How to Establish an AI Search Optimization (GEO, AEO) Strategy and Goals

How to establish an AI search optimization strategy and goals in an effective way. For today's episode, I have asked a few very experienced SEO and AI search specialists to share their insights, their tips, their methodologies that they take with their clients to address new AI search optimization processes. Tom Capper from STAT+MOSS, Judith Lewis, SEO consultant, Jonathan Moore as your consultant from different angles, different industries, different types of clients as well. And before starting, I want to thank our today's sponsor, which is Similarweb. Similarweb has also a wealth of AI insights and data that you can leverage as well. Thank you very much for your sponsorship. The first thing that I would like to address, because I know that today There's still so much misinformation and buzz, especially in social media, right? And I would love to hear which are those top errors, mistakes, and myths regarding setting AI search optimization strategy and goals in particular.

I've been working in digital marketing for over 25 years now. So, uh, kind of old, kind of seen a lot in my time, and I think some of the biggest errors are looking at trying to port the old metrics, so to speak, of keywords and measuring keywords and things like that into the new environment of AI or LLM search, LLM usage. And it doesn't quite map. I've, for many years, asserted, shall we say, vociferously, that people should not be using just keyword rankings as a measure of success. We need to be looking at the money we're making for a company, whether through generating leads or whatever. We can't be looking at how many keywords we're ranking for at what level. It's a proxy for making money. So now suddenly we don't have keywords. And I'm literally not even figuratively, I'm actually in the middle of creating my report for a client, uh, just pre-end of month because he's very interested in what we're doing in LLM Search. What am I looking at? Money. How much money did we make? How much money is coming in? I don't care what fanout queries we're in. I don't care. Really, that is money. Money and what AI knows about you. Accuracy within YK. So I want to make sure that if the LLMs are answering something, it's actually what's on the site, not something it hallucinated. So accuracy in the answers and money. Are we making money? And the biggest mistake is not Relating your actions and goals to the money that you're making the company. That's the biggest mistake.

I think the bigger issue is that we see this unfortunately, right? This extreme takes, it's like nothing has changed versus everything is changing. Right. And in nothing has changed is pretty much reusing in the exact same way, pretty much like keyword research from research and trying to just make longer keywords for prompts and use them in the exact same way when it is like, uh, not necessarily reasonable to do so because of their dynamic nature and how FireEye works and so on, on one hand. And then on the other hand, regarding that from a strategy standpoint and from a success measurement and goal setting. Unfortunately, the overreliance that we had had with traffic because it was a very easy to understand KPI, which unfortunately right now doesn't necessarily explain the full story. What I explain to my clients is that this is only the base, the floor. It is not the ceiling. And there's so much more that cannot be explained just by looking at the traffic that you can see in your analytics, right? Like the influence from a branding standpoint in AI visibility on one hand, and then on the other hand, like even the behavior that we see users. So indeed, we cannot just extrapolate them. And then on the other hand, the KPIs that they, that we use and the type of goals that we set in order to make them reasonable to communicate.

This is really close to my, to my heart. So for those that don't know me, a lot of my job is working on the sort of metrics and insights that go into the tools and data that we produce at STAT and at Moz. And so this, this is the exact kind of thing I spend a lot of my time Thinking about. I think the challenge here for the industry is that this requires a really different mindset to the one that we're, we're used to. I absolutely agree with, with Judith that it's, it's about the connection to the client's bottom line, but the, the, what that connection looks like is very different in this ecosystem. We all cared about rankings in the past, and instantly I'm not suggesting that we put that in, in, in the bin or something. But we all cared about rankings in the past because we felt that we could create this sort of story or a plausible correlation where it's like, okay, you rank for something, you're quite likely to receive a click. If you receive a click, you're quite likely to receive a conversion. You've sort of got this funnel going down and you can paint quite a clear— often there's sort of a trail in the data from the ranking right through to a conversion. Right? And that's not really how it works in this ecosystem. It's more like the kind of mindset you might apply if you were buying print advertisement or billboard advertising or this kind of thing. The biggest mistake I see people making is they obsess with the links in the AI results. Now, don't get me wrong, those are important. They are tactical information that you might use to optimize your chance of being mentioned or your chance of being mentioned positively or accurately, as, as Judith says. But the, the link is not the KPI in the way it used to be. You are not really expecting someone to click through to that link and convert. I'm not saying it never happens, but that's a fraction, a tiny fraction of the value that this system is delivering to you. What you care about is being mentioned and recommended and talked about positively. In the answer, with or without a link. If the link is to someone else, all the better. That's kind of more credible, and that means that you can go and influence that publication via your digital PR team and try and be mentioned even more positively or in even more places in, in the future. Um, you're not going to say, oh damn it, The Guardian ranked in this citation, therefore I failed. You're going to say, great, I was mentioned positively in The Guardian, and as a result I got this sort of free mention in this conversation and people are gonna think about me and talk about me. But the attribution chain from that mention to your conversion, because it's no longer via the click or anything like that, is so much more nebulous, right? So we're looking at this more like you would've looked at traditional marketing where you often can't draw a— we've been spoiled in digital marketing, right? Over the last 20 years.

Mm-hmm.

We didn't feel like it at the time, but we were. We were spoiled and we, cuz we were able to draw those really concrete attribution chains. And that's at least for generative search, that's gone. That's not how this works. It's, it's a different kind of marketing channel. It's not a performance channel anymore. And the reality is that this has probably been true to some degree for regular search for quite a while. And it's only sort of recent developments like AI overviews and ChatGPT and so on and so forth that sort of forced us. To have a reckoning with this reality. It's probably been true for a while that a lot of the value being delivered by search was actually brand awareness with or without a click. And it's definitely been true for a while that your rankings were not telling you the whole story, right? That, you know, for example, we have a metric at Stats called pixel depth where we look at how large and how high your result is rather than just And you'd be surprised how poorly that correlates with rank, right? And the whole reason we cared about rank was because we thought that being further up the page meant you'd be more likely to be clicked on. Well, what if you are ranking first but you're not even visible? That's literally the majority of the time now. The, the number one traditional organic result is not even visible. So rank was already not telling the whole story.

We pretty much were still using rankings as a useful KPI to measure supposedly like good visibility, but not anymore. And we can see this a lot in e-commerce. It wasn't necessarily like related to AI overviews, the data coupling, but it was actually more related to product packs, for example, or images.

Like product carousel, right?

100%. 100%. But yes, this is a problem that we have been like dragging for a long time. You mentioned something interesting too, as well, regarding not being only a performance metric. reinforcement channel, I believe that there's a mix now of performance plus branding and the value of being just mentioned and recommended in an accurate way, in a way that represents well and correctly your brand values, your excellent proposition, and so on. That doesn't necessarily result in a click or traffic, but just the endorsement will have value, right? And that, I believe, that is the tricky thing that potentially in other areas of marketing are much more used to communicate or to live with that uncertainty.

Right.

Since we have been always working with more visibility, more clicks, more traffic, more conversions and revenue, supposedly, right? And that to be consistent or directionally helpful. But we see that that is not necessarily the case.

So for me, the main starting point is good strategy starts with a diagnosis. And I would describe someone that's a good strategist is strategists diagnose causes, diagnose causes, right? Rather than Leading with the solution. As a freelancer, I tend to get, I'm gonna use the word like clients here. So normally people come to me with like 3 things, right? They have a vision, like we wanna be market leader. They might say, come to me with a symptom. That happens a lot, like traffic's down, that's a symptom. Or a proposed solution, which is probably not always the best one of these 3 is we wanna do AI, we wanna do a load of stuff. So at that point, our job isn't to basically validate that solution. What we should be doing is basically diagnosing like the real business problem that we're trying to solve, which goes back to what Judith said at the start, right? It's normally tied to economic activity or an economic measure. So for me, by having that clear diagnosis, the strategy then is once we've got the business problem outlined, the strategy is quite easy because it's just how are we going to achieve that, basically that problem? Of like, we've got the problem defined, we've got the strategy, like how are we going to achieve it? And then the goal then is really simple. It's what does success look like? And then how that sort of ladders down then into basically the tactics, which is the activity that we're doing. Like, the next thing I'd say where I see people maybe going astray is getting measures, which is like the effectiveness of a particular activity, getting that confused with goals. So if you understand the difference between goals and measures, you're going to be in a good place. And if you also have a good root cause of this is the business problem we're trying to solve, that's going to really help you. In the context of AI, then that actually really translates nicely to these are things like, how does AI understand us as a, you know, as a brand? And then why, why don't people have the confidence in us? And then that gets into the specific tactics. So. I tend to start really at the start with that diagnostics and getting the definition on that problem because it makes then everything else so much easier. Bad signs are just where we go straight into, like, and I've seen this, is like, yeah, we're just gonna do 200 new articles to get complete coverage on a topic, right? Or we're gonna do a load of digital PR. There's certainly situations where you may consider doing that, but it's certainly not aligned to the business problem. And basically where you want to get to.

Probably one of the biggest sources of issues that we have had historically as SEOs in the community, in the industry, being too tactical, being one-trick pony oriented historically, right? At the end of the day, that has ended up backlashing so much when these shifts happen. I would love to move on with what would you say that are the minimum viable steps to take to avoid these myths? And those steps that you also tend to take to communicate and establish good KPIs to track, to take into account in AI search optimization processes, and then the goals.

Tying it back to business goals. It doesn't matter whether we're SEOs or PPC or social media or GEO. It doesn't matter what we are. If we're tying it back to business goals, then it's easy to bring the business along with us. But I think what you said. And we've all, in fact, reflected on this a bit, is strategy versus tactics. So, you know, as, as Jonathan was saying, there are tactical elements that we can employ, but we've got to be thinking strategically. So I think one of the problems of us as an industry is that we're not thinking strategically. And weirdly, I see this in search awards where it says, list your strategy, and they just list tactics because they have no conceptual notion of the difference between tactics and strategy. And I think— Perhaps that's a seniority issue. It's a leadership issue, but we need to be thinking strategically. So we always have that North Star that we use as our guide to every other thing that we do. So once we know where we're going, that's why I mentioned YK. I really like it because of that branding element. We've all talked about that branding element. I know whether I'm accurate in Claude versus ChatGPT with YK. And that's why I appreciate that as a tool that helps me deliver that visibility for the client and say, look, you know, you've got to get over clicks from organic search versus AI tools versus whatever. Let's make sure that you're accurate and then we make the money. This is how it relates back to money because our funnels are weak when we're talking about AI. We don't know what our click-through rate versus conversion rate is. So we're still working on that. So we've got to bring people along with us and Obviously, I'm not saying that if you're a junior, so to speak, a junior or an executive or whatever in your organization, you have to suddenly skill up and understand what business goals are and how to target and align KPIs to them.

I believe that the key here, and this is something that I am trying to do also in my day-to-day as a consultant, right, is to tie what you mentioned well, Jonathan. It's like presence, readiness, or optimization, and business impact. So these are all tied together. And what you actually end up optimizing for is what is going to move the needle for those actual topics that you want to be endorsed or get endorsements because they have an actual value to your brand, how your brand is perceived, the accuracy, the visibility, the sentiment, business proposition, and that also connects well with the business rather than trying like what we used to do also, unfortunately, for way too long in SEO, trying to target keywords just because of the sake of the search popularity and so on, right?

So the way that we've structured this in our tools is we're, we're trying to focus people on presence and on sentiment drivers rather than on, for example, whether their URL is there or the order in which they appear or these kinds of like more traditional search analogs, which don't apply to the same degree. And the industry hasn't really settled on one metric at this point in the way that maybe we did before with rank. Although rank has. So much more nuanced than we ever like to talk about itself. But we're trying to push people towards those more nuanced metrics and metrics that are towards the more subjective aspects of a response and what's driving those and what sources might be informing those and this kind of thing. Use your prompt tracking solution to track nuanced prompts. I'm going to get a list of citations that are talking about these issues that are interesting to me, and then The list of pages that's there is gonna be the starting point of my ideation for my PR campaign rather than going the other way around. I feel like normally at the moment it's the other way around where it's like, okay, what's the story I can produce? Who is that gonna influence? Oh, look, there's some overlap between this and citations almost by accident. That's okay. But I would love to see it be completely from the top down. Like, okay, digital PR has never had the opportunity to do this before. And have that sort of technical SEO mindset of being, we're gonna go from the goal right through the, to the solution. And it's, it's the opposite of the story I was just telling about attribution earlier, right? For attribution, we've lost this ability to connect the story. For digital PR, they've gained it where they didn't have it before. They can now finally say, look, we did this thing, we changed the narrative. Here's the steps we followed. This is the result. This was us.

I love that you touched this because I believe that this has everything to do with alignment, right? In bigger organizations where there are actually PR departments and community management departments also that work a lot with influencers, actual channels, is that they usually don't talk about this with their in-house SEO or AI search specialists. And now it is not that it wasn't needed before, but now it becomes a must to have alignment and understanding of What they are doing, which resources, tools. We already share a few, by the way. YK is an amazing tool. Of course, Tom is working at Status Moss. We'd love to hear also about the tools that you are using.

We talked about YK there, but I think that's really such an important tool. And I think a lot of people get the understanding wrong of like why, where it adds value, but it's there. It's that understanding your reputation and what evidence there is out there. And it then breaks it down into entities. So that's like super powerful. I think prompt tracking can help again, because again, it helps you understand like essentially the features, benefits of your product. And if there's alignment there with a target audience, there's someone called April Dunford and she's done a, like a book called Obviously Awesome. And that's talking about positioning. So I like to sit down with my clients and do like positioning exercises with them to actually make sure that they're communicating like how they talk about their brand consistently and effectively. And then I also like to do USP exercises again, because there's often like misalignment there. So you have to start with like branding, USPs, and how you position yourself before then you can actually then start like, is the evidence consistent with how we want to talk about ourselves? So those are sort of the starting points I look at there. And Waikī's put a big part of understanding that in terms of prompt tracking. I use, you know, the usual. There's, I think there's over like 90 prompt trackers in that marketplace. What I would do is encourage anyone, I did a talk in Athens on like just doing some manual, manual checks initially, but you can, you can, like there is access to things, APIs. So like Data for SEO is like a tool I use heavily that I have built my own Mickey Mouse prompt tracker using that, which allows me to do a lot of this verification work very very quickly using like their stack. And that's, that's pretty cost effective if you don't necessarily have the budget initially to go out and take the plunge on some of these more expensive tools. But really for me, before buying a tool, it's actually really thinking about, yeah, how, how can we verify the evidence? And then from there, that sort of should support then that diagnostics that you have or help create that diagnostics.

To start wrapping up, I would love to hear your last tip or your last resource leverage. My actual final tip will be to more than obsessing, because I believe that a lot of people is really obsessed regarding 100% accurate prompt tracking, which by the way is like impossible at this point because of we understand the personalization and dynamic nature of prompts. So rather than, oh, I need to track a million prompts in order to better understand. No, what is important is that you have a representative AI search prompt library that actually have enough meaningful topics across your different customer journey stages and personas and product lines and service lines and so on, and take it from there and focus on those actually meaningful metrics that will really explain if you are achieving or not the goals that you want, whether from a performance standpoint, branding standpoint, and so on, right?

Definitely. I'd want people to take a look at SE Ranking. As well as search, and of course stat and Moz. Moz especially has been in the market forever as far as looking at tracking things because we can't be fully blind. Obviously, we want to have some metrics that we can feed back to the client. But my biggest tip would be, as I've said from the beginning, understand the client's goals. And it's not just monetary, obviously. I'm very fixated on money because I like to make sure that my clients exist for a long time. So they pay me for a long time, but it's understanding the client's goals. So I had a client once, his goal for him was about what he needed to reach for his bonus. And it happened to be weirdly subscriptions to the newsletter. So sometimes we have weird KPIs we've gotta reach, but understand what your client needs to make them look good and make your client look good for their boss. So really understanding the whole business's goals is amazing, but also understand what the person who hired you needs to do in order for them to look good. So definitely focus on that. And I've already said YK, SE Ranking and Semrush are 2 others. And of course Moz, which is the OG of all tools.

My tips would be if you're considering a prompt tracker of any kind, um, yeah, include some branded prompts in the set, not overwhelmingly branded, but include some branded prompts in the set because you do want to understand what is said about your brand. Specifically, make sure the prompts that are being tracked or the responses that you're getting back are grounded. Like in a lot of tools, you'll get back only the sort of in-model responses, and that's not really representative of what users see. And it also doesn't give you citation data, which is the tactical information you need. And then on the sort of accuracy and keeping it representative point, it's kind of the same pitch in some ways as a, as a tool provider. In some ways we find that easier than dealing with Accuracy and regular search tracking, because to be honest, Google does a lot to fight us on sort of deliberately mucking around with with regular search rankings to make them harder to track, and that isn't happening with AI tools yet. But keeping them representative is a challenge, and I I would say that the the cost and the scale of your prompt tracking like. A lot of these tools are very expensive right now, mentioning no names, but this should probably be about the scale and about the cost that you're used to with regular rank tracking. And that then gives you the room to have a representative sample that will allow you to overcome some of those localization or personalization issues, just like it did in regular search. And lastly, I would say, yeah, the focus on the— look at the metrics. And I kind of said this already in the episode, look at the metrics that you're actually getting out of of prompt tracking. Like, this is not an alternative or an add-on to, to put in your rank tracking report. You probably get it from the same tool because a lot of the steps that we're going through to get the data, uh, exist in a similar way. But the way that you're thinking about this within your strategy is not the same, or it shouldn't be the same. Um, you are not looking, you're not ranking in citations. Please, like, just remove that one phrase. We're already doing so much better.

100%. I believe that the naming conventions should be updated to avoid misunderstanding and the type of focus that we should have when looking at those metrics indeed.

Anyone listen to our points on strategy that maybe is like, oh my God, what do I do? Because it is hard to like retrofit a strategy once maybe you've already gone down a certain path. So think things I would like, just simple yes, no answers, right? That you may be doing some of these things. It's a red flag and that means it's not good for you. It's not good for your company and probably not good for the client. So activities before diagnostics. So again, if you haven't got a business problem, let alone a diagnostics, and you're just going into doing activities, problem number 1. I think strategy sort of confused with tactics or tactics driving strategy the other way around. So again, leading with a solution, don't want to be doing that. Goals confusing measures. So that's like problem, I think one of the main ones there. We don't actually have like where we want to get to, what needs to change. And then probably no guiding principles. We didn't talk about that very much today, but a guiding principle is essentially the— basically how we will make decisions. Again, we talked about North Star, but same sort of point really, that guiding principle there is actually saying, yeah, that is valid. We're not sort of going off-piste, we're not going off on a tangent, we're all sort of moving in the same direction. So that sort of gives you that direction and getting everyone aligned and basically, I'm going to use a British phrase, singing to the same hymn sheet. So Those would be sort of like some of my checks there, just to sort of again make sure that you are on the right path strategically. Because I think historically, isn't it like, as an industry, we are sort of— this historical problem, it's been around for years— that do a lot of checklists, don't we? Or here's your SEO, like, your 12-month plan. So if you're doing a combination of those sort of things, it's probably, yeah, probably worth having a revisit and going from there.

Thank you, everybody. I really appreciated your insights, all the inputs, the mix of resources and dos and don'ts as well. Hopefully they will be useful to address the topic in a much more reasonable way, successful way, also connected way with the actual business, a strategical mindset as well, which is important. I really appreciate it. And hopefully you learned a lot from today's guests. Please feel free to Follow them. They are quite active in social media as well. And remember to subscribe to the channel to avoid missing out the following episode. Thank you very much, and until the next edition, bye-bye.