So What? How to Analyze If you Have a GEO Problem

Well, hey everyone, happy Thursday. Welcome to So What, the Marketing Analytics and Insights Live Show. I am Katie, joined by Chris and John. Howdy, fellas. That actually worked really well.

This week, we're gonna talk about how to analyze if you have a GEO problem, if you have been living under a rock, or willfully ignoring what is happening in the marketing space, which listen, I get it. Uh, then you have not heard about GEO. Otherwise, you are probably inundated with GEO requests. GEO stands, I believe, for generative AI engine optimization or generative engine optimization. Or it's an old car that nobody drives anymore, an old geo.

And so there's a throwback. There's a throwback. What we want to talk about today, you know, so a lot of executives, here let me just sort of set up the problem. A lot of executives, or even just like marketers in general, are saying, like, do we have a GEO problem? How do we optimize for GEO?

Well, first and foremost, you need to understand if there is even a problem to solve, which is what we're gonna do today. So, Chris, where would you like to start? Uh I kind of want to start by saying you we we have a solution for the problem, and that is our new course on GEO 101, but that's gig that's for the cart before the horse. Um, however, there's the URL if you want it. Um, we have to set up the problem first and help people understand what to even look for.

So fundamentally, there's three phases of GEO. Uh, there is what does the model, the AI model of choice know natively about you and your brand and your and your company and your products and services. Second, when the when an AI tool goes and executes a traditional search, does it find you? And then third, once it finds you, does what your website serves up actually uh is it useful to the AI model? And this is occurring in many different tools Gemini, uh Claude, Perplexity, Chat GPT, but the 800 pound gorilla is Google AI overviews in regular Google search.

That is, according to uh SparkToro and Datos, as of January 26, that is 93% of all search, period, and like 95% of AI search. So when we're talking GEO, we're really talking Google still. All right. Well, I mean, that's sort of an underwhelming fact, but at the same time, but at the same time, it's good news for marketers who understand how Google works and how traditional technical SEO and so what is it on site, off site, and technical. Um, and so if you understand those things, you're already ahead of the game because if 95 to 97% of what we're talking about is still Google, then you can just continue to optimize with a few extra steps.

But first, probably we should start with how to even how do you even do the analysis so that you know what you're measuring against to see if you actually improved. Yep. So there's two places that you want to spend your time. The first place is going to be ye old fashioned Google Search Console. When it comes to uh GEO and AI overviews, Google in particular has said that AI overviews counts towards your impressions, which is the blue line in your in your Google Search Console.

If you are, I'm sorry, the purple line. The purple line is impressions. Um if you are not showing up in AI overviews, your impressions goes down. The number of impressions you have goes down. Um, and so what we see here on in this lovely chart is that in general, for trust insights, because this is our search console, there's the purple line, uh, which is you know 37,000 impressions for that week, and you can see it just kind of going up and to the right, which is great.

That means that more people are searching for things for which Google thinks we are in part or in full one of the answers that regular search and AI overviews believes is the answer. Now, what this does not necessarily do um is does not necessarily they don't Google does not break this out and say, like, this is the percentage of it that is AI overviews. They just say this is what's going on. Um, so we don't know what percentage of people are are seeing us in in the quick results versus the percentage of people who are seeing us in the AI overviews. And so if I was to go to Google and just literally tap the word trust insights as a terrible example, and actually let's do this in incognito mode.

There's the quick answers, and then you get into all of the other stuff that Google loves to throw in here. So yay, we're first for our name, which is good. Um but that's this is the first place I would look. Which direction is that purple line going? Is it going up or is it going down?

Because if it's going down, Google is saying whatever it is that people are searching for related to you, you're not you're not it. You're not the answer. So, you know, we often use the example like, give me the top consulting firms in Boston, or give me the you know, consulting firms that can educate me on AI. Those are things that we would probably want to show up for. And so we've actually seen a few people come through our contact form saying they found us in um some kind of generative AI, whether it be open AI or some other large language model.

So, you know, we're doing something right, I would hope. Um, but you know, I mean, it's like, you know, John, when you see those, you know, what are you thinking in terms of oh, we need to get more people coming from that? Or do you, you know, just thank goodness people are reaching out at all? Well, yeah, you know, especially given all the trending for organic, you know, kind of cratering over the past couple years. But this is we are definitely riding a wave of success.

I mean, I can't, you know, deny that that's just what's happening because we see prospects just coming in and say, hey, we asked Claude, you know, who should teach us about AI, and Claude said it's you, said trust Insights. So it's both Claude, Gemini. Um, I haven't off the top of my head, I haven't seen any open AI in the past month, but you know, it's a source of inbound traffic for us, and that's you know, it's kind of classic inbound um quality leads, you know, to have the tool itself tell you it's like the old days of SEO, right? Like every vendor that wasn't first for their own name, they were third tier, and so to be at the top of the list, like this is great. Um yeah, we just need to continue on our path, and it's does it's great.

It gives us credence every time we talk about this. Yep. The second place you should look is in huge surprise, Google Analytics. So uh we've covered on past episodes of both the podcast and the live stream, how to set up this kind of dashboard in Google Analytics in the Explore Hub to look at the click stream uh clicks traffic um to your website from the different major uh AI tools. Gemini, chat GPT Perplexity, Claude.

This does not include Google AI overviews because they do not come, they're not specially marked, they will just show up as regular Google organic traffic. So this is people who are in a dedicated AI tool, and the amount of click traffic you get from those tools. So here we see I put up Gemini and Chat GPT just for comparison's sake, in time series for the last uh 12 months, and we can see in general, Gemini has kind of headed up and chat GPT has kind of kind of arced. Um part of this, and this is something that before anyone hits the panic button, the market share of chat GPT itself has declined by about 20 points in the last year. Uh, chat GPT used to be the only game in town with AI, and now of course you have Copilot.

Gemini is the second largest uh AI tool used in the marketplace, followed by Copilot, and then you know, all of the other ones. So if you see this in isolation, it looks like something's wrong. But if you see this in the bigger picture of the AI industry, you go, okay, that's normal because fewer people are using Chat GPT than they were a year ago. Now, I want to go back to something that you just hit on and just um you know expand on a little bit. So the AI overviews that show up on a Google search page.

We've all heard someone say, Well, my boss told me that we have to be showing up in those. And so I guess there's two questions. One is can you do anything about that? And two, if you can, how do you measure it? And I don't want to derail this episode too much, but I think it's at least worth acknowledging sort of what the difference is between those AI overviews and actually getting seen in a large language model for this generative AI search.

So there's as is typical with Google, um, there's three different ways that Google can recommend you using AI. The first, of course, is Gemini itself. Uh, you could when you ask it a question like name me a consulting firm in the Boston area that specializes in AI, it will give you an answer. The second is in Google itself. So if we go to regular Google, we say, um, what is a consulting firm that does AI consulting implementation and is in the Boston metro area, you get your uh that you get some ads, there's your AI overview, and then you get more stuff, and then the third way is AI mode.

Uh, and this is where a lighter version of Gemini and a lighter version of Deep Research essentially kicks off and does this. So there's three different ways that you can be recommended by Google's AI systems. Of these three, the AI mode and AI overviews show up as regular Google search traffic. They do not, they're not specially marked. Only Gemini will be marked as just came from Gemini.google.com.

And I think that's an important distinction for those who are watching who are struggling with pushing back on whoever's asking, well, I need to show up in AI overview. Well, that's great, but you can't measure whether or not that actually happens. And you Googling it for yourself is always just a bad idea. Stop stop Googling your own company and asking why aren't we showing up in the first spot? Just don't do that.

Yep. Another key piece, uh, this is from uh the same Spark Toro paper, which I I encourage you to go read because it's a hilariously fun read. I believe Amanda Nativitad wrote it. Um they tested out of curiosity to see how often uh the different AI tools would recommend the same vendors in the same order. And for Claude, it would take 1,492 searches before Claude returned the same list twice, meaning that Claude just gives you different answers almost every time.

For Gemini, for Google itself, it was 124 searches before you got the same answer. So even anyone who's peddling, oh, we can tell you your what your rank is in Google AI overviews. No, you can't. You can't predict that because the tool's probabilistic, and so you're gonna get different results, even for the same query. If you just literally copy and paste the same query over and over again, you're gonna get different results.

So I mean, that to me sounds like looking at traffic data from these large language models is largely unreliable in terms of you know, where does this fit in our customer journey? Should we spend more resources on optimizing for it? Like it sounds like in some ways it's kind of a crapshoot. It is. Yep um and because of the nature of large language models even small deviations in a prompt can have different results so very famously about a year and a half about a year ago um Tim Solo from HREFs and Olga Andrienko then of SEMrush each posted on LinkedIn this is quite funny um look our our tool is recommended as number one by chat GPT they each took screenshots Tim's query was what are the top 10 SEO tools what are Olga's were what what are the top 10 best SEO tools and each company showed up number one respectively with one deflection one tiny word difference what are the what are top 10 top SEO tools versus what are the top SEO tools that linguistic twist was enough to change the results so the other caution and we say this in the course the other caution is anyone who's telling you what they know what people are typing into chat GPT is lying.

I mean I said a little more diplomatically in the course we appreciate it but they're lying this complete bullshit um nobody knows that because these companies are never going to give away their most proprietary data which is the customer interactions they are never going to give that to anybody else so any anyone who claims ah we know exactly what people are typing to chat GPT bullshit absolute bullshit so a couple things on that um one you know so the to your point the easiest way to find out how people are finding you is to talk to them, and that's something that John does every day. The second easiest is to add it to your how did you hear about us options. Like you can either have it as like they just fill it in, or you can put it in as like I used AI mode in Google, or I it was the AI overview, or just like give people the option, and when they're contacting you and if they check the box, then add it to your numbers and be like, Great. Five people came in through the AI overview because they told us that's where they found us. The other thing, and this again is sort of a not to derail, but I personally, as someone who worked in um clinical trials, take issue with that, you know, head-to-head experiment because I would give more credence to it if they've used the exact same query, but changing the query from one to this means it's no longer apples to apples.

And so I just I just want to point that out is like that's great that they changed the wording and got different results. I would fully expect that to happen, versus if they had both used the exact same phrasing and then got different results. That to me is a little bit more telling about you know, making sure that the you know, uh understanding that the algorithm is unpredictable, versus well, yeah, you both did different things, therefore you get different outcomes. That's sort of just common sense of how those things work. You would think it'd be common sense, yes.

Um, the bigger question that people need to answer in the umbrella topic of how to analyze if you have a GEO problem is is what you're doing working, right? Because that's that is ultimately the question that we want to answer is yes, we can see from Google Search Console data, like yes, the numbers, the lines going up and to the right or down and to the right. Either way, you need to be able to say, okay, well, here's what we do. How does it relate to what we see on the screen so that we can know, should I do what should I do more of? What should I do less of?

Should I do this? Should I do that? And that is not visible here. You cannot see it here. You can only see the outcome.

So what you would want to do, and we're gonna have to go back to our roots uh for this one as it as a company, is you've got to look at your data that you have and determine is the uh is the stuff that we are doing, does it have a statistical relationship with the outcome? To do this, you kind of need to have the data. Uh it tends to work better. So, what data do you need? You need Search Console data.

Ideally, you have you've been uh archiving your Search Console data because you can download 16 months at a time. You cannot go back earlier than that. So you need uh ideally, you've been downloading this whole time. We've been downloading our Search Console data as a company now for uh literally eight years. So we have eight years worth of back data, and that part is important.

Uh you also need something that that reflects activity data, and part of GEO is knowing how AI models themselves work. Obviously, you have traditional search, but you also have all the other stuff that we've been telling you now for two years. Be everywhere you can be, be on every podcast you can be on. Make sure you're posting on on major social media channels and things, because if you don't do those things, models have no ability to gather training data about you. So the big thing is go and do those things.

Well, if you've done those things, if you're using a piece of software, so our recommended vendor for this that we use, well, that we we use it is um called brand 24. Again, we've had our uh brand 24 account on and running since uh 2011, uh before that predates trusted sites, actually, because uh I've been a brand ambassador for them for forever. And this collects stuff like our our stuff, but also uh social media, etc. etc. So if you were to download this data and do a statistical analysis to say, here's all the things that we do, what are the things that matter that that have a statistical relationship with the outcome?

If you were to do that, you would be able to understand that okay, yes, these are the things that make that move the needle for us. Question. Yes. Can you use generative AI to do that statistical analysis? Or should you hand that off to someone who actually understands how to do a proper statistical analysis?

Yes to both. I feel like you're getting the it depends. And I'm guessing that the it depends depends on the tools you have access to and how proficient you are in them. And then also depends on the skill sets you have on the human side as to whether or not it makes sense just to have a human do it, who can probably do it faster than someone who maybe doesn't understand the AI tools and kind of fumbles around, that would be me. Um so I'm guessing that it's a big, big old it depends on what you have access to and what your resources are.

That's right. Um, if you have access to one of a paid plan for Claude, Chat GPT, Gemini, uh, Quen, when basically any major language model provider, and you have access to a coding environment, you don't need to know how to code. You do need to have it to explain what to do, but you you don't need to know how to code. You would say, I have this big old pile of data. I want to do this kind of analysis.

Help ask me questions to on how to do this kind of analysis, and then go build me the thing so that we could you could see, okay, this is exactly what worked. Um, because that's what you want to know. You want to know is what we're doing working. So, what would this look like? It would look kind of like this.

You would say, we're gonna do some statistical analysis today. I think make this bigger here. There you go. Uh, data analysis yielding some Python code that will do data analysis. Here's the task in the data folder are these two subfolders the coverage folder from brand 24, um, the domain rating data from HREFs, because brand 24 doesn't collect domain rating data, and then Google Search Console data.

And you say, I want you to build a tool that will clean up and blend all this data together. I'm gonna suggest things like and tell it, we want to think about things like cross correlation functions and our correlation functions, which are statistical analyses to understand the time lag between when you do something and when you see results, gradient boosting and other machine learning techniques to understand what are the activities and coverage we've attracted yields results in Google Search Console, increased impressions. Because remember, going back to where we started this episode, impressions is when AI overviews in particular, and and everything else in Google, Google says, I'm gonna show trust insights. So Google says, You're the answer. Now the clicks say the user agrees, but the impressions is what we want to calibrate on.

When I put this into Claude Code, the coding environment, and you could put this in to co-work too, it'll just take longer. What it does is it looks through all the data and says, okay, I have some questions. I have a lot of questions. And it asks you all of those questions, like, what about this? What about this?

When when I did this last night, I was um it was asking me questions. I'm like, I don't even know the answer to that question. So I would like uh explain to me what that is. So, John, you're a redis resident statistician. Have you heard I've never heard of this term before?

Have you heard of Granger causality effect? No, no, that's beyond my stats one-on-one. Same with me. I was like, I've never heard of this. What the heck is this?

And Granger Causality Ect is looking at a time lagged or a time either lagged or time leading indicator to predict an outcome. And the uh the Claude explained it to me this last night, says, let's say you want to predict uh how uh when your roommate's gonna eat pizza, and you happen to notice that every week on third uh your roommate tends to order pizza on Fridays, not always, but tends to. But every week, if they are going to have pizza on Friday, on Thursday, they buy beer. And so if you predict based on the beer that your roommate buys, um, because it has a strong correlation to the purchase of pizza the next day, you could say I can predict pizzas based on when my roommate purchases beer. So in the case of impressions, if I have all this other data from my coverage, can I run an analysis to find out does coverage predict increased impressions?

In the same way that beer predicts pizza in a roommate example. That's Granger causality analysis. I didn't know this existed. I mean, is it I'm okay? I will I will refrain from picking that example apart.

Um question though. I thought that, you know, in terms of data and statistics and code, I thought that the tool of choice would be something like a Google Colab. Is that no longer the case? In this particular example, no. And the reason for that is the sheer amount of data and and what it looks like.

So to give you a sense, just the hrefs data spans 13 files with about a hundred thousand rows. Um, Google Colab has limits. Um when you look at Search Console data and you start seeing you know years of data, you're way past collab's limits. And uh you're gonna blow up its memory too. So using Python code on your computer is the best way to do this because Python code can handle data sets that big, especially if you have a great CEO who bought you a very nice laptop with a lot of memory, uh, you can handle it.

Um there's so much data here that having clawed code build this or clawed co-work or anti gravity or open AI codex. Again, I wrote zero lines of code. I but I had to give it detailed instructions about what to do. And I used the brainstorming skill to say, hey, ask me questions. What am I forgetting?

What what don't I know? What are my blind spots? And things like Ranger causality analysis was one of my blind spots. It's like you don't need to do this super heavy, crazy, you know, um extreme gradient boosting technique that you mentioned in your prompt. That's not the right tool.

The right tool is called Granger Causality Analysis. I'm like, oh okay. So fun when Chris gets humbled by the AI. Uh-huh. And so what did we end up with?

We ended up with an analysis saying, you know, what does the data actually say? Does the coverage we earn translate into search visibility into AI overview impressions? The short answer is yes. Um we have a dividing line here uh before AI overviews went into market and then after all everything you have to write is after AI overviews went into market. Impressions is the blue line, coverage is the orange line, and it says eight since AI overviews launched, um, we have seen an 84% increase in dead impressions, 73% increase in daily clicks, and we have we've matched on 758 days.

And if I um if you look at the statistical distribution, uh you see market changes at in impressions after AI overviews came out. So AI overviews has been it has been great for trust insights. Uh looking at the coverage over time, how long does coverage take to work? Uh, in general, 27 days. So after a piece of coverage comes out, we see impressions go up.

Um there does it actually drive impressions? Yes, yes, it does. Um, the rhythm, what does it reveal? It reveals um essentially that there is a strong correlation between these things, and then how well can we predict with high accuracy? What to do more of?

And then I had it go in and break down. We have YouTube, social media, and our own stuff. What works? The things that work, YouTube works really well for us. And critically, YouTube works fast.

So when we post on YouTube or we get coverage on YouTube, because it takes into account our own channels, we see a significant uh lift. We see lift and we see that lift happen very, very quickly. On our own property, we see um that, but it's takes a couple of weeks. Social media is kind of iffy, like broad social media, particularly X, you know, because we have a lot of X uh formally Twitter data from prior to uh it being taken over and turned into uh deep fake porn site. That data no longer really correlates, but it screws up our predictions.

So essentially what it's saying is for our GEO, coverage is great. YouTube is where it's at, which again we've been saying for two years now. YouTube is where it's at, um, followed by your own properties and doing some cross-promotion, particularly on LinkedIn. Please send me that report. I can't wait.

Well, you know, and I know for the sake of the live stream, you were going through it pretty quickly, but I was like, oh, wait, oh, but you know, and so there's I there's so many things in here that I want to dig into. But what I like about this is it's not written in an overly scientific statistical way. It's the big picture. AI overviews doubled our visibility. Okay, great.

I can understand that. You know, and it's like I like that the takeaways are very quick and to the point. And that's something that I really appreciate about because this statistical reporting can feel very overwhelming. And so I also feel like I can look at the graphs, but I don't really need to understand them or dig into them if I have, you know, the headline that tells me what the so what is. Yep.

And it and it says it even does so Claude. The way I use this was I had Claude do all the data analysis first. Claude produced the graphs and the tables. And then I said to Claude, now write the report for Katie using the co-CEO gem. Uh good the co-CEO skill.

That said, tell us what's going on, and most importantly, tell us what we should do more of and what we should do less of. And what we should do more of is do as much as we possibly can on YouTube because boy does it work. And um, I didn't break it out because I didn't have time last night. Um, but what I noticed was the daily posting that I do on YouTube on my personal YouTube channel, which has links and mentions to trust insights throughout in the copy and also in the the video itself, um, seems to have a slightly outsized impact in terms of helping increase the visibility of trusted sites, getting people to search for us more. So my foaming at the mouth, waving a knife around in the kitchen on Sundays to create the weeks worth of videos, because we put trust insights in it so much, including in the the the you know, 90 second closer, the more I can mention trust insights and what we do and who we are, the more it helps contribute to our overall visibility.

It's funny because we saw this with our email channels a few years back. Um, I think you call it something like the halo effect. Um, it's how you describe it to our clients. And so now, with the emphasis on YouTube, which by the way, is a Google property. If you didn't know, um, it makes sense that you would also get that halo effect.

But please don't scare people away if you accidentally, you know, nick yourself with a knife and then your videos are just bloody. Like, don't do that. My view counts would be great though. People love a good disaster. Oh my goodness.

But I would recommend to the to this uh episode's title, how do you analyze if you have a GEO problem? Do this analysis, right? Sit down with a clawed code, sit down with your data. You don't need to do the math. You give a tool like Claude Code or uh Gemini Code Assistant or uh OpenAI Codex, right?

You give it the data and you give it the recipe and you say, help me do this. As long as you are clear. This is classic 5P framework by trust insights, right? Purpose. You've got to be clear about your purpose.

People, it's you uh doing the work. Process, you ask the AI, how do we do this? What are the best statistical methods for achieving the purpose? Right? Uh, you choose the platform, it's going to be whatever you're paying for.

Um, and the performance, the outcome is a report like this to say, this is what I expect. I expect you to tell me what's happened. I expect you to tell me the so what, and then I expect you to tell me the now what. Like, what should I be doing? It's great that I have all this data.

What should I do? And if you have that all written out, which is essentially what I put in my my recipe prompt, that big long page, the tools can do a phenomenal job, and all of it, all of them will do a great job with it. You just have to have the five framework by trust insights written out very clearly so that the agents can don't have to ask you questions for every single thing. Well, and I think it also uh is worth mentioning that if you don't know how to exactly structure the prompt to do, or if you don't have the vocabulary to say this is the type of analysis that I want to do, that's okay. Go into that conversation with the LLM saying that, like, I know what kind of outcome I'm looking for.

I don't know exactly what kind of statistical analysis this is, but here's my five P's. Here's what I have. Can you help me put that master prompt together? Because again, it's a really good uh companion tool. It's a really good supportive tool.

You just have to start that conversation and be really honest. Like, I know what I want, I don't quite know how to get there. So can you help me put that together? Don't start doing the analysis. I just want to build the prompt.

Exactly. And it will do a they will all do a great job. Any of these coding tools are using a good model. So use the the best model that you have access to on the plan that you're paying for. Um, the free plans will not be able to do this.

Um, there's there's you will run out of uh compute time before they even get a quarter of the way through it. So unfortunately, if you're on a free plan, pay the 20 bucks a month to get on someone's paid plan. Um, if I had to, if you were to hold a uh a frozen mackerel to my head and make me choose, I would say choose Claude because Claude's models, even though Claude's 4.6 model is such a good coder and it's good at asking questions, and clawed code is probably the best coding environment right now. If you're going to build the stuff. Um, so if you had to pick one, that's the one I would pick if you have a choice.

I feel like we need to get our hands on a frozen mackerel. I have some in my freezer. Actually, hopefully no one holds it to your head, but I do think you know, people are gonna that's a natural question people ask. It's like, well, there's so many choices out there, where do I start? And so if you have the opportunity to make a decision, a lot of people unfortunately are given the tools that they have.

So if you're a Microsoft shop, then you likely uh have to use co-pilot. Um to your point though, if they have a paid plan, could they do it in a co-pilot or another tool that isn't clawed? So if you have access to if you have uh Microsoft co-pilot, ask around your IT shop if that also gives you access to GitHub co-pilot, which is Microsoft's coding environment. Uh, and if you do, if you're allowed to use it, that would be the one to use inside the Microsoft environment. That makes sense.

Um, so we did the analysis, we determined whether or not we have a GEO problem, right? Like, I guess so what do we do now? So if this graph didn't look like this, if it was the inverse, it was like, hey, you're not showing up, then our next steps would be to say, okay, in the data, what does work? Like, here's the things we're doing that's not working. Is there anything that hints at what does work?

And I would also say take the Trust Insights GEL 101 course at TrustInsites.ai slash G E L 101 to learn, okay, what are the the things we should do? It's a it's a six-lesson course, takes 90 minutes, but there's a crap ton of prompts and stuff in there to help you get started and to do basic analysis. So you don't have to do the crazy coding thing. There's some basic analysis and basic measurement uh tools in there. And then we know beyond a shadow of a doubt, the things that do matter in terms of visibility.

Uh, one of which is YouTube. Um, and again, we've been saying this for two years now. Show up on every podcast that will have even if they have two listeners, who cares? As long as that podcast is putting episodes up on YouTube, uh, that is the place to be. And when you're on any show, name check yourself as much as you can around relevant terms, talking about the 5P framework by trust insights, et cetera.

Yes, you sound like a narcissistic egomaniac, but the point is not for it's not for the humans listening, it's for the machines that are listening to try and train them. And, you know, uh, Chris, as you mentioned, a lot of that so what, what do I do now, all lives within the GEL 101 course that uh we've recently published. Um and includes, you know, so you talked about it's for the machines. We dig into all of that. I think we actually did do a uh so a live stream.

Um, if you want to go back through our playlist, if you go to trustinsights.ai slash YouTube and go to the live stream playlist. I believe we did an episode around just that, how to basically create content for the machines, where to put it. Um, but if you want that more focused um instruction, so definitely take our GEO 101 course where we walk through step by step what that looks like and how to measure your results. Exactly. Um as you do this, if you fire up clug code, and we did an episode on the live stream for about getting started with Claude Code if you want to go back and watch that one.

Don't be afraid to ask questions of these tools. These they're not gonna yell at you, they're not gonna belittle you, uh, unlike me. And they will say, you know, this is a good idea, or that might not be a good idea, or have you thought about this? Um, like I said, one of the things that I thought about was, well, you know, should I differentiate between what I do personally for fun uh on my weekends on my YouTube channel stuff versus official trust insights content? And the answer is yes, that's a that's not a bad thing to analyze.

Um, when I look at third-party coverage we've gotten, you know, does it move the needle? The answer is no, it doesn't. Um, some of the third-party coverage we've gotten just isn't relevant enough compared to the stuff that we do, which I think is a very interesting finding because we doubled our visibility in in Search Console since AI overviews came out. But the majority of the influencing factors is us making stuff, which means that's kind of weird. It means that we have more control over how we show up than we did prior to AI, which is a really weird thing to say because for years we were like, you have no control over what Google does.

Well, now it turns out that because to your point, Katie, because Google owns all these properties and they're building their models off of them, the more you feed Google directly through channels like YouTube, the better you'll do. And if you need a quick Search Console refresher, because it's been a minute, um, and for some people, they don't even know that the data exists in there. We do have a course for that, which you can get at TrustInsights.ai/slash search console. Um, sure, to say we have a course for that. We do.

Well, we have a framework for that, we have a course for that, we have a John for that. Uh John, what's I mean, what's your takeaway on all of this? You know, I know that you've been hearing from a lot of um prospective clients, like, I need to know about GEO. Like, what kind of questions are they asking? Yeah, it hasn't been as much of like, hey, I need X, Y or Z done.

There's been a lot of, we don't know anything that's going on with any of this. Can you just give us some direction of some kind? So that's good that um there's not a lot of SNECLA. There has been some snake a lot, and there's an interesting take on this. I want to make sure that you guys are square with this.

But from the stuff that I've seen in the course, I haven't finished it, but at least I've seen enough to know that there are a lot of people promising that they'll deliver specific results, which is we, you know, as we've already talked about here, is complete ridiculousness. But there is the fact that we know how these models work. And if you don't do what's in the course, we can guarantee that you won't be in there. So and I think, well, no, and I think that that's a good way to look at it. I think that you have to do something, but you have to know the right things to do.

Otherwise, you're gonna stop showing up at all because other people will start doing things. And so you may, you may still show up in Google searches, and that's fine. But if you're looking to optimize specifically for those AI searches, those different large language models, and looking at that data in your Google Analytics or your Adobe or wherever you're looking at your web traffic, you have to do something. It's not just gonna magically happen. Like you have to actively do something.

So you're absolutely right. Um, doing nothing means getting no results. Well, and a lot of it is like classic technical SEO stuff of the worst place you want to be is be cranking out a lot of great videos and stuff that you're putting on YouTube. But because you've got some website that was designed like eight years ago by some contractor that you can't even get in touch with, like you could be missing metadata and other structural data that prevents that from getting into YouTube. Like you could be you know sabotaging yourself, which is the kind of thing you need to stay away from.

We have a tool for that, and it's free. It's called AIView. If you go to TrustInsights.ai slash AIVU, you can put in a URL. I'll put it in the Trust Insights website here, and put in your email address. That is now required because we're markers.

Um, and uh it's gonna go through and look at the page as machines see it. So, what are the things that it sees on your page that would indicate like yes, you're being found or no, you're not being or not being found. So let's see how we're doing here. It's calculating. And there's an ad for the course.

Um and to be clear, when you say you put in one page, this free tool, uh totally free, gives you literally one page. So this is not analyzing your entire website. This is analyzing one page at a time what you've put in. If you want to do something like your entire website, you can reach out to our chief statistician, John at TrustInsights.ai slash contact, and he'll set you up. Yeah.

So in this case, to John's point exactly, um, this will say, like, hey, your descriptive link text sucks on the Trust Insights homepage. You have click here, get rid of link text like that, and make it more descriptive to improve the alignment between what is the text that's on the page and the actions you want people to take, right? So some basic things. Structured data looks our structured data looks in pretty good condition, our metadata looks very good, and again, these are all data structures that AI tools can see. And then you can see what we call linearized content, which is what does a page look like when a machine lands on it and looks at it.

And that's what you want to uh to John's point. You want to make sure that your site, when a machine retrieves it, goes, yep, this is relevant. We got a tool for that. We got a lot of shirts. We do.

That's okay. We never have to make decisions about our wardrobe again. This is true. Any final parting thoughts? But now that we know that we know that for us, we do not have a GEO problem, but we do have to continuously try to create more stuff so that we keep feeding the machines.

If only in the background, I had just created some sort of a strategic plan for that. I used a framework. Um, no, it's you know, start start to actually take a look at what you can measure. Definitely start with the free tool, trustinsights.ai slash AIView. If you see, oh man, we didn't get a good grade.

There's a lot that we have to do. Definitely reach out and contact us or take our course, trustinsights.ai slash G E O 101, which is, you know, I hate trying to say GEO, but here we are. I didn't choose it. Nope. All right.

That's gonna do it for this week's episode. Thanks, folks, for tuning in, and we will catch you all on the next one. Thanks for watching today. Be sure to subscribe to our show wherever you're watching it. For more resources and to learn more.

Check out the Trust Insights podcast at TrustInsights.ai slash TI podcast at our weekly email newsletter at trustInsights.ai slash newsletter. Got questions about what you saw in today's episode? Join our free Analytics for Marketers Slack group at TrustInsights.ai slash analytics for marketers. See you next time.