and understand that all these models behave differently, it actually and then but the but the real trick is having a good data strategy under the hood, you're gonna get pretty far. Right? And so, for example, like, Gemini, like I I already said, it's very search grounded. Right? Google owns Gemini. They have their own index. They really what we've seen is they really kinda lean on the EAT formula, the you know, is it, like, the authority and experience and all that kind of trustworthiness? That seems to carry over. Claude, like I said, leans really heavily on, you know, citations and reviews more than others do. It's not, like, fully skewed, but there's definitely a tilt towards that. You have a, you know, ChatGPT and OpenAI. They have they're kinda all over the place. They have really high variance, but depending on what your industry and sector is. And so, and so just knowing that and knowing how variable they are, you kinda really like, you don't want it to try and tune in on just, like, one thing. You really wanna focus on having that really consistent, you know, like, data footprint, information footprint, having content and for your conscious conscious strategy that answers questions, that that marketers or or that people can, you know, they can use because these models are gonna consume that. You know? And so, yeah, in the aggregate, websites make up about 45% of your citations. Your listing's about 44%, but these averages kinda mask those model level differences. And so that big takeaway is, like, don't really tune per model. Try to tune holistically across you know, trust just try to have, like, a really, you know, well thought out kinda consistent digital footprint there. Awesome. Well, thank you very much, and thank you for all the work that you've done with Yext Research. I think we're ready for some q and a. We got a few there in the hopper already. Here is the first one. So are AI citations meant to reflect the sources used to generate the answer, or are they sources that simply support or align with the answer? Yeah. They are so the way that the Rag process works or results augmented generation is the model like, the model has a certain level of kinda training data under the hood that teaches it kinda how to respond and how to think. But when you ask it, like, a question, like, you know, where's some coffee in Salt Lake City? And what it does is it doesn't just have all that information inside. So it actually goes out to an index, and it looks for websites. It quickly parses those websites, and then it creates an answer with the information that it parsed. And so it very strongly reflects the sources used because that's what it those those were the things that it used in order to to generate that response. Excellent. Okay. Next question. Is there any insight from your research so you can shed light on regarding home service space from this research? I hear a lot of storefront business, but rarely anything about home services. Sure. I can take this one. In the docs that I think Logan shared or or the admin here shared, we've released the ELA research. And I just opened it up after I saw your question publicly. It looks like we've got about 12,000 matches for home services. We can get more specific in what those keywords exactly were, but we've tagged, and it's in that in that article that we wrote, that essentially in 12,000 matchups in home services, there was about a 1.4 position bump for managing versus not managing their data well. So, yeah, we got in those almost 4,500 different keywords, we got we got very specific. So, essentially, if it's a business that we work with and it has a primary category, we'd probably run it within this research. So, yeah, plenty of homes at least 12,000 home service businesses and matchups. Thank you. Here's a good one. How are you able to decipher between business names? There are a lot of business names that may be similar, and how can you ensure that you distinguish between them? I Sure. I'm sure that there's a thousand or thousands of DailyBrews, but if I only want a specific brand Yep. how do they go about that? there's two there might be I I from the timestamp of when this was asked, I think I know when it was asked, on which slide. So there's two ways to answer this question. For the Elo research, it's very easy because, when we get the data back, we also get a Google CID, and the CID is unique to a business listed on Google. So, as far as did this business go up or down in the second or third or fourth match, That's that's pretty simple for us to say, you know, CID123 is the same on on those three matches. You didn't ask this question later in the presentation because where this got tricky for us was trying to say whether or not Daily Bruise was a small business or part of a larger enterprise business. So in the Elo research, fine. Perfect. We know who Daily Bruise was. We know who it was, throughout the matches. But for telling you if it matters for an enterprise business versus a large business versus a small business versus a a single location business, that got trickier. So we did first use URL to find out whether or not they're all part of a larger URL. We we then tried to use some fuzzy match for names. It gets difficult when you're looking for doctors or financial advisers who may or may not say that they work for a particular a particular house. But I I think if we weren't positive whether it was a small bank or part of a larger bank, then we probably just, you know, kept it out for the the brand size analysis. But for the Elo research in general, it's very easy for us to determine if it's the same Daily Brew. CID. Thanks, Anthony. Alright. One last question here. Could this be why we saw a 200% increase in impressions for Internet providers and Internet providers near me in January that didn't result in an increase in clicks? Looks like an earlier timestamp. I'll take this one. It could be. It could be that they're doing their research on Google and going somewhere else to act. I think that's that's the AI answer here. Right? Do your research here and act here or vice versa. Now. we'll see fewer impressions and just as many actions on Google. It could be seasonal. I don't know. If you we saw plenty of plenty of impressions for for those search terms on Google in January in general. I think, realistically, not to get too much into the weeds, when you see a crazy bump in, Google listings impressions, here's a fun trick. If you're looking at the GMB data, the bump probably wasn't in all impressions. It was probably in search impression sorry, in map impressions. And the bump wasn't just in map impressions. It was probably for those unbranded ones. So we see all the time. If you get something like a 10% bump, it's probably real. If you see something like a 200% bump, could be AI, or it could be just that, you know, for a particular month, Google was playing with the, when to show your pin and when not to show your pin on a on a a map. So if you didn't get that again in the next month or the next month, Google saw that you didn't get any actions. And if you didn't get any actions, then it probably wasn't a good time to show you. Thanks, Anthony. All right. Actually, we have one more. Is distance considered differently when comparing nationally ranked hospitals? For example, where patients travel long distance for treatment. It's a good question. Hospitality and hospitals are both unique in this because of the way you search. Right? Right. If I'm searching for a hotel, I'm probably not gonna be next to the hotel. And if I'm searching for a hospital, I probably if it's not an emergency, willing to travel a bit further for for care, For the best care. best care. Yes. So yes, we have all of this data in here, and I think we kind of simplified it for this presentation. But yeah, you're thinking about the right stuff here. Excellent. Let me check the chat one more time. I don't see any more questions from the attendees. So I think we'll wrap with that. Really appreciate both of you, taking the time today to walk us through the Elo research and the AI citation research. And thank you, everybody, that joined as well, keep those questions coming. Again, I saw a couple of folks that wanted to talk to sales as well. There is a link on our website and, or, also, we can also connect you with, sales as well. So thanks again, everyone, and hope everybody has a great rest of their day. Thanks, guys. Take care.