SEO for Brand Visibility in LLMs [Webinar Recap]

Dana DiTomaso: [00:00:00] Hi, everyone. Welcome to the Kick Point Playbook webinar. I’m Dana DeTomaso, founder and lead instructor here at Kick Point Playbook. And today I’m joined by Crystal Carter, head of SEO communications at Wix. And today we’re going to be discussing how to get brand recognition in LLMs, a very timely topic that I know a lot of you have questions about.
So please make sure to ask questions in the chat and we’ll cover those as we go. And if you have to drop out early, we’ll be emailing everyone who registered a recording of today’s session. And now let’s get this rolling. Hey, Crystal.
Crystal Carter: Hi thank you so much for having me. I’m really pleased to be here.
Dana DiTomaso: Yeah, thank you so much for bringing this presentation. So I saw an earlier version of this presentation at Brighton SEO and it was one of my favorites. It’s like, absolutely, we have to have you on the webinar because I loved it so much. It’s funny, but also like important information for people to know.
So as I mentioned, please ask questions in the chat. We’ll get to those at the end or during if it’s really timely for what you’re talking [00:01:00] about. And yeah, if you want to get your screen sharing started, we can get rolling with the presentation. Okay.
Crystal Carter: All right. I’m just going to put it into slideshow and then we’re going to click on to slide share, share my screen,
wait for it. And
Dana DiTomaso: here we are. All right.
Crystal Carter: Great. Wonderful. Okay. Thank you everyone for joining us. I’m sure just like everyone else, you have been having these conversations around your organizations, about how you track this and how you’ve been Find out what’s going on with all of these LLMs, because everyone’s using them all the time.
I know I certainly am. I have one, I use the Co Pilot Assistant and it’s got a a lovely voice and I call him Bay Eye, for instance. So like, there’s just like a thing, like, it’s Everyone’s using them all the time and everybody wants to see what’s going on with it. So, I’m Crystal Carter, I’m the head of SEO communications over at Wix Studio.
I do enterprise SEO, I do thought leadership, I do comms leadership, I [00:02:00] do course instruction. You can see a lot of my exploits over on the Wix Studio Learning Hub and lots of other places as well. And I’ve been very lucky DiTomasso on more than one occasion, which is always awesome. Always a joy. So today we’re going to talk about lots of things AI, and I’m going to be helped by my twin sister, the other Mrs.
Carter, Beyonce Knowles, throughout this throughout this presentation. So today we’re going to cover why generative Search Engine Visibility Matters. And I’ve got engine in quotation marks because it’s not exactly a search engine. However, it’s, many people are using it in that way and I’ll get into that.
We’re also going to talk about how to monitor visibility and anyone who’s been following Dana is I’m sure very interested in that kind of thing. And then also we’re going to talk about how to optimize for generative search engines. And people have not quite settled on a title for, for this. So some people are calling it LL.
LLMO, LLM Optimization. Some people are calling it GEO, Generative Engine Optimization. There’s lots of [00:03:00] things like that. But we’re going to talk about visibility in LLMs. So one thing that’s really important is, yes, we are talking about that, but we are not going to talk about AI overviews. And the reason why, you know, you might be asking, I thought this was an AI conversation, so why aren’t we talking about that?
But the reason why is it essentially comes down to user journey. So LLM is the tool for the search. So if you were carrying out a search, if you say, you know, tell me what was the impact of Demi Moore’s speech at the Golden Globes, you’re using the LLM to give you information. You’re using it as a search tool.
If you go on to if you go on to Google and an AI overview shows up, it’s just a search feature that shows up. And that’s not the same, right. Part of that is because users choose an LLM for, for a search. However, users just happen upon an AI overview. It just sort of shows up. And that’s, again, is not the same.
Additionally, users interact with LLM responses. They go actually that’s not what Demi Moore said, or, you know, that, that’s not correct, [00:04:00] and, and I think you should do it this way, or or can you explain it to me like I’m five, or that sort of thing. And users just receive information from, from AI overviews, there’s, there’s not a lot of back and forth with it just now.
So that’s not the same either. So we’re going to be talking about optimizing for LLM brand visibility and essentially that is the process of increasing references to a brand or website in LLM responses and this is something that people should be thinking about for lots of reasons. So why does it matter?
Essentially, and these are numbers from October, I will be uploading these again, but basically the reach of these tools Is very prolific. So we’re starting to see, you know, millions of people using Gemini. And these are just people who are going directly to the app. This isn’t even people who are going through, you know, some of the integrations that we’re seeing.
So for instance, co-pilot is integrated throughout Office 3 6 5. Claude is Claude is things still standalone but using a few of other things for sort of things. But Gemini is integrated [00:05:00] throughout loads of loads of tools and Google continues to integrate that across. The people who are directly using these apps.
We’re coming in around 300, you know, just under 300 million for Gemini. We’re coming in 30 million for Copilot, 70 million for Claude, 60 million for Perplexity. And, you know, Of that, even though that might seem really big, ChatGPT is getting the lion’s share of this, coming in at around 87 percent of LLM traffic.
And that means that they’re getting, as of October, they were getting around 2. 6 billion visits, and which accounted to about 9 billion monthly page views and 6 minutes per session, which is a lot. And you might say
Dana DiTomaso: that. That’s incredible, the length of session time Google wishes. Right. Six minutes per session.
Crystal Carter: Precisely. So there’s a, there’s a lot of that there. And you know, these numbers are continuing to increase as, as all of these tools add more features. So Perplexity recently added something where you’re allowed to see sports type, sports type content. Sports Times and sports, sports scores and things in those, those elements.
Again, ChatGPT and Gemini is adding itself to lots of other features and all [00:06:00] of the sort of things. So people are going to be more and more familiar and more actively using these tools as well. And so we should expect to see these sessions go up. That’s a lot of time. That’s a lot of energy. That’s a lot of pages that people are spending.
So even though Google is currently still on top, we’re not seeing And we’re not quite seeing any confirmed drop in Google’s Google search share, you know, their LLM use is still growing rapidly. And so to the point where Gartner predicts that there could be as much as a 25 percent drop in search engine use by 2026.
And they’re also predicting that organic traffic could drop by 50 percent by 2018. And this all might seem really dark and sad and worrisome. However, we’re SEOs. We built the internet, we ruined the internet, and we can do anything. So, if you’re ready, I’m ready, let’s get into this. How do we optimize for these generative search LLMs?
Well, the first thing that we need to do is we need to understand what we’re working with and what kind of type, what types of LLMs there are. So, I’m, I’m [00:07:00] calling them static pre trained data LLMs. versus search augmented pre trained data LLMs. And essentially, when you’re thinking about your static data pre trained LLMs, you’re thinking about things like, things that are not necessarily pulling through, pulling through web data, and, or not consistently pulling through web data.
Now, Gemini, Gemini does this sometimes Claude, Claude does this, does this sometimes where they, where they don’t pull through information, but Perplexity is always online. Right. So they’re always search augmented. And then when you have, when you have Chat GPT, you have a split. So earlier versions, earlier models of Chat GPT are more of the static end.
And then later the, the Chat GPT4 is, is search augmented. And if you’re moving through Chat GPT4, Current such situation they will also for most people they’ll be you’ll have access to the search search there as well. Now the reason why I left chat GPT instead of the static pre trained LLM area is because for things that are like for things where you’re using a tool like say a GPT for sheets or where you’re [00:08:00] using an API integration You might not have the search enabled element if you’re using the chat GPT 3 for some of those, for some of those tools you, you will have the API connected to, to the earlier model because of cost reasons, because it’s, it’s cheaper to use the earlier model.
You can also upgrade to the other one, but you should be thinking about, about that when you’re thinking about the kinds of responses you should get. Similarly, with Copilot if you’re using the Copilot app, wait, my bae eye, as I was saying, that does, that’s not always pulling through live data. And if you’re using the Copilot within Office 365, then you probably are getting, getting web data.
And if you want to check that, you can literally just ask them, like, are you, are you able, are you able to search the web? Are you able to get live search results as you’re, as you’re going through? So let’s talk about the Static Retrained Data LLMs. The way we can describe these. is essentially you have data sets that are a fixed data training set.
And so, for instance, when ChatGPT first dropped, they had information up to around 2021, and that is the [00:09:00] information that they had, and they didn’t have anything past that. Right? So that’s their, their information training set. And they will have, their training sets will vary by platform. And I think that this, as we go forward, could even get more acute.
So I think at the moment, lots of people have lots of shared data, but I think that we might even end up in a situation where, where some of these things, tools have exclusive rights to certain data, and some of them don’t. That’s me, that’s me projecting, but that’s, that’s where I think it might go.
Dana DiTomaso: That’s, that’s interesting.
So you think like, for example, like the Reddit deal with Google, have a Reddit deal
Crystal Carter: with
Dana DiTomaso: Gemini?
Crystal Carter: Yeah, I think, I think we could certainly end up in a situation where like, in a place where data is king. right? Maybe content’s not king anymore, but data is king. Then we could seriously have a situation where let’s say ESPN or something has like all the sports data or the MLB has all of the sports data for, for for the, the 2025 26 baseball season.
And they will sell the rights, like people sell the rights to TV. Right to one of the LLMs and to not one of the other, to not, you [00:10:00] know, some of the other, potentially, potentially. I think that that’s, there’s certainly a potential potential there.
Dana DiTomaso: Yeah, that
Crystal Carter: makes sense.
Yeah. And then I think, so the other one we have is like visibility.
So you’re within the visibility for a static, a static data set model. Your visibility will change as the data gets updated. So if chat GBT is going, has data up to 2021, you have to wait. until they add more information before you can, before you will see any shifts in visibility. So, if you had a thing that happened in 2022, and you were looking to find it on something that was trained on data, but only goes up to 2021, you have to wait until they upgrade it to another set of data before you will see that information.
And then with some of these things, you maybe will have links and you maybe will not. For some of the ones that have, have data that’s a little bit more up to date they might have links within them, but they might not be live links. So that might be that they’re, that they have links that are referenced within their, within their data set, but it’s not necessarily the one from last week.
And so it’s a little bit like the encyclopedia or the phone [00:11:00] book. If you remember the phone book, if you’re old enough to remember when the yellow pages would drop and you’d get things from like Aardvark Taxis and that sort of thing, you’d have to wait until the next week. Year if your business opened in between, before you could get in the phone book.
So it’s a little bit like that. So you need to have a think about your training cutoff dates. And these will, these will vary and they will update. And if you are particularly concerned about where you are with the with your models, then you need to keep on top of those. I have an article that Dana will be sharing, sharing later on, which has links to where, where all of these LLMs have the information about their, about their datasets.
So, in terms of optimizations, don’t expect to see brand new info for some of these more static LLMs that aren’t, like, search enabled. Make sure that you’re monitoring the training updates for this, and also tailor your approach for each LLM. Now I’d like to talk to you about a little case study, because I’ve talked about this and people say, Oh, you know, it’s all very theoretical, you can’t do things with it yet, and stuff, and I’m like, well, I mean Kind of.
There was a project that I was working on a few months ago and this is something that we [00:12:00] launched from our Wix Studio team and it is called Site of Sites and it’s essentially a gallery of websites that are built on Wix Studio and that show that show some of the things that the websites can do and different that are built by different people and we launched this in the spring of 2024 and I was very curious.
to see how long it would take for this to show up in some of the LLMs, because it’s a brand new property and it’s something that we don’t always see. So I asked Claude, I said, what is Site of Sites? And they said, oh, it’s a concept in archaeology, blah, blah, blah. No. So then I asked I, then I asked ChatGPT, what is Site of Sites?
And they said, oh, it’s a web development concept. It’s metamorphic use. Yeah, not quite. Then I asked Gemini, I said, what is Site of Sites? And they said, the term can mean a couple of things. And one of the things it said is there’s literally a website called Site of Sites. And I said, okay. All right, that’s good.
I like this. And then I asked it again in 2024, and by, or in September 2024, and it said, Yes, Site of Sites serves a curated collection of [00:13:00] exceptional websites. Blah, blah, blah. Fantastic. This is exactly, this is content, you know, remixed from our website. Great, wonderful, fantastic. And because I am And because I am an SEO, my next question, after I got this fantastic information, was can I get a link?
And, and when it said to me, it said, I apologize for my oversight, there seems to be an issue with the URL I provided earlier. It’s not its own website, it’s part of Behance. And this enraged me, because Behance is a competitor, and I was very, very upset, because honestly, Behance, hey, guys, that’s not what we’re here for.
It’s about, we’re talking about site of sites. So I told this little robot, I said, bad robot. Absolutely no. The website for site of sites is@siteofsites.co, and it’s not part of Bee hands. Down arrow, no bad robot bat. So then in November, I asked it again. I said, what is Site of Sites? It said, Site of Sites is XYZ.
And then I said, what is the URL? And it said, the URL for [00:14:00] Site of Sites is siteofsites. co Fantastic. Thumbs up on that. Now for this, I would say this is a great, great technique for anybody who’s looking at, who’s looking at a small site, big site, whatever, what have you, to give feedback on good and bad responses.
If you go to the Gemini, the Gemini documentation, they actively say we’re actively looking for feedback on good and bad responses, and this is particularly important if they are giving you factually incorrect information, and factually incorrect information that is verifiable, right? So if I say that I don’t If I say that, if I say If I ask ChatGBT, is pineapple acceptable to have on a pizza, and they say yes, and I think no.
Like, that’s not really gonna get, that’s not really going to be something that I can argue with. But
Dana DiTomaso: actually, the answer is no to that. The answer is no! Nobody eats
Crystal Carter: pizza with, like, nobody eats pineapple with hot tomato sauce. No, that’s
Dana DiTomaso: incorrect. Yeah.
Crystal Carter: I’m glad we’re aligned
Dana DiTomaso: on this. Yeah. Okay.
Crystal Carter: [00:15:00] Completely.
Okay. So, so when we move over to the the search augmented pre trained data LLMs, we see sort of similar things. So all of the things that you were thinking about for some of the other things still apply, but there’s a few other, you know, a few other things to consider as well. So when you’re thinking about your data set, it will still have a fixed training data set.
Okay. However, Eric says it’s also no, I agree with Eric. Thank you, Eric. However, you’ll also have the search will be augmented by search engine results as well. So you’ll have the fixed data set, but you’ll also have the search engine results. And your visibility will thus be impacted by search engine results.
And we’ll also include links and references as a rule. So I was talking about Perplexity. Perplexity always has links and references. If you’re using SearchGPT, then it will include links and references as well. And the way we sort of think about this, Prometheus is the backing behind behind Copilot, and essentially these kind of work, they work in an interplay.
So you [00:16:00] have your orchestrator, you have the GPT, you have the, the way that the index is looking at their information as well, and they will cross reference all of this to give you your chat answer. So when we think about, when we think about you know, the search augmentation, it’s important to Look at where some of the information is coming from.
So with Copilot, it’s Bing. With Perplexity, they’re actually backed by Brave, which is a new search engine. So it doesn’t have lots of legacy data legacy rankings. Gemini is going to be backed by Google. They’ve not confirmed if they’re using other, other data as well, but they’re almost certainly using Google and then also ChatGPT also has Bing, but they’ve also said that they’re also pulling in other data points as well.
So this is important because a lot of people think that the, you know, the difference between Google and the difference between Bing are really, really similar, but actually they’re not. So if you are considering thinking about, you know, ChachiBT and Copilot, which are both backed by Bing, then you should also be thinking about Bing specifically.
So if you, for instance, I went through a, a data [00:17:00] set of around 400 keywords from, from a website that has lots and lots of ranking pages. And I found that there was actually a really big difference between the top ranking keywords on Bing and the top ranking keywords on Google. So of this, of this set of information, There were 10, 000 pages that were indexed on Bing, whereas there were 15 times more on Google.
And I found that for the same set of keywords, they had, they had 26 fewer top 10 keywords on Bing than they did for, than they did on Google. In terms of traffic forecast for, for, you know, when we compare, look at it on, across search engine tools or search tools. we saw that Bing’s traffic forecast was five times lower.
And in terms of number one ranking keywords, even though it was the same data set, I looked at top 10 keywords. I found that there were about 104 fewer keywords that were ranking, that were ranking top on Bing than there were on Google. So this means that if you are just assuming that your Google ranking will see you through to chat GPT, then you should make sure that you have a look through the [00:18:00] actual data.
rather than just assuming. So make sure that you have a look and treat, treat them specifically. Here’s an example of what, what we see for Site of Sites on Bing, for instance. So if I go to Bing and I look up Site of Sites, I can see that it’s pulling out a few pages there. The Collections page, the About page, the Main page.
And if I go to Copilot and I say, you know, tell me about Site of Sites. They’re pulling out some of these things and they’re pulling out the references. I mean, they’ve also added an ad because they’re Bing. But they’re pulling out these references that we also see in search. And so this, this is something that’s really important to think about.
So when you’re, when you’re seeing your performance on Google or on Bing, then it’s worth thinking about in terms of your internal links that you don’t ignore your search engine ranking. And that you query your, your, your core queries regularly and prioritize the pages that are showing in the LLM. So I know that those pages are showing on, on Bing search, but I also know that those pages are showing on, on Copilot already.
[00:19:00] And that means if I’m thinking about my internal linking, then I should try to prioritize those pages as the sort of, you know, the hub and spoke of your internal linking strategy. so that you’re linking pages out from there if you want to get more traffic from an LLM. Now let’s talk about optimizations that apply for every LLM.
So when we think about LLMs, we need to think about a few things. Optimizing for LLM crawl intent, optimizing for augmented LLMs first, managing your brand entities, and getting involved with the platforms. So let’s talk about LLM crawl intent. I think it’s important to remember that LLMs are not trying to click or rank your content.
LLMs are trying to interpret your content. And this is challenging. This is a little bit, this is very tricky for us as SEOs because that’s not, that’s not, hasn’t been the name of the game for, for us this whole time. But it’s, but it’s something that’s important to think about. So they have a different intent from human users who want to click on your information.
And they also have a different, different intent from traditional search, search agents who are [00:20:00] trying to, who’re trying to, you know, pull the, pull the full navigation and all of that sort of information. So where are all of these search bots? We have lots of information about web crawlers from, from the tools themselves, so ChatGPT Copilot, and Gemini, and Perplexity and Claude, they have, they have their own crawlers that go around and look at things on the web.
Again, we’ll send the article we’ll send the article after the webinar that talks about this as well. And we’ll give you links to all of them. And OpenAI has been really, really generous, so they’ve got all of their information about literally all of the IPs, and you can go through and have a look at all of them and they’ve listed them publicly.
Common Crawl has been used to train all of the LLMs, so if you have legacy data, if you did all of the things that Dana told everyone to do before UA was retired and stored your data, then you can go and have a look at some of the older versions of, of the of the Common Crawl bot that was crawling your site before.
You can look at those IPs. They’re currently using AWS. However there are some tools that are that are able [00:21:00] to to pick it up still. So I’ll show you an example of that later. So for instance, on the Wix Studio bot log report, which is super wonderful and y’all should all check it out. In the bot log report, we can see lots of bots.
lots and lots and lots of bots. And I had a look at at Common Crawl and how it performs across, across the website. And I also compared it to I also compared it to Google Google Bots mobile bot. So I’m just checking this one over here. And essentially when we look at this one, we can see these smaller ones.
I don’t know if you can see my mouse. Can you see my mouse? Yes. Yeah. Okay. So these smaller ones here, these are Google’s mobile crawl bots. Right? And this is a fairly small website. And these are, but these are Google’s Google’s mobile crawl bots. These are the common crawl up here. And this, this site is pretty small.
It’s literally like under a hundred pages and common crawl is coming to almost every single page, almost every single time, because they are have, they have a different intent. So you can see that they’re crawling the whole time. And when we pull, pull it out, We can see that they’re crawling lots of pages that aren’t necessary [00:22:00] in terms of, in terms of LLM interpretation.
I’ve seen that they’re crawling a lot of page 1, they’re crawling a lot of categories, they’re crawling a lot of hashtags, they’re crawling a lot of things that don’t really matter. And this, and so I think there’s lots of people who are going around saying, you should block every LLM, you should block ChatGPT from crawling your website.
I don’t think that that’s, I personally don’t think that that’s, that’s the, the move here. However, you should be managing your crawl. So remember, they’re not trying to click or rank your content. They’re trying to interpret your content. So they don’t necessarily need navigational pages in the same way.
They’re trying to understand that, understand your content. And so there may be, may very well be better pages than certain like pagination pages, et cetera, et cetera, that are, that they don’t need to see. So have a look through your, through your call, crawl and prioritize. So I went through and I, and I
Dana DiTomaso: blocked specifically.
If I can interrupt for a second. So are we going back to like the HTML sitemaps? It’s like, hey, here’s my most important pages. Hey, Common Crawl bot,
Crystal Carter: look at this. I mean, we’ve, we’ve, we’ve opened up a lot of can of worms in this talk. We’ve [00:23:00] already like ignited a pineapple pizza argument in the comments.
Hi, Haley. Hi, Eric. And sitemap. I’m not gonna lie,
Dana DiTomaso: like I don’t like them, but I also, like, understand the point of them. It’s like, I have, I’m really torn on them, but I also wonder too, and by the way, this is a great feature Wix Studio, if you don’t use Wix Screaming Frog has a log file analysis tool as well that is free up to a certain number of pages.
You can do this with any site at all. You just need access to your log files. And yeah, I also see this when I do log file analysis and non WIC sites that they just hang out in the tag files and it’s like, you gotta, you gotta control what they’re looking at. Otherwise they’re going to waste their time.
And this is where Crawl Budget most of the time is like. You’re not, you know, Amazon, you’re fine, but for smaller sites, like you don’t need to worry about crawl budget, but maybe you do need to start worrying about crawl budget when it comes to AI crawlers.
Crystal Carter: Right. Absolutely. And I think that this is, this is the kind of exercise that like, you don’t need to do this every month, right?
Like if you’re thinking about [00:24:00] retainer or like, let’s say you’re thinking like, I dedicate like half a day of SEO to my every month or something to, to what I do or a whole day or whatever. Like this is something that you could do every quarter. Like, you can just have a look, see what’s going on and update that.
But yeah, if they’re spending time in pages they don’t need to, like, I go for an all killer, no filler approach to, to crawl budget. And, and they don’t need to be wasting time on, on those pages. And I think that you need to be driving traffic to the content that makes it most clear about what you do.
and, and how you do it. And you can have a look at the kinds of, you can have a look at the kinds of sites that they are aware of, the kind of pages that they’re aware of on your site, and they will give you an idea of what they know about your site. I also have another tool that I’ve, I’ve I’ve added to the resource collection, which Dana will share later that allows you to get an idea of, of what a brand knows, or what, and LLM knows about your brand and you can ask them, like, what can you tell me about this brand?
What does this brand do? And then you can go back and you can, and you can, [00:25:00] you know, correct them with, with some of those things, but we’ll also give you an information about where they’re getting that information on your, on your website and which pages should be prioritized in the crawl.
Dana DiTomaso: Awesome. Yeah.
Like if you take nothing away from this presentation, this is, this is the thing. I
Crystal Carter: hope they take things. Okay. So let’s also think about optimizing for augmented LLMs first. So this is all about timing and this means that essentially you are, first you need to rank on search and then you’re, you will be visible in search in a search augmented LLM and then you will be visible in a pre trained model LLM.
So essentially this is like how the data flows. So, so this will be like later on. So essentially like when you’re visible in search, you can get, you can rank on search quickly. Like you can be indexed on a page very quickly or on, on Google, on Bing, on whatever. Very quickly, you can, and then you can be accessible to, to a search augmented LLM because you have been crawled on that website.
I mean, I, one of the things I talked about between the time that I started looking at site of sites, for instance, [00:26:00] and the time that we started showing up in in Gemini, for instance, I was doing work. on indexing during that time. So that’s what we can do. So in terms of thinking about ranking on search, you need to rank on search first, then you’ll be visible in a search augmented LLM.
And then when they update the, the training sets, you will be part of the, you’ll be more likely to be part of the, of the training model for the further on. And it’s also that the, the training model updates happened more infrequently than the search augmented update. So that’s just a question of timing.
Okay, now let’s think about brand entities. And for this, I need help from Abarbe. So this is really important. When I was thinking about, when I started thinking about this I first searched for myself on ChatGPT. And I got that, I essentially got the equivalent of that Mariah Carey meme, that I don’t know her meme.
And they just didn’t know who I was. Then months and months after that, they, they said, Yeah, no, we do know Crystal Carter. She’s that person from Wix who does this, this, and this. And I was like, oh, thank you. Wonderful. Fantastic. And [00:27:00] this made me realize that these things can change. So when I go, one of the things I had to look at was I had to look at the Barbie movie summary.
And the Barbie movie summary doesn’t actually say anything about a movie. Doesn’t actually say anything about a toy. Doesn’t say anything about Mattel. Doesn’t say anything about that stuff. It says Barbie and Ken are having the time of their lives in Barbie land, blah, blah, blah, blah, blah. Joys and Perils, etc.
I ran this through TextRazor, which is an NLP tool, which is an oldie but a goodie. They have a fantastic free demo that I use all the time that’s really, really great. And, and what it does is it pulls out the entities from your text. So what it did, what it pulled out from this one, even though it doesn’t say anything about Mattel, even though it doesn’t say anything about dolls or toys or anything to that effect, because you have Barbie in close proximity to Ken, the entities that it’s able to, to identify are Ken doll.
Barbie Doll, Mattel, Fashion Doll, Mattel franchises. They can understand that specifically just from this text. Why? Because they understand entities and they have named entity recognition. And this is something, and this is something that [00:28:00] you see materializing in LLM. So I went to Claude and I said, name a fashion doll.
And they said, Barbie. Great. I went to ChatGPT and I said, name a fashion doll. And they said, a well known fashion doll is Barbie. Great. Then I went to Copilot and I said, name a fashion doll. And they said, the most iconic fashion doll is Barbie. is Barbie. And then because they’re Bing, they put a lot of ads that weren’t even Barbie.
Then I went to Perplexity and I said, name a fashion doll. And they said, Barbie is the most iconic fashion doll and and well known fashion doll created by Mattel. Great. Wonderful. Then I went to Gemini and I said, name a fashion doll. And they said, here are some names for a fashion doll. And the first name that they said was Barbie, but they didn’t quite understand what I mean.
And they, and they asked me, do you have any preferences for the themes in mind for your doll’s name? Which was ridiculous. They did get there in the end. So overall, all the bots did pretty good. We’re pretty happy with this one. So how did Barbie become so good at LLM optimization? Well, the Barbie entity is pretty rock solid.
You know, they’re a very well [00:29:00] established IP. And if you go to their Wikipedia page, the first thing that it says is Barbie is a fashion dog. The first thing it says, found it. Founded by Ruth Handler, or created by Ruth Handler and run by the company Mattel. And this is something that’s important to think about.
LLMs are all trained on Wikipedia because Wikipedia is 66 million pages of human vetted, always up to date information that is very, very, very well organized. And I think that, and I think that this is something that people, people overlook. but because it’s all chained on Wikipedia, they’re trained on entities and thus on the knowledge graph.
So this is important to think about when you’re thinking about how you do, how you do your LLM optimization. So if you don’t have a Wikipedia page, if you don’t have a presence on Wikidata, if you aren’t connecting with entities, then this is something that will, that will that you should invest in. In order to do this.
So think about managing your entry on Wikipedia or Wikidata. Think about claiming knowledge panels and people might say, well, I’m not that big and fancy. I don’t have a knowledge panel. Well, guess what? If you’re in a local business, you kind of [00:30:00] do. So your Google business profile counts as part of your knowledge panel.
So that’s worth thinking about. Implement Schema to make sure that you’ve got, you’ve got good information and that you’re under, that you’re organizing your information in an appropriate way that that works well for that and make sure that you’re aligning your content and links to known entities.
One of the examples I like to give is if I said the phrase, Steve McQueen is a man known for his films. That actually applies to two Steve McQueens. There’s Steve McQueen from like back in the day, driving fast cars, etc, etc. And then there’s Steve McQueen who’s currently working, who’s, who’s, you know, does lots of artsy stuff and it’s like super cool.
Now if I said Steve McQueen is an Academy Award winning Turner Prize director known for his film, then that’s only one Steve McQueen. Why? Because I’ve got the Steve McQueen entity, I’ve got the Academy Awards entity, and I have the Turner Prize entity. So even if you don’t have a Wikipedia page, if you went to a college that has a Wikipedia page, if you work for a company that has a Wikipedia page, if you [00:31:00] are connected to certain entities, then it’s worth making sure that you’re defining those entities when you’re describing yourself and your business.
So make sure you’re thinking about that. Also, make sure you’re getting involved with the platforms. Now, there’s a lot of people thinking about LLMs, talking about LLMs, meditating on LLMs. We actually have to get involved, right? So, this isn’t something that is new. That we have to do you know, that’s, that can be really, really complicated.
We can make our own custom branded GPTs. And this is really important because they, when you go into chat GPT, any custom GPT that you’ve touched upon shows up in the top left of your, of your screen. Right. So that’s brand visibility. Okay, and not only that, but it also has a link in your website. So we have, we have a a web builder GPT that’s in ChatGPT.
It’s had over a hundred thousand chats. It was featured, a featured one in ChatGPT before. That’s one that’s available there. And these rank. So not only do you, do you show up in chat GPT, but it also ranks in search. So this is also ranking in [00:32:00] search. And this also gives, you know, two opportunities to increase your brand visibility, both using the LLM and, and because of the LLM.
So build custom GPTs for brand visibility with billions of GPT users. And, and they, they, they search for you in chat GBT and you’re there. That’s great that, that puts you as a forward thinking company and gives you an opportunity to connect with folks. And there’s also people who are actively embedding links in their outputs from, from their chat, GBT for the, from the custom GBTs.
And, and there’s, there’s one that I was seeing that was one that was a a visualizer creator and that that was super, super useful and it’s easy to do, to do it that way. And, and yeah, it’s really, really valuable. Another one to think about it is thinking about publishers. So the team over at Detailed have published this.
I’m sure we’ve all seen it. Or it talks about some of the top publishing networks in Google and how there’s, there’s not that many of them. That there, lots of people are in different groups. So there’s 16 different publishers who are on Google and ChatGPT is essentially taking, taking a Thanos approach, going around just capturing many of them.
So in 2024, [00:33:00] they partnered with three out of 16 top publishers. I think they recently just announced that they’re also publishing. I think they just announced another partnership as well. And essentially they have collabs with loads of folks. We’re talking Hearst, Fox Media, Conde Nast, Time, Atlantic, News Corp Prisa, Le Monde, etc.
But they’re also not the only ones. So Perplexity also has collaborations with folks like Time and Entrepreneur and for some reason the Texas Tribune, but hey. Go for it, Texas. And Der Spiegel as well. And so you might be thinking though, Hey, I’m not the Times. Like, I’m not Conde Nast. I don’t know how you think this is going to be relevant to me at all.
However there’s a couple of great ways that this can be useful. If you are a big player, you should consider becoming an LLM partner because this means that you’re in, that you can manage the data that you’re giving to the LLM. Think about Reddit’s partnership and things like that. Then this means that you’re able to become part of the training data and presumably you can manage and negotiate how you’re going to have your data shown there.
The other one that you should be thinking about [00:34:00] is prioritizing backlinks from those partner websites. So, This might, again, people might think, oh, but I’m never going to get a backlink from, you know, from Time or from the Atlantic. I’m a small player, etc, etc. However, a lot of those, those websites and a lot of those news teams have smaller, smaller offshoots.
So my local newspaper is part of, is part of the News Corp syndication syndication collection. So have a look at Which, which new, which news outlets that you have access to are part of, you know, which companies and where that would, where that will play in because that could give you, again, could you give you access to more visibility within LLS.
Perplexity Pages. Perplexity has pages and they create pages and you can basically create a page on a topic and it will include links. This is one for instance that has 57, 000 views and this is and they also rank. So they’ve started this in the spring of 2024 and you’re starting to see visibility across them and essentially they’re AI generated.
So you would ask for a [00:35:00] topic like let’s say, you know, best Beyonce outfits or something and then it would pull out all of the different links. through there, and it would start generating content for you, and then you can edit which ones will show there. And not only that, but those pages rank on Google, and you even get featured snippets.
So, build pages on perplexity is another one that you should be thinking about. Now, once you’ve done all of that, how do you actually monitor any of this? Good question. There is The fantastic Jess Schultz, and there are other folks who have also done this as well, but Jess was the first one that I saw who was really getting into this, who has a Regex, and this is it in in bigger form and it basically pulls out all the different, all the different AI tools, and it allows you to see what’s going, see what’s going on there, and you can use it as an explanation.
And if you don’t know how to do an exploration in GA4, talk to Dana, she can help you with all of that. I was going to say,
Dana DiTomaso: we also recently published an article on how to build a channel for AI traffic in GA4, so you don’t need to go into explorations, because no one likes doing that. Yes, and so that’s on our website, we’ll include that in the email that goes out afterwards as well.
And I will be reading that and I’ll be adding that to my deck. [00:36:00]
Crystal Carter: Wonderful. So yeah, but basically you can see this in, in ChatGPT, or so you can see this in GA4 whether you’re using Regex, whether you’re using Dana’s fantastic methods it’s something that you can have a look at. And there’s also folks who are offering this as a service.
So CIR Interactive are really leading the charge on this, and they have been doing some fantastic work around ChatGBT tracking for months. And they offer this as a service, and they can do this at scale for folks, and so that’s worth doing as well, and I think they’re doing something similar to bulk queries and things like that, and that’s worth looking at as well.
And also, on my article on the Wix SEO Learning Hub, which Dana will also share as well, you can find a link to my LLM tracker. Which also allows you to see your brand visibility in LLMs. And you can ask lots of questions about like, what is, you know, what, basically you can enter your brand name and it’ll say like, what is my brand name?
What do you know about it? What is good about it? Those, those sorts of things as well. And then you can get those [00:37:00] sorts of information about your brand and your LLM. And with that. I wish you good luck.
Dana DiTomaso: You always have such a great gif game in your decks.
Crystal Carter: I must say, I’m a little lazy. I picked Beyonce because there’s a million Beyonce gifs.
And also she’s very gif
Dana DiTomaso: able. It’s true. And most people know who she is. It’s perfect. It’s the combination of everything you need. All right. So we have a few questions from the chat and if you have a question you haven’t asked it yet, please ask. I am keeping an eye on the comments coming in from LinkedIn and YouTube.
All right. So first question, is Perplexity most updated when compared to Gemini and ChatGPT? I mean, they sort of have a live feed, so I would assume they would be the most up to date, right?
Crystal Carter: Perplexity I find to be really, really good. I recently, I mean, I, I sometimes check, check these things with live, with live questions or like really, really recent questions.
So I remember checking checking Copilot like two days after, after the Oscars and I said, what did Zendaya wear to [00:38:00] the Oscars? And Chad Copilot said, said, Zendaya did not go to the Oscars. She was hanging out with her boyfriend, Tom Holland. And, and that was really good because on all over Twitter at the time they were saying, they were saying, Oh, Zendaya wore this or Zendaya wore that.
And people were posting pictures and like news outlets were reporting that she’d worn this outfit, but they were all AI generated images. So I think that it’s worth the best thing, the best way to do it is to, to, Ask the same question across most of them to, to confirm, but they’re changing every day.
So I w I would probably say that the last time I checked, I literally just checked what was the response to Demi Moore’s Golden Globe speech just before we started on Chachi BT, for instance. And they were pulling out, Chachi BT was pulling out great references. So they referenced, you know, the kinds of folks you would expect to know about this.
Grazia, Buzzfeed, Glamour, New York Post, for instance, they pulled, pulled that out straight away. So I think that. It’s worth looking at that. Perplexity is moving more towards being like real time information. So they just launched their [00:39:00] their sports tracking information, which again is a real time, real time feed.
So I think that Perplexity is, is definitely trying to hold that space. Mm
Dana DiTomaso: hmm. Yeah. And it’s interesting about the sports stuff. So I thought about using, I play a lot of fantasy football, which anyone who follows me would know this. And I thought about using, I use Claude mostly for like fantasy football analysis this year.
I did not, but next year might be the year that I try, although I did fine without it. So maybe I don’t need it. I mean, it’s, you know, it’s all percentages either way and weird emotions. And I like this team, therefore they’re going to win. So yeah, we’ll see. Sometimes it’s a little faith. Yeah, just a little bit.
I, I, you know, I thought for sure Buffalo’s gonna win and they did. So hey, you know, you just gotta, and you gotta go through your routines. That’s obviously superstition is true, right? Very important. All right, next question from Sophie. She had a question. So this is when you were searching for yourself on Chat GPT.
And Is this because your own ChatGPT is learning about you, or do you test it anonymously? And I don’t know if [00:40:00] ChatGPT holds onto stuff that you teach it. I know that Claude does not, because I’ve asked it repeatedly, but
Crystal Carter: Yeah, so I, I tested it I tested it anonymously to, to confirm that. I, I, I, I’ve, I’ve Googled myself enough times to know that you should, that you should check those things.
But I’m also, I’m also the, the kind of person who will check things on seven different, seven different platforms and seven, from seven different devices. Whenever I’m testing anything, that’s just a little bit of a, of an obsessive testing method that I have. So I think that it’s, it’s worth, it’s worth doing that.
And I think that. In terms of training, I know folks who will go to ChatGPT and they’ll say, tell me about myself. And that’s, that’s, and that will tell you how much, how much ChatGPT is storing about you in terms of that kind of information. But I think that, If you want to see what’s going on with your brand then it’s worth, it’s worth logging in separately, logging in separately to see how that works and or [00:41:00] asking a few different folks to log in for you.
So in my, in my resource, my LLM Brand Tracker, I, I linked to a few different resources. There SpyFu, for instance, which has queried, it. ChatGBT, I think two million times or something like that, or, or many, many more than that. And so you can go to that and enter in your brand name and see what kinds of outputs it, it, it got about your brand.
So you can, for instance, enter in you know your football team and see what it says about that particular football team. And that is a more, again, a more anonymous thing. sort of element. And I think also one of the things I recommend. So my LLM brand tracker, I set it up so that it was like free to use for anybody to use.
However, if you were to connect it to something like like GPT for sheets, for instance, you can, you can scale that and you can essentially like enter in a enter in a like a formula and then it will, and then it will scrape it for you. And that again, will add some more, some more anonymity to it.
But I do think it’s worth checking and checking again on some of the [00:42:00] queries that are most important to you.
Dana DiTomaso: Cool. That’s great. Okay. Mackenzie would like to know, how do you recommend brands? Balance, Prioritizing Natural Language Content, and Technical Elements like Schema, Markup, and Optimizing for LLMs.
And it’s also been interesting because I’ve been seeing a lot of discussion on LinkedIn that schema doesn’t actually matter for LLMs. I’m like, yes, it doesn’t yet, but, you know.
Crystal Carter: Yeah, so this is, I’m having, I’m, I’m joining that webinar with with the team over at Sightbulb in February. And I will be doing some information in between.
And so, so the digging that I found was that, was that they’re not reading it in the same way. So, Yeah. However, Google has been reading schema, schema markup for years. Bing has been reading schema markup for years, and again, Google. Google and Bing are both using, are both using that to help guide their understanding of search.
So if it’s a search enabled if it’s a search enabled, GPT or so sorry, search enabled LLM, then as I said, first you rank on search. Then you rank on then you rank on you know, something else. And I [00:43:00] think that it’s important to remember that, that first, that if ChatGPT is part, or sorry, not ChatGPT, Schema Markup is part of you ranking first.
then it will contribute to your LLM further down the line. Is it a direct correlation in the same way that like, does, is Schema a direct ranking factor? No, but Schema, in terms of, in terms of standard SEO, Schema isn’t a direct ranking factor. However, the Schema, like the rich results show above everything else.
Therefore, therefore it’s a ranking factor. Right. Therefore you kind of rank ahead of everyone. So what I would say is that I think at this, at this stage, LLM optimization should be treated kind of similarly to CRO, in that, if you have high volume, high traffic and, you know, if you’re a big, big brand, like Wix.
com, for instance, like we’ve got billions, you know, millions, millions of folks coming to, coming to our, our URLs every day, we’re able to get a lot more signals and a lot more information about what’s going on with, with with folks who are looking us up on LLMs, et cetera, [00:44:00] than somebody who’s, who’s getting like 200.
Great. A day, for instance. And so similarly with CRO, you, if you’re able to test it very quickly because you’re like, okay, well, we’ve had 400, you know, a million hits on this and therefore we can take a decision that like, you know, 10 percent of people are more likely to do this or people are 10 percent more likely to make this decision.
Similarly, you will need some time to see, to see the queries come through the LLM if you’re a smaller team. So it might take you longer to do. So as I was saying, like with the, with the the bot,
You can update your crawl. You can update your crawl maybe that month. Maybe that’s what you do that month. Maybe the next month you spend some time on your Wikipedia. Maybe the next time you like spend some time going over your entities, making sure that they make sense. So I think that in terms of balancing, I think that it’s, it’s it’s an interesting approach.
It’s like very, very similar to CRO. If you’re a small team, then maybe add in a little bit of time. If you’re a bigger team, then maybe add in more time because you will have more queries that are more that are more relevant into to that space. If that makes [00:45:00] sense.
Dana DiTomaso: Yeah, that’s great. Thank you. Okay what would your recommendation be for brand optimization when you have brands that are easily confused due to their name similarities?
Crystal Carter: Yeah, so this is fun. So, so yeah, I think that, I think that this is, this is something that again comes down to thinking about your, your entities. And if you find that when people Google you, like, for instance, if you think about Chicago, the band, and Chicago, the city for instance, or if you think about, you know, other folks that have the, have the same kind of names you will know that, that when people go to, oh, and there’s Chicago, the musical as well, so you will know that when people go to Google, you you will that there’s that disambiguation that needs to happen that sometimes people need to go, no, I mean Chicago, the musical or no, I mean Chicago, the band.
And if that’s, if that’s the case, then you will know that that’s something that’s, that’s something that you need to think about. What I have seen, and this is again, why I think why I, I emphasize the entity element is that I will see, I spoke to somebody who is doing it, who had an agency and they were like, we never show up in chat GPT.
We’ve been trying to show up in chat GPT [00:46:00] for ages, blah, blah, blah, blah, blah. And I went to their website and they had one of those websites, which is classic where you can’t, you go to the website and you don’t know what they do. They’re like, I don’t know what you do. I can’t tell.
Dana DiTomaso: Like, just say it on the homepage.
Just say what you
Crystal Carter: do. Just say it! Just say it. So I’m like so I’m looking at their website and I can’t tell what they do. And similarly, I looked at some big, big agencies and they were saying like, we are an award winning agency that works across you know, 200, you know, 150 different countries. And we are, you know, great at doing creative campaigns.
And then another big agency, we are an award winning agency that works in 150 countries. And we did it. And like, basically they’re, They’re identical. So I think that uniqueness is really, really important. And again, defining yourself alongside the entities that, that are related to you. So, you know, we think about Barbie.
Barbie is a fashion doll, right? It’s a doll about fashion, right? And it’s owned by Mattel, which is also, which is also an entity. Ruth Handler, also an entity. These are, these are important. But I think about myself, [00:47:00] for instance, I’m Crystal Carter. I work at Wix. I went to Kenyon College. You know, I’m from, I, you know, I’m from California.
Like there, there are entities that I can attach to myself, even if nobody knew anything about me, even if I didn’t even have, have my own, my own page. There are other entities that can distinguish me. from other Crystal Carters. And that’s really important to, to think about. So how is your, how is your brand distinct and how are you articulating that on your, on your website?
And how, how can you make sure that when people look for you, look for you that they’re, that they’re able to find you. So if somebody like, let’s say, let’s reverse engineer it, right? If you were to say, I’m looking for I’m looking for a company that specializes in Claude based websites. Right?
There’s only gonna be so many websites. So many folks would do that and we’re gonna show up there. And I know that because I’ve done that query. So it’s one that you can, it’s one that you can, that you can look up. So, so reverse engineer it. And again in my, my brand, my brand, a [00:48:00] book or my brand brand visualization tool, I talk about you know, I have some of those questions that you can ask.
And again, they’re kind of seed questions and can also go to an LLM and say like, what questions would you recommend that I ask an LLM? to find more information to find out what this LLM knows about my, knows about my brand. So I think it would be useful to do that. And you could also take the same LLM document that I have and ask it about a, about an competitor and see, see what the difference is and see how you can make it more distinct.
Dana DiTomaso: That’s
Crystal Carter: great, great
Dana DiTomaso: advice. Okay, we’re going to end things off with a couple of non LLM specific questions for you. So, if there was such a thing as digital marketing crimes, what would be the number one you would want to prosecute? You’re now the Attorney General of SEO. What do you got?
Crystal Carter: We are on the internet, right?
And, like, as somebody who deals in analytics and metrics and things, like, you know that we can see what’s going on in a lot of ways. And maybe we can’t see everything. However, what I cannot stand is when we have a situation where [00:49:00] someone’s like, this is our goal, and I’m like, how are we going to measure it?
And they’re like, I don’t know. Michael, what do you mean? I don’t know. We have to know. We have to know. I need my gold star. I need to be able to say it’s gone up and to the right, or maybe it’s flat or whatever it is. I need to know if it worked. And if when we have goals that are there that are, you know, Just hopes and prayers, like hopes and dreams.
It’s not enough for me. Aspirational.
Dana DiTomaso: Yes, that’s, that’s what I call them. Yeah.
Crystal Carter: Yeah. We can’t, you know, we can’t pay the bills with hopes and clicks.
Dana DiTomaso: No, no, no, we cannot. All right. And what is a tip that you recently learned?
Crystal Carter: I mean, I was thinking about this, but like, I think I’m going to, I think I’m going to go back to, to my, my book.
Jack, I was really surprised at how effective the up, the up arrow was, or like the up thumb was. And that’s something that I recently learned. And it just made me think like with all of these machine learnings, like they’re machines and they’re learning. And so it’s really important that you take the time to make sure that they are learning the [00:50:00] appropriate things about you.
So I was incredibly, I mean, I honestly jumped out of my chair when I, when I saw that it had done the thing that I wanted it to, and that’s really important. And I think that make sure that you learn. Spend the time learning the tools and learning how they work and learning how they respond to, to inputs.
Dana DiTomaso: Mm hmm. Yeah. And I also like your tip about asking, just ask the LLM. How can, what questions should I ask you? Like it, it will tell you. It will. It wants to help you. Help it help you. Yeah. Yes. All right. Thank you so much again for joining me today. So, as we mentioned, we are going to send out an email afterwards to everyone who’s registered.
So if you’re watching this on YouTube, head on over to our LinkedIn page and make sure to register for the webinar there so that we will make sure to get your email address. Or you can just DM me if you want to get a copy of the resources. They don’t show up in your inbox. That is totally fine. DM me on LinkedIn though, please.
I don’t check. X slash Twitter. Anymore. All right. Our next webinar actually is a great jumping off point from this one. It’s so [00:51:00] I already, you know, was going to get Alayda who is, Alayda Solis is a fantastic speaker. One of my favorites as well. And I said to her, what do you want to talk about? Anything totally up to you.
And she said, what about SEO for brand visibility and recognition? Like, Oh, this is so perfect as a follow up to Crystal’s talk. So join us on February 8th, We’re going to have that published to our Kick Point Playbook LinkedIn page later this afternoon. Please make sure to register there. We’ll also have it on our YouTube as well if you prefer to follow the stream on YouTube.
And definitely watch your email for those follow up resources. You’ll get a replay of this webinar and a discount on Kick Point Playbook courses if you would like to learn more about using analytics for agencies or Practical GA4. And all the comments have been Deeply complimentary. Thank you again, Crystal, for joining us and this fantastic presentation.
I’m so glad you were able to bring it to our Kick Point Playbook audience.
Crystal Carter: Thank you so much for having
Dana DiTomaso: me. Yeah. All right. Thanks, everyone. Have a great day.