Get Found & Recommended by AI: Understanding GEO

Download Slides | Download Transcript | Download Audio

Machine-Generated Transcript

[00:00:00] Today we’re gonna be talking about generative engine optimization, or GEO, how we can get these large language models, LLMs, like ChatGPT and Claude, to know us, to cite us, to recommend us, and maybe even drive traffic to our websites or retail spaces.

[00:00:20] Now, there’s a lot of mixed feelings about AI these days, and that’s probably putting it lightly. There are real concerns from data centers, privacy issues, environmental issues, job loss, Skynet, and those are important conversations to have, but that is not the focus of today’s talk. Today we’re gonna be focusing on just one thing.

[00:00:44] Your customers are already using ChatGPT and Perplexity and Google’s AI Overviews. They’re asking these tools for recommendations. That is not a prediction of what’s coming next. That is what is happening today, yesterday, and every day going forward. So the question in front of us today is not whether AI is good or bad for humanity or somewhere in between, but does your business show up when somebody asks AI for a recommendation?

[00:01:10] So let’s get into it Here’s what I’m hoping to accomplish in the next hour. Let’s talk about why this matters, because we are going to need to put an investment of time, energy, and possibly money into our GEO efforts, so is there gonna be some sort of ROI? I’m gonna share with you how AI works today, how AI search works today, how you can benchmark where you are right now, what are the steps that you can take to improve, and then what to measure to see if the work that you’re putting in is actually making a difference.

[00:01:47] Before we jump into that, you don’t know me, if you haven’t been on my email newsletter for long, you may not know who I am. My name is Rich Brooks, and I am the president of Flyte New Media. We’re a digital agency located here in Portland, Maine, and we have been in business for nearly three decades.

[00:02:03] Decades, helping small businesses and medium-sized businesses with branding, with web design, and with digital marketing, things like SEO, GEO, paid search, paid social. I’m also the founder of the Agents of Change, which was, and perhaps will be again, an annual conference here in Portland, Maine, and it is still a weekly podcast.

[00:02:23] We just launched, I think it was episode 635, 36, maybe a little bit higher than that. Been doing that for over 13 years, where I interview digital marketing experts week in and week out about different ways that you can reach and engage your audience. I’m also the tech guru on 207, which is the evening news program here on the NBC affiliates in Maine, where I do hard-hitting news stories like what are the best recipe sites online, and how can you take better pictures with your smartphone?

[00:02:53] And lastly, I wrote a book a few years ago now called The Lead Machine: The Small Business Guide to Digital Marketing. So that’s where I spend a lot of my time. I’m very passionate about this, very passionate about marketing, and very passionate about AI. Since I was a kid watching Jetson- Jetsons reruns, I’ve always wondered what it was like to have an intelligent conversation with a robot, and now I’m living through those times.

[00:03:17] So I wanna share with you some of the research that speaks to the impact that AI search is having on your business. So let’s start with one stat about local business. What percentage do you think of consumers use AI tools to find local recommendations? 45%, which puts it in third place behind Google and Facebook.

[00:03:44] For those of you in B2B, you may be wondering what percentage of business buyers use AI in the buying process. How many of them turn to AI at some point during their decision-making? 94%, almost everybody. 19 out of 20 business professionals use AI when they’re trying to find a product or service Let’s talk about conversions.

[00:04:08] W- Will people take action when they actually follow one of these links at your website? Adobe found that AI-referred visitors to US retail sites converted at 54% better than non-AI referred traffic. So when I say AI-referred traffic, I’m not talking about bots that are going to your website by themselves.

[00:04:29] I’m talking about somebody asked a question of ChatGPT or Kloud or one of these other platforms, they included a link, and this person clicked on it and visited your website. Not only did they convert 54% higher than the average traffic going to that website, but they generated 53% more revenue per visit.

[00:04:50] So really good conversion rates there. And then you may say these, these stats may be misleading,” or maybe they’re not dramatic enough for you because … But here’s the thing, AI referral traffic is only part of the story. Because 50% of people are more likely to visit a brand’s website within seven days after ChatGPT recommends them, often via a branded search So in other words, somebody’s doing a conversation with ChatGPT.

[00:05:20] They’re having a… maybe it starts with something like, “How do I get rid of my crabgrass?” And it elevates to, “What are some local lawn care companies near me?” ChatGPT makes the recommendations. They don’t click on any links. But a few days later, within the next week, they do a Google search on your company’s name.

[00:05:39] They find you, and then they go. You didn’t even know that ChatGPT played a role in this, and this happens time and time again. Oh, wait a sec. That 50% is wrong, though. It’s 250%, so two and a half times more likely that they’re able to go to a website based on a search you didn’t know and you’ll never know that they used AI to find you, but there it is.

[00:06:03] Very powerful numbers showing us the impact of GEO or of AI search, and you’re not even seeing it even with the best measurement tools. Now, getting back to that conversion conversation slide that I showed a few seconds ago, why is it, and this- we see this across the board, that AI-referred traffic converts at a much higher rate than, say, traditional search traffic or traffic coming from another website?

[00:06:29] There’s a number of reasons that this is happening. First of all, AI sends transactional traffic, transactional meaning that somebody has an intent to buy something. So when somebody has the intent to buy, they’re already further down the funnel, and they’re going to make a buying decision. They arrive later in that buyer journey.

[00:06:49] They’ve had the conversations. They’ve done the research. When they click on that link coming from an AI referral, they’re ready to make a purchase. That may be buying something on your website, or it might be about filling out your contact form or even picking up the phone And lastly, and this is just my hypothesis because I go through this myself, when you do a Google search, I think there’s an understanding that even with these 10 blue links, I still need to do some vetting and research on my own.

[00:07:16] Google’s just, before AI overviews at least, Google’s just pulling in places for me to continue my research. With AI, there’s this sort of understanding or belief in AI’s magical abilities that it’s already figured out what the one right answer is. So whether it’s giving us a short list or it’s giving us a singular answer, why do I have to do any more research when didn’t AI already do that for me?

[00:07:41] So again, these are all reasons why AI traffic is likely more val- or is more valuable than other types of traffic coming to your website. So now that we know that AI search traffic is better than traditional search traffic, we just need to replace all of our SEO tactics with GEO tactics, right? It’s not that simple or that straightforward.

[00:08:01] First of all, be- besides all those or despite all those stats that we just looked at, AI search is just a fraction of traditional search today. If you would somehow remove all of your SEO and replace it with GEO, you would get high-converting traffic, but not enough yet to sustain your business. And there are other challenges as well.

[00:08:26] First of all, excuse me, it’s a brand-new discipline. We don’t have enough data yet to know exactly what works and what doesn’t work. We have early research, we have informed hypotheses, we have best practices, but almost anybody who says they’re an expert isn’t. We also have multiple platforms. When it comes to SEO, all you really need to do is pay attention to Google.

[00:08:49] But in AI search, we have ChatGPT, Claude, Gemini. We also have AI overviews in Google search results, and then we’ve got AI-powered search engines like Perplexity. These models are going through updates all the time, and not only does a new model mean that it’s got more information and data that it can reach without doing any searches, it also means that the way that the model thinks is different.

[00:09:13] So even if we figured something out, the next model update might give us a completely different engine. And unlike the early days of SEO, and to some degree even today, this is a black box. In the early days of SEO, we got a list from Google that told us exactly what people were searching on when we were looking at our analytics.

[00:09:34] Even today, where we don’t get all that data, we still get a sense, we still get some reports on the type of searches people are doing to end up on our homepage or one of our service pages. There’s nothing like that in these in these LLMs. They just don’t share that information with us. And then also a lack of consistency because LLMs are probabilistic, not deterministic, and we’ll take a look at that in just a minute, and I’ll explain a little bit more about what that means.

[00:10:04] But we definitely have some challenges with GEO where we are today It might be helpful to explain a little bit about how AI actually performs searches and provides the answers and recommendations that we all get when we turn to a pro- a platform like ChatGPT. And most all of these platforms work in a similar way.

[00:10:26] AI has knowledge that it’s been trained on, or pre-trained, as it’s referred to, and that’s where the P in ChatGPT comes from. The knowledge here is represented in that golden circle right there, and this is often evergreen content like US history or math concepts. From a business standpoint, this might represent what AI already knows about your company and your industry.

[00:10:52] If AI feels confident that it can answer your question with its pre-trained knowledge, then it will. But if it doesn’t feel confident in the answer it’s gonna give you, that’s when it starts to perform searches, and it will search the web. That’s referred to as retrieval knowledge, and this is information that’s rapidly updated or new, so it can’t be found in the training knowledge.

[00:11:14] Think sports scores, weather, news, stock prices. And from a business standpoint, it might be hours, reviews, new competitors, new products, social media activity, and so on. Even content that might ultimately be considered evergreen content, if it happened after the last time that the information in the pre-trained knowledge was updated, then it’s gonna have to search the web to get that information When your prospect goes looking for answers from AI, this is the process.

[00:11:48] First, the AI tries to understand the intent behind the question the person’s asking, and then, like I was saying, it tries to determine, do I have enough information in my pre-training knowledge to answer the question? If not, it starts to generate some searches. This is where it starts to formulate not one search query, but multiple search queries called a fan-out, which we’ll talk about– a query fan-out, which we’ll talk about in just a minute.

[00:12:14] Then it goes, it does the searches, it brings back that information, it pulls out the the facts, the information that’s in there. If there’s discrepancies, it tries to figure out what the right information is. It really wants to see consistent information in multiple– from multiple sources so that it feels confident in giving you the answer, and then it synthesizes all that information from those multiple sources and gives you the answer you’ve been waiting for.

[00:12:42] Now, the query fan-out is something that Google pretty much invented, but is really how all these AI platforms work. They turn one question into many questions and then synthesize all that back into a single answer. So it takes the initial question that you asked, how do I get rid of my crabgrass, say, and then it breaks it down into the underlying sub-intents, all the different things it believes that you want the answer to when you ask that question.

[00:13:11] It runs each one separately, and you can see here in the model subquery one, two, three, four, but it could be twenty-five, it could be fifty questions. It just depends on how complex the answer is. And then it goes ahead and synthesizes that information and provides the final re- result to you. This is a popular example.

[00:13:32] Someone searches or prompts an AI for best sneakers for walking. That will get broken into parallel sub-queries like best sneakers for men, best sneakers for walking in different season, sneakers for walking on a trail, best slip-on sneakers, and a host of other q- related queries. Now, if you ask that question, you never see all of the individual searches that were done.

[00:13:55] You just get that final result that tells you, “Here’s the top,” or, “Here’s the top five shoes that might be a good fit for what you’re looking for.” So we really need to stop thinking about AI as a search engine and really start thinking about it as a recommendation e- engine that uses search engines to do its home- to do its homework.

[00:14:16] So SEO still matters. It will probably always matter because it gets you into that research. But getting rese- into the research isn’t the same as getting recommended. Andy Crestodina, who is a friend of mine, brilliant mind when it comes to GEO, he was on my podcast and he said, “People are spending so much time chasing AI visibility, but what you really want as a business owner is AI recommendations.”

[00:14:39] So the question becomes, why would AI recommend you or why would it recommend the competition? And to get a better sense of where you are now, there’s three questions that you should be asking of AI or through AI perhaps. The first is, what does AI say about me? The second is, who is AI likely re- to recommend in my niche or industry or geography?

[00:15:06] And the third question I wanna come back to because it’s a little bit different, but I also wanna state at this point that is not Comic Sans font that I’m using in this presentation. I work for a design firm or in a de- I would never use Comic Sans. This is Chalkdust, which seems very appropriate for this image, but I just didn’t want you all thinking that I would stoop so low to use Comic Sans All right.

[00:15:31] So the first thing we want to find out is what does AI know about you and your business, if anything? So a simple prompt, “What do you know about my business?” And you just put in your business instead right there, and then see what the answer is. And I would recommend that you do this across multiple platforms.

[00:15:50] If you have paid versions of it, great, but you could also use the free versions, ChatGPT, Kloud, Gemini, Perplexity. And I might recommend that you start a document so you can start documenting what the responses are, especially so you can do that over time. So you wanna s- once you get these results, look through them.

[00:16:09] Do you agree? Is the information correct? Is it up to date? I was on a call with a client earlier, and they noticed that they’re being listed at an address that they haven’t been at for 10 years. That’s problematic. If you’ve got AI checking a series of different sources and the sources can’t agree on where your business is, AI’s not gonna feel confident in giving you as a result, and so you’re less likely to show up.

[00:16:33] Now, for flyte, we’re coming up on our 30th year in business, and for as long as most of us can remember, we’ve been offering web design and digital marketing. But in the last year or two, we’ve kinda upped our game and started doing more in the branding arena, even though our current website, which hopefully will be replaced in the next month, doesn’t talk much about branding at all.

[00:16:53] But for the past year or two, ever since I knew about how these AI search engines worked and what they were looking for, whenever I had the opportunity to describe flyte somewhere online, I would say that we’re a digital agency offering branding, web design, and digital marketing. And for several months, nothing changed.

[00:17:12] But then, when I was preparing for an earlier version of this presentation, finally in Perplexity It says that we help businesses and nonprofits generate leads and sales through branding and then a series of other things. So this was the first time that I saw AI recom- recognizing us for offering branding work.

[00:17:32] So that showed that a lot of the off-site work I was doing was paying off even though our current website has very little to say about any of our branding work. So do that. Take that first step. Find out what AI knows about you and your business. Document it. That’s phase one. The next thing you wanna do is find out who is AI recommending in your niche.

[00:17:55] And I recommend that you include relevant parameters like location, audience, and so on, whatever feels relevant for your business. And that is a very it depends category. Different business… or different industries might ask very different questions, and there might be very m- much different criteria.

[00:18:11] Some of us may be able to take on clients from around the world, where other ones are only interested in very localized business. And again, do this across multiple platforms and document what it says. Benchmark it, benchmark this so you can later measure your results. So you might ask a question like, “Who are the top five forensic accounting firms in Greater Boston?”

[00:18:33] If you were a forensic accounting firm in Greater Boston. Or maybe you might ask, “What are the best low-cost time tracking software tools for freelancers?” If that was your product and your audience. Or, and we’ll delve into this one, “What are the top five retirement communities in Southern Maine?” So I went ahead and I performed this, and I went to, I believe this was ch- yeah, ChatGPT.

[00:19:00] Now earlier I mentioned that AI search is probabilistic, not deterministic, and if you don’t know what that means, you’re about to see it in action. So I asked ChatGPT that, or I prompted ChatGPT with that, and basically ChatGPT came back with this list. Stroudwater Lodge, Piper Shores, Huntington Common, The Cedars, and Thornton Oaks.

[00:19:23] Okay, great. I guess we have our answers. Then two minutes later, I opened up a fresh window in ChatGPT, and I asked the exact same prompt, copy and paste right in there, and it was Piper Shores, Ocean View, Huntington Common, The Cedars, Atlantic Heights. Now in case you didn’t memorize those two lists, don’t worry about it.

[00:19:43] I got them here side by side, and we can see here was the first five and here were the second five, that Piper Shores moved up a spot during those two minutes. I don’t know what Piper Shores did in two minutes, but it seems to have worked. Huntington Common also stayed there. The Cedars stayed in its position, but Stroudwa- Water Lodge, what happened there?

[00:20:02] What happened to Thornton Oaks? And why did Ocean View suddenly make the number two position when it didn’t even make the top five two minutes earlier from the exact same platform, and same is true with Atlantic Heights? I’m internally curious about stuff like that, and really if you wanna get to heart of improving your GEO, you need to be curious, too.

[00:20:26] One of the things at the bottom of this list, ChatGPT, as it often does asked a follow-up question. And you’ll notice here that there’s a sources link. So I clicked on the sources link because I’m always curious to know where AI pulled its answers from. So I click on this information and I see…

[00:20:43] And what was interesting to me in this particular time is ChatGPT actually brought up two parts of its memory about me to answer my query. One was that I asked the same exact question two minutes ago. Okay, interesting, but not helpful. And the second one is that it remembered that I lived in Southern Maine.

[00:21:00] So that may have also skewed some of the results as well, but then I can also see… Oop, I skipped through it. Actually, let’s go back. Let’s just take a look at this. Also see that the, it pulled 30 different sources from the web, including Piper Shore’s website including Ocean View’s website. So those, it’s looking at websites of the people it might list, which means that you do have some say on your own website about whether you’re gonna get listed in this short list.

[00:21:28] And getting listed in a short list is incredibly powerful. I didn’t have this stat in the slides, but it was something like 95%, and this was specifically to B2B, but I’m sure it’s, it works other places, too, was that if you made the short list You were, companies that made the shortlist were 95% likely to get the business, or one of those companies in the shortlist.

[00:21:49] Only one out of 20 times did somebody start with a shortlist and then actually choose a company that wasn’t on that list. So getting into that shortlist is as important as getting into the top three positions in local search. And we also see down at the very bottom there’s a, a link to Seniorly, which is like a Angie’s List or Yelp for senior living.

[00:22:09] So we also need to pay attention to those kind of things as well. So this sent me down the rabbit hole and I asked the question, how did you make your decision on what to include and in what order? And ChatGPT got back to me and said, “Good question. I should’ve been more explicit that this was a curated shortlist, not a formal ranking based on published methodology.”

[00:22:30] And then basically it showed me how it made it. Weighted continuum of care heavily. Now, if you’re not currently looking at senior living, continuum of care is the idea that you can start in one place fully independent, but maybe over time as, oh, your condition deteriorates, that they o- have other services.

[00:22:50] Maybe they could move you into memory care or maybe they could move you into assisted living or something like that. And so that was a big… I didn’t tell it that continuum of care was important to me, but it knows enough about how people search for these and make these decisions that it weighed continuum of care near the top.

[00:23:09] So if there was a very highly ranked senior living facility near me that didn’t have continuum of care as part of their offerings, that might’ve knocked them off the list right there. Balanced reputation and longevity and nonprofit s- status. I would’ve never thought of nonprofit status, but these also tell me how AI’s making its decisions.

[00:23:29] It considered the location and access, and it did not simply rank by Google reviews. And finally, it mixed different retirement models to create that curated list. And then it told me that if it if it was doing more rigorous ranking it would use a scoring model like this that I thought was interesting.

[00:23:45] Back to those sources, I mentioned before Seniorly, so that’s one place where it’s pulling from If I were a skilled nursing facility or other retirement community, senior community in this area, I would be looking at how do I get Seniorly to be talking about me? You can also see that it pulled from Reddit and Reddit is often cited as a place where a lot of AI answers come from.

[00:24:11] Now, I think any discussion forum is a place where it might come from, but Reddit is a popular one. Whether or not Reddit should be part of your marketing strategy is a very case-by-case basis and I don’t find it super valuable for smaller businesses where most of the people on Reddit aren’t talking about you, but conversation for another time.

[00:24:30] It also looked at A Place for Mom, again, another Yelp-like service specifically for senior living. And then it also looked at Down East magazine that had an article or had a community directory all about Maine retirement. So these were some of the sources that it looked at. So your work, if you were running one of these places or marketing for them, would go to visit those sources.

[00:24:53] Make sure that you’re listed. This happened to me where one of the places didn’t list Flyte New Media as a top five agency in Maine. I said, “Why?” They said no, you tur- so certainly could have made the list, we just chose these other five.” And I asked, “Where did you go to, to source this?” And it mentioned a company called Tech Behemoths that I wasn’t familiar with.

[00:25:13] So what did I do? I immediately went to Tech Behemoths and set us up with an account over there, so next time AI went looking, we’d be visible there as well. So make sure you’re listed and make sure you have consistent, correct NAP, which is name, address, phone number. So make sure your name, address, and phone number is consistent across all these different platforms.

[00:25:35] And then also consider advertising, especially if you’re just getting onto this platform for the first place. It gives you additional benefits, additional visibility, and that may be just the kind of thing that AI at this point in time is looking for. So now we know what AI says about us. We know who AI is likely to recommend in our industry, in our vertical.

[00:25:56] The next thing is, and I abbreviated this incorrectly, what would my buyer persona ask of AI? What is my ideal customer asking of AI? What does that conversation look like? What are the questions they’re asking? Because I wanna be able to answer all of those questions, both at my website and, if I have the opportunity, other places as well.

[00:26:17] Now, if you’ve never used a buyer persona, and I’ll be honest, I never really got into buyer personas for a long time, but AI makes it so easy. This is a great use case for AI, is creating a buyer persona. This is one that I developed for a Maine bankers presentation I did earlier this year, so let’s just take a look at this.

[00:26:38] And I basically just prompted it with some information that I thought would be for a fairly typical user of a Maine bank. Create a buyer persona for a second-gen owner of a Mid-Coast Maine seafood processing company doing 4 to 6 million in avenue- revenue, and then I asked it to include some things.

[00:26:54] I probably didn’t even need to, because AI knows what a buyer persona looks like and would’ve completed it anyways. A few moments later, I get the buyer persona. It gives me the demographics of this person, it gives me the business profile, it gives me the financial pain points like seasonal cash flow volatility, it gives me growth goals over the next two to three years.

[00:27:17] Where does this Susan Willette, as they named her, where is she looking to take her company? What is her current banking relationship? What does she want from a bank? How does she make her decisions? Where does she get her information from to make these decisions? And you can see it’s broken down by industry and trade, business and local, and so on.

[00:27:42] And then also how to best reach her. This is a great buyer persona right out of the gate, only took a minute or so to create. The downside of this, or the shortcoming of this, is that it’s based on really good research that could be for anybody anywhere in the United States. It’s really not that targeted.

[00:28:05] You just pull from different resources. It didn’t necessarily pull from, the bank of flyte New Media, for example or whatever the bank is. It’s just using generic information. So we can make this better by including first-party data. We can make Susan Willett come to life, or Sarah Willett come to life.

[00:28:24] So first, we can add our own first-party data. We can record interviews with our best clients, asking them relevant questions. “What were you struggling with before you came to us? How– why did you choose us? How have things gone? Would you recommend this to a friend? What are the outcomes today?” All those sorts of things.

[00:28:40] Ask the same questions, do it via Zoom if you can with something like Fathom recording in the background. You get five transcripts. You can have AI summarize them, then upload that document to the AI chat where you first created your Sarah Willett and say, “This is first-party data,” and you can overwrite…

[00:28:57] use this data to, to flesh out this buyer persona. And now you’ve got a buyer persona that’s much more actionable and much more accurate to your own business. It still may not be correct. You still may read through your Sarah Willett buyer persona and feel that some things are off, and you can just tell AI that.

[00:29:15] “Hey, she needs to be able to do this,” or, “She needs to be able to do that,” or, “She doesn’t know this information.” Based on your own experience with your own clients, iterate until you feel like your Sarah Willett is really who a good representation of your buyer persona. And then when you’re happy with it, you can just ask AI, “Would you download this as a buyer persona doc?”

[00:29:36] And now you can use that buyer persona doc in future prompts, or if you’re building out an AI assistant or using scripts in AI, you can add it to that as well, so the AI will always know who your buyer persona or buyer personas are, and it will craft messaging that’ll be more appropriate for your business All right, so we talked about the fan out method that AI uses.

[00:30:00] It’s meant to replicate the buyer’s journey, and whether the buyer asks these questions themselves or the AI does it on th- their behalf, we want to know what those questions and problems likely are so that we can create content that gets us cited and recommended. So you might say something like, “Based on the attached buyer persona, please develop a list of questions or prompts that this person might ask of AI to find a company like mine.”

[00:30:25] And then you just attach your buyer persona doc right here. This is called RAG or retrieval-augmented generation, and what this does is it makes it less likely that the AI’s gonna hallucinate. So if you feed the AI more information in that session, like by attaching your buyer persona, you’re gonna get better, more accurate results.

[00:30:46] There’s less chance that the AI’s going to to hallucinate information that’s just not there. We have this information. I did the same thing with Sarah, and some problem searches, how do seasonal food businesses cover cashflow gaps between paying harvesters and getting paid by distributors?

[00:31:03] Which main banks actually understand seasonal businesses, not just tourism? SBA 504 versus conventional commercial mortgage for a facilities expansion. How to prepare financials for a growth stage loan conversation. Best banks for b- … So on and so forth. So these are legitimate questions. Wait, I got one more.

[00:31:22] Let’s see. And sh- also, it’s telling me that she wants quick answers because she’s typing on her phone or during her commute. This is all great information, and this is helpful, and it’s going to help dictate what type of content that I should be creating for my website, for my newsletters, for social media, for my YouTube channel, what have you.

[00:31:39] All of which is gonna help me capture and convert Sarah. But here’s the caveat. Sarah’s not a real person. She was created by A- AI. And these are not real questions. They were developed b- by AI. But even though these are fake questions from a fake person, they’re still valuable. A buyer persona, even before AI, was never a real buyer, but it’s still valuable.

[00:32:04] A map of Portland, Maine isn’t Portland, Maine, but it’s still helpful if you’re trying to find the local coffee shop. All right. So let’s take a look at some of the other actions that you can take to improve your own GEO. Now, I’m gonna categorize these into three different areas, and they’re the same areas that I talk about when I talk about SEO.

[00:32:22] Things you can do on-site at your own website, things that you can do off-site, and some of the technical things that you should keep in mind as well

[00:32:36] So here are some things that you can do for your own website based on the research paper that first coined the phrase Generative Engine Optimization. Now, keep in mind this research is a few years old now, and things have obviously evolved, but it is foundational to everything that’s going on today The first one is stats and quantitative data.

[00:33:00] This one by itself gave a 40% lift to the chances of being cited by AI platform. So you wanna put stats and quantitative data on your website, and I just saw another report come out from Andy Crestodina, name-dropped him twice now, that they did research into this, and they found that the number one thing that you could do is have factual statements on your homepage.

[00:33:27] Right now, the AI seems to pay more attention to your homepage than any other page on your website, and it has the biggest impact there. But of course, you should be putting facts all over your website. Citations are also important, so quotes expert quotes, fluency, how well it reads. Using simple, straightforward language is valuable, and the one negative that the report cited was keyword stuffing as a negative, and this was true with search engines, so it’s the same thing with AI.

[00:33:55] You can’t just re-re- repeat the word dog park 100 times and think that AI is gonna say, tell everybody that you have the best dog park around. Also how you structure your content using clear H1, H2, and H3 hierarchy, which is HTML speak for headers and subheaders. Using the inverted pyramid approach, which is how news articles are written, where the most important facts are right at the top of the page.

[00:34:22] In fact, one of the biggest, I find struggles I have with writing copy that’s optimized for AI is that AI wants the answer right at the top. And as somebody who loves creative writing sometimes it’s not always easy. I don’t wanna give away the ending right at the top of the page. But for AI, having a header or a subhead and then having the answer that directly refers to that can be beneficial in terms of helping you get found and cited by AI search And then QA, Q&A formatting.

[00:34:55] So the, if you do a search on, “How do I optimize my site for AI?” All the information or all the advice out there is very eerily similar. One thing it tells you is that you want to do FAQs. And the thing with FAQs is it’s not that FAQs are so powerful, it’s just that it’s a great format for asking a question and giving an answer.

[00:35:20] And if that question matches one of your buyer persona questions through the previous exercise, and then you immediately answer it, that’s something that’s really helpful to the AI who’s trying to find the right answer for that person. So something to keep in mind, but it doesn’t mean that you have to have an FAQ at the bottom of every page.

[00:35:39] That, that’s not necessarily any better. It’s like in the old days when people would say blogging is good for SEO. Yes, but it’s not that a blog is good for SEO, it’s just at the time, search engines were looking for a lot of new, fresh content, and blogs were the easiest place to find it, so that’s why it worked so well.

[00:35:58] Now, a regular page on your website is just, if not more beneficial than a blog post would be. People just got in the habits of saying that blogs are good for SEO There’s a lot you can do offsite as well, and I mentioned the fact that I took every opportunity I could to talk about our branding services whenever I had an opportunity to describe what flyte New Media does.

[00:36:21] Because AI’s not gonna take your word that you’re the best, fastest, cheapest, offering the best customer service, or most complete bumper-to-bumper warranty. They’re just not gonna necessarily trust you out of the gate. They, or AI’s gonna look for, look outside of your website for confirmation and context.

[00:36:39] It’s gonna look at major newspapers. It’s gonna look at online guides, at blogs, at Google business profiles, and a whole lot more. So of course you can influence what AI uncovers using many of the same tactics from traditional PR and SEO. So one of the big things is that it’s gonna look at reviews. So making sure that you get as many positive reviews on platforms like Google or Facebook or Angie’s List or Yelp or whatever’s specific to your industry.

[00:37:11] Like I was saying before Seniorly and A Place for Mom would be two platforms that are appropriate for senior living. So do your best to generate, to do great work and to generate positive reviews for your business, ’cause that will make a big impact. Also, directory listings and citations. So if there is…

[00:37:29] down East magazine had that list of senior, a senior guide, and so you wanna make sure that your senior living center, if that’s what you do, is in there

[00:37:43] Excuse me, earned media. This is traditional PR, right? Or digital PR these days. It’s just about getting, whether it’s a national magazine or a local newspaper or an industry mag, to be writing about your business. And for AI, it doesn’t even need to create a link back to your website, it just needs to keep on seeing your name come up over and over again.

[00:38:04] One of the shortcuts and ways that work today but may not work in the future is there’s a lot of PR distribution services out there, and there are ones that are really good for big stories, and there are ones that are very cheap and you can do a lot of and will distribute to a lot of places that nobody really cares about, except for AI.

[00:38:23] So you can find some cheap services that may be under $100 a press release and send out some press releases, and this is just gonna basically seed the internet with a bunch of articles linking back to you but also talking about your business and what you do. So when we launch our new website, I’ll be sure to mention that we offer branding services as part of this new website that we have.

[00:38:46] So those are some things that you can be doing. Owned media, we talked about your website already, and this is offsite stuff. But think about different platforms that you might control or a blog that maybe isn’t on your website, anything like that. And then social media. In the past decade or so, I’ve definitely moved away from pushing clients to do organic social media.

[00:39:08] I just don’t think that there’s a lot of visibility, value. It doesn’t drive a lot of traffic. I think it’s decent for B2C brands. I think there’s no reason for B2B not to be on LinkedIn, but I don’t think of it as this direct generation machine for new business. But AI is an argument that we should be getting back to some of the basics, even with social media, because AI looks at our social media profiles where it can.

[00:39:31] It looks at our posts where it can. And that also gives it a sense of who we are, what we do, and who we serve. And the last group of things is the technical side of things. I’m not sure if I mentioned this today. I apologize. Been doing a lot of talking about AI recently. But I was talking to a business that had their website built by an old employee, and it turned out the way that he built it, AI couldn’t parse the page.

[00:39:58] So AI couldn’t tell you anything about what was going on in that business, and before we could do any GEO work, they needed to completely overhaul their navigation, otherwise AI would not be able to see their website. That’s an obvious problem there. But so you wanna check or, and even limit the JavaScript that your webpage is using Another thing is a lot of companies were trying to protect the integrity of their Google Analytics perhaps, and they shut off access to AI bots, to bots, to their website.

[00:40:29] And on one hand it makes a lot of sense. Unfortunately, now the AI doesn’t know anything about your business. So we wanna make sure that we’re allowing AI bots to our website. Schema is something that I have mixed feelings about. If you don’t know what schema is, it’s a special type of markup language that’s used in search.

[00:40:50] Whenever you go and you search for a recipe, and you’ll see the recipe is there, and it’s got a photo. In the Google search results it’s got a photo, it’s got number of stars, all this sort of stuff. That’s because somebody on that page set up the schema for recipe sites, and there’s hundreds if not thousands of different schemas.

[00:41:07] There’s one for local businesses, for travel, for events a whole bunch of ones. You can find them online. I bring this up because one of the other things that’s in every list for what to do for GEO is add schema to your page. But the truth is AI can probably parse your page anyways. It’s a good idea to do schema for a number of different reasons.

[00:41:29] It probably, I would call it table stakes. If you on- if you do nothing else and you only do schema, I would not expect to see big changes because AI can already read and understand your page. But it is best practice, so if you’re looking to check every possible thing off your list, sure, go ahead and make sure there’s schema on your pages, at least the important ones Now, if you make these changes, you may wanna find out hey, did this actually move the needle or anything?

[00:41:56] And as I mentioned, measurement can be very difficult when it comes to this AI work. And a lot of it might be more manual. In fact, I just got back from a conference in Boston, the B2B Forum, and one of the recurring themes because of AI is it’s much harder to prove that marketing is working in the AI age because all the KPIs that we’ve become addicted to over the last 20 to 30 years are starting to become irrelevant and not telling the whole story, and getting real data is becoming trickier and trickier.

[00:42:33] But some of the things you can do is do that spreadsheet, and, or I said spreadsheet. I think more, more or less it might be better just to have this as a Word document or a Google Doc. But document what do you know about my company queries, and may, and maybe you wanna do that mul- multiple times. And then also the question about, oh, so I’ve got them in the next slide. Sorry. So ask the question, who am I? And document that across multiple platforms. Who do you recommend? And put that in this document as well. Create the buyer persona, develop about five to 10 buyer prompts that you want to track, three to five platforms, as I mentioned, and then maybe a monthly or a quarterly cadence and run these things multiple…

[00:43:16] each quarter or each month if you really want to, and just see, are the changes I’m making getting me into this shortlist of recommendations from AI? You can also check your GA4 reports. Th- this is fairly new category in GA4 reports. It was only introduced this year. But you’ll see AI assistant as a traffic source, and you’ll probably discover that it’s a much smaller amount of traffic than most of your other channels.

[00:43:47] But again, it is growing, and as we’ve talked about, it’s not the only metric that’s important because AI conversations are having a much bigger impact that you’ll never be able to measure. You’ll never be able to attribute it to somebody’s ChatGPT conversation or chat that they had. I do know that Google says they’re rolling out their own AI tracking tools within Google Search Console as well, but I haven’t seen it in our Search Console yet or any of our clients’ Search Console.

[00:44:16] And in a recent interview, Google admitted the tool wasn’t really ready for primetime, so I’m not sure when we’ll see a tool like that. We are seeing a lot of AI visibility tracking tools in a lot of the SEO tools that we’ve seen in the past, like Ahrefs or Moz and some of those come with caveats as well.

[00:44:34] I don’t think you should really trust any one number So GA4 is not the end-all and be-all of determining the question if is your GEO working because GE- Google Analytics, GA4, only measures traffic to your website, doesn’t ma- measure citations, doesn’t measure recommendations. And if somebody has a conversation with ChatGPT and then does a branded search in Google, there’s no way of connecting those dots to find out that it was actually ChatGPT that was the original source.

[00:45:05] And as we’ve discussed, AI traffic is very likely underreported, and the more I learn, the more sure I am that it has a much bigger impact than shows up in your GA4 reports. Because AI influence does not equal AI traffic

[00:45:22] Now, as we wrap up today, and like I said I wanna take some questions after this. I don’t see any questions, so I must have answered every question along the way. But if you have some questions, now is the time to start dropping them into that chat. If you’re gonna walk away and think about… Yes?

[00:45:42] It looks like there is one question in the Q&A section. Oh, okay. How do I get to… let me finish up this, and then we’ll hit the Q&A. Thank you. So benchmark where you are. That’s the first step you wanna take. Create content to answer your buyer persona’s questions, and I showed you how to do that.

[00:45:59] And then optimize your website for those kind of answers. Optimize your offsite. If you have the opportunity to do podcasts, to create a YouTube channel, like these are opportunities for you to get your name, your business name, who you serve, how long, all those details out into places where AI can find them.

[00:46:19] And then also just make sure that you’re not preventing the AI bots from visiting your website and learning more about you. And then measure in GA4 and manually. That’s what I have for you today. I do have one question I saw come in the chat, and then I’ll switch it over to Q&A All right. Kate asks, “Do you still need strong SEO to be found by AI?”

[00:46:44] I would say yes. The bottom line is there is so much overlap between SEO and GEO that it would be foolish to just put all of your eggs in one basket, and you should continue to optimize. And a lot of the things that you were doing that were best practice for SEO are going to be best practice for GEO, and are also best practices for PR.

[00:47:07] So a lot of these things, these traditional marketing things that kind of went out the window when we got all these great platforms and measurements are all becoming very important again. And then also, “Would AI see recommendations on one’s LinkedIn personal profile?” That’s a good question because I don’t know that AI can search Link- I know AI can search LinkedIn because I know that AI, when they did this recent ranking of sites that AI is most likely to pull from, LinkedIn was near the top.

[00:47:39] What I don’t know is whether it can get into personal profiles or if it’s just going through the newsfeed. So that is a question that I would have to get back to you on. And I am going to be sending out an email in a couple of days with links to the video once it’s been processed, and if I can get an answer to the LinkedIn question, I absolutely will.

[00:47:59] All right. Let me just make sure I’m not missing any here. Wait, bring up that takeaway screen back. Okay, I can do that. Oh, hold on One, two, three, four. Okay. There, and I’m gonna be sending slides too, so if you miss anything, don’t worry. Lori, “I work with AI often, and so when I asked in your homework exercise, it answered our company a top of recommendations each time for the b- various product lines we offer.

[00:48:26] Is AI just trying to give me the answer I want to hear, or is it accurate?” That’s a very good question. And one of the events, one of the sessions I went to at the B2B forum was talking about the difficulty in reporting GEO because he would, in a big conference room, he would ask 10 people to put in the exact same prompt about a business question, and they got 10 different responses.

[00:48:50] He’s like, “How am I supposed to report on that?” So what you might do in that is you might ask somebody else or use somebody’s account to do it or do a free account and ask those same questions as well, so that way it doesn’t have a history or a memory of previous conversations. And so it won’t likely, it won’t be as likely to recommend you, and that would be one sure way of finding out for sure.

[00:49:15] And the other thing would be is, again, look into the sources. How did you make your decision? What sources did you pull from? And make sure that you’ve got a strong visibility in each one of those platforms that it went to make its decision. Yeah. Carrie says, “Do your search incognito.” So I think there is an option in a lot of these tools that you can do a an incognito prompt or that isn’t connected to your memory, and those are all beneficial ways as well.

[00:49:43] I can’t promise you that it would be the same as going to a, a chat where you’re not logged in or anything like that, but it might be helpful. A lot of thank yous. You’re all welcome. What percentage of AI users are paid subscribers, and is there any data on how different the responses are? April, that’s an excellent question.

[00:50:01] I don’t have an answer to that question. I’ll see if I can get it for you. And also if you’re a paid user, that may mean that you’re using it a lot more often as well. I will say for anybody out there who’s using AI, you should be using the paid versions, period. The bottom line is if a product is free, then you are the product.

[00:50:22] So on all these platforms, if you’re using the free version of it, it’s training off your data. So anything you share with it is gonna go into the knowledge base. So any, sensitive data that you share about your numbers or your business goals, that’s all going into the cloud and may end up in s- a competitor’s query down the line.

[00:50:41] So you definitely wanna make sure you have… you’re using the paid versions. Or even better, you use a local AI, which is free. You can just download them. I’m not gonna go into the hows. You can Google that. But that would be another way. But again, most people aren’t using those platforms. We wanna be using the platforms that our ideal customer is going to be using.

[00:51:00] So I’ll see if I can get some data on the percentage of AI users that are paid. Those are probably, Power users, but it might be a good idea to do a search logged in and not logged in so that you can… If you’ve got some users who may be more consumer-based and just using AI for one of the first times, those would be important questions to ask and ascertain as well.

[00:51:24] Everybody is very welcome. And somebody says, “Personas too, that will help zero in the answers.” What paid platforms are you comfortable recommending? Because of my job, I have a number of paid subscriptions for AI. Claude is my go-to, but I also use ChatGPT and Gemini. I use ChatGPT and Gemini primarily for image generation when I need to do images from scratch or creating a chart or something like that.

[00:51:55] Those are some of my go-tos in terms of large language models. Although Google’s also great for deep research, but then after I run a deep research on Google, I always say, “Okay, now reword that so a layman can understand it,” ’cause it tends to speak in highfalutin words whenever it does a deep research project for me.

[00:52:12] Those are the LLMs that I recommend. If I could only have one, it would probably be Claude, and then I also have some special use case AI platforms as well. Descript for all my audio and video production work I use Descript. Just absolutely brilliant tool. I use Fathom for note-taking on my Zoom calls.

[00:52:30] I absolutely love that one as well, and I do the paid version of that. There are some other tools here and there that are escaping me, but those are the ones that I go to most often. You’re welcome. Oh, hold on.

[00:52:46] And a bunch of these just disappeared. Maybe Lauren’s helping me out behind the scenes. Lauren, did you also mention that there was a question in the questions? Yep. I didn’t remove anything, but- Okay … there are three open questions in the Q&A section. Okay. Would you read those to me? The first one- One at a time

[00:53:02] and let’s see. Karen asked, “Let me know if this is outside the scope, but how does it come up with false answers and links that do not exist? Does it use its own logic to come up with answers, which can be very weird?” Yes. So I’d have to see them, but sometimes it works off of resources that are out of date, and sometimes LLMs work, like I said, in a probabilistic manner.

[00:53:27] So if it guesses at the most likely next word every single time, which is why I got two different results when I did the same prompt on the same platform two minutes apart. If I said in front of a group of 1,000 people, “Peanut butter and…” and everybody answered, most people might say jelly, but some people might say chocolate, and occasionally somebody might say fluff.

[00:53:49] So that’s a good example of how these AI, these LLMs work, is they’re just guessing at the next word, which is why you get hallucinations. One way to cut down on hallucinations, like I briefly mentioned before, is to provide it your own documents, your buyer persona, your your brand guide, your brand voice doc.

[00:54:08] All of those will help lock it into provide less hallucinations, and it will make less stuff up because it’s working off of very specific documents that you’ve either created or approved. All right, next question. Okay, one more here from Zach. “When I asked to tell me about the business, about my business on ChatGPT, which I use, it pulls in information from my work.

[00:54:31] How do I get it to tell me what it knows about my business from a public-facing side?” I would try logging out and maybe trying the free version of ChatGPT. If you’re not allowed to do that at work, ’cause there’s maybe some security issues, that might be something that you just wanna do at home.

[00:54:47] You can set up a free account with a different email address, and then perform all your searches like that. Probably be my recommendation.

[00:54:57] All right. Was that it, Lauren? That was it. All right. I want to thank everybody for the time you gave me. Great questions, by the way. I will be sending something out to you in the next couple of days with a link to the video and everything like that. Share it with your friends. If you need any help on any aspects of this, I love talking about AI.

[00:55:16] We’d love to help you out here at flyte. If you need branding… See what I did there? Branding, web design, or digital marketing, we’ve been doing this for 29 years. I expect to be doing it for at least another 29 years, so please feel free to reach out to me either by following up on one of those emails I sent, or the one that I’m about to, or just give me a call or find me on LinkedIn.

[00:55:36] I’m The Rich Brooks on LinkedIn and everywhere else. Other than that, have a great rest of your day, and I will talk to you soon. Thanks so much.