Subscribe
Share
Share
Embed
James Dooley, Jabez Reuben, Malte Landwehr and Kasra Dash discuss the best AEO and LLM visibility strategies digital marketing agencies should focus on in 2026 to build consensus, define their entity and win at both discovery and decision-making.
This video explains which digital marketing strategies digital marketing agencies should focus on in 2026 to improve LLM visibility, entity definition and conversion at the decision-making stage. James Dooley and Jabez Reuben, Malte Landwehr and Kasra Dash start with KPI tracking because prompt tracking and share of voice give agencies directional data to know what to improve rather than relying on vanity metrics. They cover brand SEO, AI visibility and Google Business Profiles because stronger search presence improves trust and conversion rates.
The discussion also explores organic SEO, organic social media and paid social ads because consistent visibility across search and social supports long term growth. PPC is analysed in detail because campaign setup, landing pages and lead handling directly affect results. They also discuss Reddit, Quora and paid AI ads because diversified enquiry sources and early adoption can strengthen digital marketing performance for digital marketing agencies.
PromoSEO lead generation for digital marketing agencies recently received recognition as the "Best Digital Marketing Agencies Lead Generation Agency."
Best AEO Strategies for Digital Marketing Agencies in 2026: James Dooley, Jabez Reuben, Malte Landwehr and Kasra Dash on LLM Visibility is available on:
AI SEO and Business Automation Podcast is founded by James Dooley because he teaches business owners how artificial intelligence accelerates growth. The show explains AI models, agents, and automation systems because practical guidance helps companies scale with less effort.
James Dooley: All right. So today we have a very special episode. We have Malte, CMO of Peec AI. We have Kasra Dash, founder of The Masterminders, mySEO App, and we have the newly crowned king of AEO, James Dooley, with us. Thank you so much for joining, guys. The most important question we ask every single guest, your top three ranking factors for LLM visibility. Malte, why don't you start?
Malte: Consistency, reputation, being recommended in all the sources. Right, Kasra?
Kasra Dash: Understanding your query fan-outs, making certain that you actually have a trustworthy brand that at least looks trustworthy when LLMs are doing research, and good foundations, so good citations that are citing you. Defining the entity. So, ideally getting a Knowledge Panel if you can, getting a KGMID, um, would be the foundations, third-party corroboration and consensus. Repeating, like what Malte said, is the, the confidence and clarity of who you are, what you do and why you're brilliant.
James Dooley: Talking about consensus, if consensus is the core for all major LLMs, then why don't we see consistent results across all LLMs?
Malte: I think that's a great question. I think the, the way that each LLM gets the information, whether it's from Bing, whether it's from Google, whether it's from Brave, is one reason why. And also what Kasra just spoke about there, understanding your query fan-outs. The query fan-outs can be different. So ChatGPT might be doing site colon on certain websites where Claude or Gemini and stuff might be using different trusted sources. So I think it makes a massive difference from each LLM because of that.
Kasra Dash: I'll try and give a different answer to what James said. I, I think the, the black box plays a big massive part in it as well. For example, how I like answers to how you like answers to how James likes answers to how Malte likes answers. It's very different. So, for example, my ChatGPT might know that I have got severe ADHD and I ask you 10 million questions in a day and I want short answers, right? And it's going to give me what I like. Whereas James might be a little bit methodical with his actual answers and he might want some actual scientific-based theory. So, I think that plays an important part that not many people speak about as well.
James Dooley: Yeah, and additionally the, the LLMs have different partnerships, right? So, for example, OpenAI has this partnership with Yelp. Google has access to Yelp via the regular Google index, but other LLMs might not. Google has unrestricted access to Google Maps. The other LLMs might have to do some scraping, some working around the limitations. So, everybody is trying to protect their data sources. Also, how good the access of each LLM to something like YouTube or Reddit is can vary. So that alone will have an impact on what kind of documents are actually considered for the grounding process.
Malte: A bit more about consensus. Do you think we can establish consensus right away or do we need consensus over a period of time to start ranking for decently competitive keywords or terms? I'm not asking about just made-up queries that definitely you can easily start ranking as soon as some content is out. But for queries that have decent competition, can you start ranking or showing, or can you start showing up in LLMs with sudden spike in publication and you build consensus right away?
James Dooley: I, I'll start by saying absolutely 100% you can build consensus and manipulate what the AI answers are within 15 minutes. And I can show you hundreds of different examples of this. Long as you get on the trusted sources and in the sources that's being re-crawled during that live RAG and live search. And it's... You got to remember the LLMs aren't as intelligent as what we think. A lot of the time they're running a live search. They're just getting the information of what's ranking within Google. So how fast can you get a source ranking? I mean, some of them are using different... We spoke about some might use Bing and stuff like that. If you can get rankings, how quickly can you get rankings on the first page of Google can be how quickly you can start being cited in the AI Overviews or in the LLM.
Kasra Dash: I'm going to slightly disagree with what James said because, like, let's say, for example, we set up a, a new CRM to compete against GoHighLevel and a couple others out there, right? If we do, like, a load of listicles, yes, we can probably spike and get into certain LLMs, but we will probably at some point fall off because what I've noticed is that there's other reference points that it's looking at. So, for example, it could be looking at case studies of people who have, who have used your actual CRM. It could be looking at what people, like what influencers are actually talking about your, your CRM. Like, is potentially Matt Diggity or Neil Patel using your CRM? Because if that's the case, then you're probably going to stick a lot longer in actual the, the, the LLM answers. So for certain industries, yes, you can spike it and you can get maybe 20, 30 listicles and that's probably enough. In other industries, you're going to need to figure out other parts of the puzzle as well.
Malte: Yeah, I, I think it's a bit more nuanced, but, but I actually agree with both James and Kasra. So if in theory you can make it happen that tomorrow you are a top 10 CRM on G2, you have a Wikipedia entry, you have case studies, everybody who is comparing CRM is suddenly including you, everywhere where HubSpot and Salesforce and Attio are mentioned, you are mentioned as well, yes, you can form consensus overnight. But it's not realistic. And for a very, very competitive topic, you will not be able to do it overnight. But of course you can pick an aspect like CRM for dentists, etc., etc., or CRM in Paraguay. Like, you, you can always get to consensus quickly in some niche. But for the super hyper-competitive things, to have it long-term and sustainable will, will take some time.
Jabez: That's an interesting take. What's a GEO tactic that a lot of experts have declared dead but you think it's still working well right now?
Malte: For me, advertorials. I hear it very, very often, they don't work anymore. I think about a year ago first people started telling me, ah, advertorials are not working anymore. To this day, I still see a lot of advertorials in the sources. The only requirement is they have their own URL. They're crawlable by, by bots and they are not served via an ad server, via JavaScript, but they look like a normal article. It can say very, very prominently in there that they are an advertorial. It can even be in the title that they are an advertorial. If they are well targeted on a very, very high-reputation domain, LLMs will continue to use them as sources and even cite. Is it same like publishing a press release that XYZ brand launches their service in this location or XYZ brand wins this award in certain category, similar to that?
Kasra Dash: Yeah, I mean, those also work. It obviously depends on the publication and how much you can push them in one direction, what they write about you. But it can also be, this is the car insurance company that everybody should switch to this winter, and then it's just you are the one that's being mentioned.
James Dooley: James? For me, social media. I've seen quite a few people saying that, um, Reddit now is not forming part of, um, the decisions. It is still there and is still using, and I've got data to show that Reddit is still being used and the information that's being placed on Reddit. It's just they're not using... They're not actually the front-end sourcing Reddit as anywhere near as much. So, Reddit's dropped a load, but they're still using that information. And there's one or two other social media platforms as well that they are using, but the front end they don't seem to be having it as being the link source, but they're using it in the background.
Kasra Dash: I'll give you a super basic one and, and Kasra from three years ago would have laughed at this answer. Would have said it's a waste of time. Things that are, like, on your website and it's, like, the easiest, the cheapest one, right? So, like, your About page. Has it got when you were founded, who the founder is, where the headquarters are, like your Meet the Team page? If you've got, like, let's say you're an accountancy firm and you've got seven chartered accountants that have all got 10 years plus experience. Have you mentioned stuff like that on there? Weirdly, I've noticed that ChatGPT actually goes to privacy policy pages as well, just to check if, like, there's a business number or company number on there. Like returns policies if you're, like, an ecom store. Again, these are all things that you can change literally after watching this video. Doesn't take any money. You don't need to pay anyone to do it. It's literally just five minutes' worth of work on each of those pages.
Malte: I 100% agree with Kasra. ChatGPT opens so many pages, especially if you just check with the fan-out and the reasoning section. We've seen this in local niches where ChatGPT is looking for who is the oldest and most established, and then it goes and checks the About page to see which one is the oldest, which one is the most established. Besides FAQs, reviews and all that, there's so much more you can do on your site to still boost your ranking without worrying about publishing hundreds of listicles.
Jabez: Getting to the next one on similar lines, what's, what's a GEO LLM visibility tactic which is definitely not working right now, which was probably working six months ago but is not working right now?
James Dooley: Good question. Uh, majority of things that's rank... Obviously, if something's not ranking, then it's not, it's not working. If it's not being picked up by the LLMs, it's not working. But if something's ranking, then it's forming part of the retrieval. So I'd say if something's not working for you in, for normal SEO in rankings, then therefore it won't be working for GEO as well. But yeah, I mean, generally speaking, if it's working in SEO and it's ranking and you're defining the entity and you're defining why you're great with reputation, then it's working in GEO as well.
Kasra Dash: I think I'm going to go back, refer back to what I was saying before, right? Like, literally just spamming listicles, you probably will get a spike and then you'll probably fall off. I think there's, there's more to it. And I think, like, the, the answers that we give today, like all of us, is probably going to change in, like, six months' time as well. You're probably going to need to do more and more. So I think just, just solely relying on listicles. Yeah.
Malte: Nothing to add from my side with, with regards to that. You said in six months. At present, probably every six days things are changing. It's changing so much every single day.
Jabez: Crazy. Malte, you shared an interesting study about recycled Reddit strategy some time ago on LinkedIn. Can you explain a bit more about, about that strategy and is that strategy still working?
Malte: Yeah, so first it's not working as good anymore, but I believe it still works. Back then I noticed a website called GummySearch showing up in our top sources and I looked at it and it was a actually defunct project. And their idea was to aggregate what people are saying on Reddit about brands, and then they closed the whole project because Reddit deactivated the API access, but they still had all of these landing pages, uh, open. What was happening back then? ChatGPT was adding the term Reddit to their fan-out queries without site:reddit.com. So they just added the term Reddit. And of course these landing pages from GummySearch, they were perfect. It was something like Levi's jeans and then it said in the title, opinions about Levi's Jeans on Reddit. It would list the exact subreddits, the top 10 subreddits. It would have exact quotes from Reddit with the username, the timestamp. It had the term Reddit over there everywhere. And it was structured in a way that if you took the Google Quality Rater Guidelines with their E-E-A-T focus and applied that to how a human or an LLM might judge something, would say, "Wow, this is an amazing source." They aggregate these real user opinions. They provide source for everything. There's a link to every comment. I can click on it. And of course, users tend to write in a different way from professional copywriters. So they often use unique terms, bring in unique perspectives. So that's why this domain for about one or two months was an absolute top source in, in ChatGPT. Now that ChatGPT has the term Reddit less often in their fan-outs or is using a site search, this works less well. I would still recommend if you are, for example, Salesforce, you could create a page Salesforce/Reddit and also call it honest opinions about Salesforce on Reddit and just cite a couple of users there that have said something about your brand on Reddit. It can make you become one of the sources during a grounding process when the LLM has the term Reddit in the fan-out queries, but it's not working as well anymore as it did a couple of months ago.
Jabez: Are you doing any tests on those lines, Kasra, James?
James Dooley: Yeah, so obviously the... This isn't new, by the way, with regards to appending Reddit to the end of it. It is for LLMs, but this has worked well for the last two years. So trying to do... Add Reddit into certain pages saying you might have got certain bits of information using Reddit, and then straight away anyone that has your main keyword. So ours might be, like, best lead generation companies Reddit. And I think just so any of the viewers that's watching this, of why people was doing that. The search within Reddit itself is garbage, right? It's just no good. And people was preferring to do the search in Google and look for best SEO experts Reddit. They wanted a Reddit thread, but they wanted Google to bring back what they thought was the best SEO experts thread and forum for them to be able to answer into it. So the amount of people for many years, and this is why I think Reddit exploded initially, and I think this is why Google did the relationship with Reddit all those years ago, because people just started searching in Google and appending Reddit to the end. So if you add... A nice little tip. If you add it in the footer of your page, if it was an affiliate site, um, we provide official case studies and we might get certain information from Reddit. Um, we've got thousands of five-star reviews on Trustpilot and Clutch, and you name a few things. They're site-wide n-grams that you're bringing up that might be being pulled in for every single page because it's, it's on that at the bottom of the page. And just by having that mention in the footer could make a difference between you showing up for the keyword plus Reddit and not showing up. So that has been a great SEO strategy, I'd say, for a few years, but it became even more important when LLMs just appended Reddit as well to the end. Right now they are doing more than Reddit now, especially in the 5.6 high mode. You will see CNET, TechRadar, NerdWallet, depending on your niche. Right now we are doing tests where we are adding, you know, those queries or those terms in titles just to see if, you know, it's influencing in any way.
Jabez: Moving on to the, to the next one. If there's a new brand that is nowhere in the LLM visibility landscape and it comes to you, how would you plan the first 90 days of the LLM visibility campaign?
Malte: I can answer because I just, uh, went through this because my wife started a new company recently. Describe yourself on your website very, very clearly and agree on one or two industry terms that you use to describe yourself, and then create profiles everywhere. LinkedIn, Yelp, Google Maps, whatever fits your business. And describe yourself the same way everywhere because the first step is that if you ask an LLM, tell me about brand or what is brand, that it gives a clear, definite answer and that it doesn't say, uh, I don't know, maybe it's this brand or maybe that brand or maybe it's a typo. Like, that's the very first thing, that it understands what you are as an entity. And I will leave it up to the others as to what the next steps are because this shouldn't fill your first 90 days, but it's, it's what I would do on day one.
Kasra Dash: I'll add some stuff to what Malte said as well. So I would basically focus on the next three pages of Google. So if somebody's searching for your brand, what's... All right, let's say the first four results or the first five results. Result number one's your brand name and then the next four results maybe, like, your social media profiles. So that takes us to, what, six? Position six, position seven to Google, and position seven to 10, and then positions 11 to 20 and positions 20 to 30 on Google. What I'm then worried about is guest posts, but not in the traditional sense of trying to get the guest post to rank for a keyword that you're trying to rank for. I'm doing branded guest posts. So on every single one of those links that I'm acquiring, it's literally just defining the actual entity. And I'm, I'm just making a super strong branded window that all just reflects back to who I am, what I do and who I actually serve.
James Dooley: Yeah. On that, a similar framework to what Kasra said there. Um, who you are, what you do and why you're brilliant. So Malte there defined who you are. So having a clear and concise manner and they've got confidence and understanding who you are as an entity is step number one. You need those foundations. What you do are all the different services, the products and stuff of what you try to do. So you're appending the keyword with your brand. So you're connecting that semantically. And why you're brilliant then would be, have you won any awards? Do you have any reviews? If you've not got any reviews or testimonials, you need to go out to your clients to get some reviews and testimonials. It might be that you do, if you're in the SEO industry, you might do a free site audit for someone in, in exchange for a testimonial. So then you start building that up. And over time, you've got to keep defining who you are, what you do and why you're brilliant. The first steps, like you said, is what Malte said, define the entity, then what you do, and then the reputation arm comes in time. But on the reputation arm, you can keep redefining who you are on the reputation part and saying what it is that you did, which is then defining all three. Very important to, for people, in my opinion, to have specific frameworks, um, that you're working towards and ticking the boxes off. Otherwise, the staff are just all over the place and they don't really know what the next steps are.
Jabez: How do you decide which subcategories to target first? For example, if there's an invoicing software and then under that there can be so many subcategories that it can target, maybe one category that you will go after. What's your strategy?
Kasra Dash: You could either go with the easiest one that you're going to easily win on or you can go with the most profitable one for the business. I, I think that kind of comes down to the actual business owner.
Malte: Yeah, very similar answer for me. I think this is not just an SEO answer. You really need to think as a business, what do you want to own? Can you own? And then go for that. Would most of the time advocate for the one that is the easiest and least, uh, contested, um, if they are all somewhat equal from the economic point of view. Of course, if you will... If it's not going to be profitable, ju... Just strike that, uh, subcategory. But in general, it's always easier to win the bigger categories if you're already established in the smaller ones. But I would also look at, are there maybe categories that don't mix? So if I position myself as the CRM for freelancers, it's not a natural progression that in the second month I go after medium-sized and then in third, third month after enterprise, right? It's more like a long-term decision. Whereas if I go after a certain industry, it's more easy, natural to then expand to more industries. I'd also look at it from, from that perspective.
James Dooley: Yeah, similar things to what Malte said there. I'd create the category and then I'd then get all the, and the topics or the industries that we want to work with. So let's say it was CRM for accountancy practices. I would just focus on the easiest ones to go after to start with. But then also, if you're just focusing on accountancy practices, then what you're trying to do is the ones that you're working with that you have, you're getting the reviews for an accountancy practice saying how amazing your CRM is and stuff like that, and you're building that part up. You're staying in your lane. You're staying consistent. And then once you've kind of built up a little bit, it becomes easy then for CRM for builders and then CRM for HVAC companies and CRM for plumbers. And you can just keep building it up. But go with the ones that you think you can win on to start with. I think there's a few different ways of explaining it. Some people say work through your traffic tiers. Some people, Chris Carter used to call it the SEO Avalanche approach. David Quaid talks about cornerstoning. So it's trying to get... Trying to win for something, trying to get those clicks for something, and then the next tier up becomes easier and you can internal link from the pages that have got traffic to that and then work your way up to the top. It's hard to go straight top of the level. It might be the most profitable, but it's hard to start there until you've started to win the secondary and supplementary kind of services.
Jabez: Has your keyword research strategy changed for LLM visibility or GEO campaigns?
Malte: So, mine has changed massively. Depends a little bit. If it's an established business that already has good SEO performance, of course, I would continue all the traditional keyword research. But if it's a very new business in a very competitive space, I would actually often now recommend them, like, forget about SEO for the first year. You're not going to win anyways. But with AI search, you can win. And then it's just a different strategy and different, um, things you should look at because suddenly fan-out queries are, like, your... One of your main inspirations for, for content creation and your main keyword source basically. Whereas for traditional SEO, it's fan-out queries are, like, one additional input but not the main one. So for me, it's changed very massively.
Kasra Dash: I think the, the biggest difference previously, like let's say four years ago, is obviously everyone would do keyword research and it was all going on their own website. Now I'm experiencing two different buckets where it's, like, you look at your query fan-outs and then you're like, oh yeah, I can upload these 15 query fan-outs to my own website. But there might be a query fan-out, it's like review of James Dooley, right? And it's like, you can't really review yourself and upload it to, to James's own. I mean, may... Maybe you could, right? But I think certain query fan-outs, it's a little bit cringe to, to upload to your own website. And I also think on a, on the flip side of it is if it's like a review of Kasra Dash and it's on Kasra Dash's website, you're probably going to get less clout from certain LLM. So that... How I'm bucketing the, the keyword research is keywords on my website, keywords on third-party sources. So now I'm basically needing to do two sets of keyword research.
James Dooley: Yeah, I completely agree. You put the words out my mouth. It's like when you said that then, it was like writing, um, a Valentine's card to yourself. To James, you look sexy, love from James. But yeah, I completely agree that, Kasra. So my actual keyword research for my website hasn't changed that much, but where Malte is saying things have changed a lot is I now do two topical maps or two semantic content networks. I've got an on-page one, which remains the same. I want to cover every single entity, every single attribute of those entities and the topics and the services of what I do. That's actually not changed that much. But the off-page signals, the query fan-out terms, third-party corroboration, that building of that consensus. I never even used to do it. And I'm looking back now and I think I used to... I always say I'm a bad SEO, but I used to think I was actually okay. But now when I look back and I go, cringe, how bad was I as an SEO? I never used to care about my content and my guest posts. I used to just think, oh, it's semi-relevant and I'm getting the dofollow backlink through to my page. I'm getting the PageRank distribution and that's the only thing I cared about. I didn't care about what the other page said. Now I'm, I'm obsessed just as much as, like, I am on, on my own site. I'm obsessed with the semantic triples and defining the entity and who you are, what you do and why you're brilliant on those third-party sources. And that, Malte, is what I'm saying. I agree with you on it's changed a lot, but for my off-page because I never used to do it. So now it's like, wow, every query fan-out term, not only am I wanting to get rankings for it, I'm now almost, like, treating it like reputation management. I want to own that first page for every query fan-out term as best I can. I think GEO has just made, or AEO, whatever you want to call it, I think that's just made SEOs a lot better marketers.
Malte: True.
Jabez: What do you think about prompt tracking or prompt volume? Let's say, does Peec provide prompt volume?
Malte: We provide prompt volume in a limited manner. I see the topic quite critical. So real prompt volume doesn't exist, right? Like an exact prompt often has a volume of one because it's very unique. And also we need to understand that if a human writes a prompt, that is not what the foundation model answers, right? It answers this prompt plus whatever user context says, which can be location, time, but can also be chat history, the history of that specific chat and also whatever the, the model has learned about a user. So I see very little value in trying to even have that on a prompt level as an absolute number. However, of course, people have an interest of, okay, here's the topic of content management system. How many prompts are... Should be branded or unbranded? How many prompts in reality are informational versus transactional? Which is why we try to take a prompt apart into subtopics and then see what is the demand for each subtopic. And there traditional Google search volume plays a big role as a proxy metric. But then also look at how connected are these subtopics actually? Because you could, of course, have a prompt like money, Wikipedia, Trump, USA, travel, flights, insurance. Like all of these are terms with, like, very high demand, but of course their combination makes no sense. So then using real user prompt data to, like, combine these topics. And in theory, in the end we have an absolute number, but we do not trust this number enough because we believe there are too many biases in the data that we just scale it down to a one-to-four scale that is relative to each project, depending on what prompt you have, to give people some guidance. I know there are some tools in the market that give you an absolute number. And funnily, some of them now stop doing it, and others are now very openly saying, yeah, it's just Google search volume multiplied with the market share of ChatGPT, which is probably not the same as prompt volume in ChatGPT.
Jabez: Then how do you plan which prompts to track? Because that's what most of the tools, the core service is prompt tracking, visibility tracking. How should people plan which prompts they should track? And the second part to the question is, with slight variation, with but same context, do you see drastic changes in the results? Say, just before you start on that, because I just want to revert back to the prompt volume a minute. I want to have a bit of a rant, if I'm being honest with you, because it pisses me off with certain people in the SEO community. And what really annoys me in the SEO community is when people slander a tool, right, for having something like prompt volume, going, "No, you can't have prompt volume." I get it, you can't. But also they're all manipulated too. That's just a guesstimate, and the search volume in all these tools, like within Semrush or Ahrefs or any other tool of what you do, they're all guesstimates, right? So all these people that are coming back slandering the AI visibility tools for having prompt volume, every single metric, Trust Flow, manipulated tool, toxicity score, you've got to take all of them with a pinch of salt. But as a data scientist, you get all the bits of information and you base your decisions on that. And these people that are on Twitter and on LinkedIn slandering these things, I'm just like, you just don't get it. You don't understand how, how you analyse the data and what you can turn around and do. And you got to look at it almost from a branding point of view, the share of voice, not just in an isolation going, "Oh, they're just manipulating numbers." All numbers are manipulated. Unless you're in Google Search Console and you've got... And then that's not even showing all the data. So, you've got to get every little bit of data and then being able to use it. And it frustrates me when people are sat there like keyboard warriors saying, "Oh, this is a pile of shit." Well, what, what, what metrics are you using? And how are you deciding whether you're going up or going down on the AI visibility? So, if you're not tracking it, you should be tracking it and at least be moving towards the correct manner from a share of voice point of view. So, I just want to get that point across, that all these people that are out there slandering things, but then they're using DR and using search volume on just Google thing. Well, they're well terrible. They were absolutely shocking, right? But I don't slander because I still use them tools as well as part of all my other data of what I have. And it comes in the, in the use of how you use the information and how you use the data of what it should be, not just whether it's good or whether it's bad. I'd sooner have a metric than not have a metric. And I can base my decision then of whether I want to use it. I can hide it if I want to. I'd sooner have a metric there that people are trying to predict what that could be, as opposed to there isn't a number for it. I remember the, the, like years and years and years ago I done a talk at BrightonSEO and I done the exact same talk at SEO Estonia. And people were saying that the talk is flawed because Semrush does... They done the, the toxic links, but it was a completely flawed metric anyway. Because if you had a link, like, for example, in, in iGaming, you need to link to Internet Gambling Association or something like that. And if, if there was a link in your footer that was linking to a website, it would automatically deem that backlink as being toxic. And I'm like, guys, come on. Like, if that's the case, then this Internet Gambling Association, if they had to do a disavow, according to what Semrush has given the, the stats at, they would need to disavow their entire backlink profile. Like, this is super flawed. And then it also said, like, certain directory websites were flawed, which you just knew they weren't. So, yeah, I think that, that that's super frustrating with SEOs when they just believe a certain metric and it's like, guys, come on. DR, I can manipulate that in, in an afternoon to DR 76 website.
Malte: Yeah. And even, like, Google search volume, right? Google is now grouping the terms SEO and SEO tool. But probably these are different intents. Or something simple like, what is a user and what is a bot? Like whoever had to do bot detection for a large website knows how impossible that is to answer because the data centres used for bots are the same used by the VPN providers. And if you're a very large website, you will have some percentage of your users being legitimate VPN users. And so just giving an absolute number, like, how many visitors did we have yesterday, is technically impossible to answer. It's only ever possible with one definition of what is a user, what is a session. So yeah, for a lot of data, we should just look for directionally correct stuff that we believe in and make actions based on that. Now coming back to the original question. Sorry for the rant. All good. Great rant. No, very informational. That's, that's a very good point because we should take all of these tools as just a direction, but with a pinch of salt and that, instead of fully relying or fully cancelling them out.
James Dooley: Yeah. And maybe should make tools accountable that if it's just directional data, to say it, and if it's absolutely, um, accurate measured, to also say that.
Kasra Dash: You asked about prompt differences. There was this experiment from Rand Fishkin where he asked his followers, please write a prompt, how would you would look for a noise cancellation headphone? 144 people participated. 144 different prompts were written by these people. The good news is that we ran these prompts many, many times because Rand was nice enough to share them with us. And for 90% of them, the same brands are being mentioned in the answer if you run them over and over again. Of course, if you run each of these prompts once, you will think every answer is different. But you could run the same prompt three times and get three different answers, right? So if you run them over and over again, 90% actually bring the same brands. And this also makes sense because LLMs are so good at understanding the topic, the intent and the limitations you have. I mean, you can literally start writing in English, then in the middle of the sentence switch to German and then put a sentence in Klingon at the end and the LLMs can understand what you want. And, and if your last sentence is answer in Spanish, they will, they will give you a good answer in Spanish. So it really, really works that they understand what you want. It doesn't matter how you express it, how many typos you make. You can even write and change the order of letters in every word and usually ChatGPT will understand. You can even put your hands on your keyboard, move your hands one letter to the right and start typing. And if you write a longer text, an LLM will understand that it needs to shift every letter and give you an answer. So you do not need word for word the exact prompts to give roughly the same information. Now in terms of how many prompts to track, I would always start with what are my topics that I want to report on in the end. So I want to filter and drill down. Do I have different intents? So commercial, transactional, informational. Do I maybe have different personas? And do I have sales funnel stages if I have a complicated product like in B2B SaaS? And do I have other qualifiers like sustainable and not sustainable, cheap, certain materials, etc., etc.? And then I need to think about which of these combinations do I want reporting on? Like, do I need reliable weekly reporting? What is my SEO visibility for green iPhone 16 phone covers? If yes, I will need a couple of prompts for that. But if I only need reliable data for what is my visibility for phone covers and for iPhone phone covers and for green phone covers, I can do that with a lot less prompts. And then it's actually a simple question of maths. How many data points do I need per week to have reliable data? You can think that every chat is a coin flip. Either my brand is mentioned or not. If it's a fair coin, so if I have in reality 50% visibility, probably with 10 flips of the coin, I'm somewhat in the right direction. Right? It's very unlikely that if I have a fair coin, I get 10 times one side. It is possible, but it's highly unlikely. Now if I have 90% visibility, in reality just five chats might be enough to, with reasonable error margin, give me something that is correct. If I have 2% visibility, even 100 chats are not going to give me an accurate data because maybe I'm in one chat, maybe I'm in three chats. It's already mixing. Maybe even in 10. The lower my visibility is, the more data points I actually need for accurate data based on what my requirements there are. You can mathematically come to a set of chats that you need. And then for a typical brand it's something like, for each of these combinations you want to track, have at least five prompts in there for that topic, with that intent, for that sales funnel stage, and run them once per day and then look at weekly aggregates. If you need daily reliable data, you need to track a lot of prompts because there's always some fluctuation in the data, right? It's not like Google Search that is repeatable. The LLMs have always some level of change in there. So that's how I think about the topic.
James Dooley: With regards to just, just expanding on that, just internally the way we do it, and Kasra might... Kasra might talk about a different way because I'm a business owner, so everything that we're doing is for our own assets that we own. And if I was an agency, I'd probably do it in a different way to this. So a caveat, Kasra might want to talk about how an agency owner might want to do it. But for me as a business owner, I want to track as many different queries as possible. The more data I have, the better. The minute I've got more than 50% visibility for a term means, means that I've kind of, okay, I've covered that term. Now I take it out and I'm, I'm consistently looking at stuff of what I can improve on. I'm not bothered about AI visibility, what percentages I have and stuff. Like, I'm more worried about what can I improve on? What am I missing out on? What, I'm only getting 2% of the share of this. How do I improve that to get it to 50%? Once I've got it to 50%, I take it out. I'm putting new ones in. I've only got 10%. I'm improving. I want to know everything of what I can improve on to strive to get the business to the next level. I'm not doing it for glamorous reports and shiny reports at the end of the month to send through to my client because it's for my own sites. I just want the team to be able to know what areas do we not do well in that we can improve upon, and they improve upon that part, and then we're consistently deleting the ones out that we've done well on and consistently striving to try and improve where we're at.
Kasra Dash: Yeah. Then I think it's very situational. What I quite like about James's approach to it is if you're trying to get, like, a keyword or a phrase, let's say, from 60% to 70%, the effort that's involved in that, you could have probably picked four more other phrases and got them from zero to 50. So, you're getting more coverage across a broader selected set of words and phrases as opposed to just homing in on getting one or two phrases from 73% to 81%. And also we found from a lot of testing, once you got a load more terms, secondary and supplementary terms, from zero to 50%, it actually bolstered the ones that was at 60% to 70%. It was like, stop chasing this one prompt. You've got it to 60%. Leave it. Move on these others that you're not doing well with. It's almost how you building... How you build good quality topical authority. Look at the ones you're not doing well on. That starts moving up. These move up another rung. These start moving up another rung. Moving up without actually trying to touch that. So, I think it's important, but I think... But I do get there certain people that don't want to run that methodology because they're wanting to show nice glossy reports that, look at this, you've got 63% on average for the same keywords that they keep tracking every single month. And it's like, where's your growth? And they just track the same ones. And it's like, come on, guys. There's so many long-tail keywords. Like you just said there, Malte. Rand's test showed that they did so many different search queries. There'll be some... I know there's a 90% overlap. I'm more bothered about, okay, what's the 10% differences? That's the, that's where I'm looking at 10% differences. And then that then opens up more queries that start showing up for that you might be showing in Google Search Console. You can consistently keep looking for new queries and adding it in, but then I want to be taking the ones out that I'm doing well on. I think probably the biggest issue that a lot of agencies have is how they're selling the AI/LLM or whatever you want to call it, GEO services. Um, if, if you've... If your sales guys went to a company and said, "Yeah, we can get you, I don't know, 90% visibility for whatever e-commerce store you're in, right?" They probably need to go down the route of, "Okay, yeah, we've got it from 5% to 10% this month, then next month we've got it from 10% to 25%, 20%, 25% to 30%, so on and so forth." Whereas if you actually sit down with a business owner, you're like, "Look, we're going to just do a holistic approach and we're going to go after everything that's related in your actual, in your actual industry, and we're going to go after that as opposed to a singular set of, let's say, two or three keywords." That's probably just a way better and also future-proof strategy as well.
Jabez: You, on the research part, something that we have not already discussed. What else is part of your research before starting a GEO campaign or LLM visibility campaign, whether for your own brand or for your clients?
James Dooley: Happy to go last. Okay, I'll go first then. I mean, for me, I do, like, I'm doing fishing. I get the biggest fishing net I possibly can to catch every single fish. What I'm doing is I'm doing keyword research. I'm using Semrush. I'm using Ahrefs. The more data points I can get, the better. So, before when I was having a rant, I weren't having a rant saying all data is crap. I want as much data as possible. I want all the Google Autosuggest. I want all the People Also Ask. I want all the related searches. I want all Ahrefs information, all Semrush's information, all my Google Search Console information. And the more information I can get, the better. And then I'm trying to go after everything. And then I've got then prioritise what lanes I want to sit in to move through my traffic tiers. Go through the easiest ones to build up from. So instantly I'm starting to get some sort of traction because you do know that less than 1% of the internet actually gets clicks. So you've got to try to start moving through those tiers to get Googlebot coming visiting your site often and then working it that way. But I'm a big fan of holistic keyword research, going after everything related to your topic from top of the funnel all the way down to bottom of the funnel. And now it's just added 10 times more work with query fan-out terms that started to come in.
Kasra Dash: That's right. Yeah. Just to expand on what James said, I think one data source that not many people take a look at is PPC AdWords. Like some, some brands have been running millions or hundreds of thousands of pounds and dollars on their PPC campaigns and they never really create SEO-optimised pages. They, they all just have, like, PPC landing pages. There might even be, like, noindex so Google doesn't even crawl or index them. I think just by doing, doing something so simple like that. One other thing as well that I've started to implement a lot more, which I don't think I used to do like four or five years ago, is comparison articles. Going back to what James was saying, like clearly defining the entity. If you can get your brand mentioned alongside your competitors, then that's again feeding Google and it's shown, like, look, these guys, they're an entity. We should also be an entity, especially if you're a brand-new website as well. So th... Those, those are probably my go-to.
Malte: My experience is mostly with established brands moving into the space, so I maybe have slightly different view. I like to look at as many internal sources as possible. So, are sales conversations recorded with Gong.io? I want to know what kind of problems the prospects I'm using, uh, mentioning, what kind of wordings they are using to describe it. Are there customer support test tickets in Intercom or Zendesk? I want to know what problems customers have. Again, what exact words they are using. Is there an internal search? I want to know what kind of searches people are typing in. Is there maybe an AI chatbot on the website? I want to know what people are writing there. Like, these, these internal sources I think are super, super, super interesting, especially for larger companies, to actually understand the audience that they have and actually using the, the language of their target audience. Because many, many companies become really, really bad at actually speaking the language of their audience. My favourite example that I feel I mention in every second podcast I'm a guest on. I worked with pharmaceutical company. They spent thousands of dollars on each content piece, had everything double and triple checked by both somebody with a PhD in medicine and also a lawyer, and then multiple rounds of that if they edited each other's content. And in the end they were wondering why they didn't rank for the keyword heart attack. Yeah, because they use the term myocardial infarction. And while Google might know that these are synonyms, the target audience that is Googling for heart attack does not want to read an article about myocardial infarctions, even though they are one and the same, because they, they just don't know it, right? They need it in simpler language. So, anything that tells you what language the audience speaks is helpful. And when I was at an agency many, many years ago, I always liked to look at the website of trade shows and conferences in the space and look at the titles of the talks that people are giving. That would always give me a bit of an input on, okay, these are the topics in that space that we might have to cover with a blog post or something similar in the future. Especially if you work for, as an agency, with a company that doesn't have their own marketing team. So nobody there actually thinks about what to do and gives you any guidance beyond, we want to make more sales, and then you have to figure out what is the content strategy, what can we do. Then trade fairs and these kinds of things are interesting inputs.
James Dooley: Yeah, for sure. Just one more thing on there. Um, Jabez, right? I absolutely love what he did. Was it last week in The New Search on Skool, Kasra's Skool group, right? He was going through some keyword research and there was something of what he did exactly what you're talking about there, Malte, where we had all the recordings of our sales calls and Kasra used to run all of the Facebook Ads and he used to extract all of the pain points and then use them for ads. And it used to be, like, talking their lingo and he used to do it for retargeting and stuff like that and it used to work so well. And that's where then, like, obviously The New Search, that's what he was initially based that Skool group on, which he's talking about may have been related to AI, but also you need to be now looking at query fan-out and you need to be understanding what pain points your, your potential new customer has and covering that information. So yeah, that's absolute dynamite just there. It works so well. And also it could be, a lot of time, it's bottom-of-the-funnel terms. They're specifically wanting that problem being solved and you're then the brand that's showing that you are the solution to their problem. It's key. I love it.
Jabez: I recently did a podcast with a really prominent AEO expert, Irina Maltseva. She's worked with a lot of YC companies and one of her first steps when she's onboarding any new client or campaign is she gathers all the queries from the help centre of the brand, every single query that has come, and then builds the whole content plan according to that. Which I think is a very sharp technique to work on because that gives you exactly those, maybe not queries based on search volume, but very hyper-specific queries that people really looking for solutions.
James Dooley: My last question for each one of you, what's a test or, or a ranking factor that's worth exploring right now? Maybe very less used right now, but it's worth exploring when it comes to GEO or LLM visibility. Something that we have not already discussed. What one thing that we've noticed, and this works so well, is if you've got, like, a load of directories and, like, a lot of people, part of their plan is, all right, we, we need these 50 directories because our competitors have these 50 directories. Let's just go and get these directories and then that's pretty much it, right? They, they'll go and get these directories. They're, they're, they're happy with that. But what we've noticed is there's certain directories that if you get reviews on it, it, it helps massively. DesignRush is one. Clutch is another. Yelp. There's, there's all of these industry-based directories that you can get listed on, but I feel like not many people are, are driving it forward and, and getting the most bang for their buck. And some of them, by the way, are, like, premium directories. Like, I think DesignRush is maybe three, four, $500, um, for the year. So they're, they're already paying for it, but they're just not doing anything afterwards with it.
Kasra Dash: Adding to that, even the exact text used in these reviews can have a huge impact, which is why some restaurants now make their employees wear T-shirts with a Google Maps and Yelp logo and then a text like Tasty Burger on it, just to trigger them use the exact wording tasty burger. Or you can... This is actually happening. Or you can write it on the wall, like, these terms that you want to have in the reviews. So just to put it in people's mind. Ah, yeah, I like the crispy fries. I like the juicy burger. And then, then people are more likely to use that in the reviews.
Malte: I would add as underexplored, BM25. It's a, uh, technique, algorithm to measure how relevant a document or text passage is for a search term. And I see a lot of evidence that various LLMs are using this in some of the steps used for the ranking and re-ranking of documents for their fan-outs. It can be very educational for you to take one of your prompts or keywords or fan-out queries and your text that you are writing and just look at the BM25 scoring of that and documents who are beating you and see if there's something in there you can find how you can optimise your content.
James Dooley: Yeah, for sure. Just move it on with regards to what Kasra said about the reviews on the citations. The minute you go and get two reviews on, like, let's say a Yelp listing, might be if you've got... 93% of the listings, we went and did a thing, and this with which I think we checked five categories, and 93% of the companies had zero reviews. So the minute you got one or two reviews on that Yelp listing, you instantly, within the category, went from maybe scored in on within, like, on page number six within a category that, that Googlebot's having to crawl, to being on the first page. You're being crawled so much more, which was huge. Hats off to Kasra on that, like the reviews on them directories. But the biggest one for me, and I want... I don't want you to cut this podcast short because I want to start asking you one or two questions, Jabez, as well, and you, Malte, about something that I came across last week, right? And it's discovery engine optimisation, right? Which is what everybody talks about. So what we're... Every, every single podcast that I've done with you, Jabez, all talked about how do we get more citations? How do we get into AI Overview? How do we get into ChatGPT? And it's all about new visibility or being, having the visibility. Nobody talked about bottom-of-the-funnel decision engine optimisation, right? And what I mean by that is we won a job and we've got one or two case studies that's coming out where we've won the job. And I asked the headteacher of the school, why did they go with us? And I found out that we was the most expensive quote. Right? And what they went and did is they went and loaded in... The headteacher thinks she's really advanced with regards to AI and stuff. She loaded in the, the quotes into ChatGPT and said, here were the four companies, or been five companies. Here's the five companies. Here's the quotes. Which one do you think is the best value I should place the order with? And it came back with us. And we wasn't the cheapest. And they came back and said that you should go with them with the best value. And I said to that teacher, if it would have said my biggest competitor, which at the time this other company is called RTC, if it would have said RTC, would you place the order with them? And they went, absolutely I would have. So I would have lost that job. Now this isn't prompt volume. This isn't anything. People are not realising the reputation that's at hand at the 11th hour, that zero moment of truth where they're comparing you against your competitors with the quotes and allowing ChatGPT or AI to answer who are they going to go with. I don't think people are talking enough about decision engine optimisation. I don't think people, when we're talking about SEO or GEO or LLM SEO or whatever you want to call it, we all care about how many new citations, and we're not thinking about actually when we do these comparison articles or when we do these listicles, that could skew the difference between you winning an order against a competitor or not. And to leave it on that, right, your answer, your question of what kind of hack would you do, is go and find five of your competitors and ask ChatGPT or whatever LLM model that you want to use. If I was £10,000 and this company was £10,000, give me three reasons why you wouldn't go with us, right? And find out why. It's got to try and find things that, that might be this company might be better than you. And some of the things what we realised that it came back with, which we had no clue about, it didn't... It didn't come up in query fan-out. It didn't come back in prompt volume. One of them said that we don't... We didn't know what insurance you had and these have got £3 million public liability insurance, which made me then go, oh, on my website I now need to start making certain that I've got to say that we've also got £3 million in public liability insurance. There was an accreditation that we have that we didn't have on the website that they had on the website. And it said these are the, the, the Constructionline Gold accredited, Constructionline Gold accredited. But so are we. And it told us that they would go over them, over us, because the Constructionline Gold. I'm like, oh, they gave me so much information that isn't part of your normal GEO or AEO research. And by asking it why wouldn't you use us is one of the best questions that you can ask. And it forms then down to the entity, defining the entity of who you are, what you do and why you're brilliant. And I think people need to talk about decision engine optimisation a lot more than just new discovery.
Malte: Also, just before making the decision, people research and then, and they bounce between ChatGPT and Google a lot. They will search for something and go back to Google. Uh, we had a client, we tried our best and we were actually able to increase their visibility across LLMs, but they had really shitty Google review profile. Or probably it was real. They had 3.2 stars. And we kept pushing on them that, hey, no matter how much we try, even if you're recommended right on top, when it comes to making decisions you will suffer because people will not just trust LLMs. They will bounce back between ChatGPT and come back to Google to search who you are and how you're doing. And that's why, I mean, adding to what James, you're doing, it's not just about optimising for LLMs. It's very important to, you know, optimise for every single platform and establish your entity. And I think you had shared a case study, similar case study, where you got a similar contract where they decided because you had a Knowledge Panel, right? And that's, that's a big winning factor. When people have to put money in, they want to spend extra, but with someone they can trust, someone they know will deliver more value. And I think that's the part.
James Dooley: Yeah, on, on that, what it was was that was to do with me personally where, and I said, "Oh, you want to go to this, this person." I'm not going to name the person, but I said, "You want to go to this person. He's got a great following, um, and he'll promote your tool." And they went, "Don't go to... He's not got a Knowledge Panel. We only want to go with people that's got a Knowledge Panel." I'm like, really? And this is an SEO tool. They did it. I'm not going to name who it was, but they're like, I'm only going to go to influencers that's got a Knowledge Panel. And I'm like, that's a strange one to go after, but that, that was their internal policy, to find SEO influencers that's got a following that has a Knowledge Panel. If this one over, over here had an amazing following, they're like, "Nope, can't use, got a Knowledge Panel." Am I that strange? It's what, what different people want. But I don't know. It's a strange one.
Malte: I actually once gave one of my teams at a former job this as a guidance. We only take external experts that have a Knowledge Panel because I want people who are recognised as entities by Google so that we can profit from their trust that Google has in them.
Jabez: Yeah. Just one, one thing going back to what James was saying about decision making optimisation, and I've got a, a similar story to, to Malte's burger story. I remember I was speaking to an SEO. I don't remember who it was, but he was an SEO for a massive enterprise company, and we were just talking about which tools and stuff we use. And I was like, "Oh, well, I use Ahrefs for this. I use Semrush for that. I use Surfer SEO for X, Y, Z and stuff like that." Oh, well, what, what tools do you use? I always enjoy listening to people that are in enterprise because, in certain cases, they spend, like, hundreds of thousands, in, in certain cases, on tools. And he was like, "Oh, yeah, we just use Semrush." I'm like, "What about for your content?" He was like, "Yeah, just Semrush." I'm like, "Well, why?" He says, "Well, our finance team, every year they do an audit of what we're actually paying for and we just want the one invoice. We don't want an Ahrefs invoice, a Surfer invoice, an X, Y, Z invoice." He says, "Surfer ticks every single box for the finance team." I think the two tools that he used, one was Surfer and the other was Screaming Frog. Because he was like, "Screaming Frog does certain things on the audit side of stuff." And he was like, "Yeah, that's literally all our team looks at." He says, "They'll go, they'll load up the websites, they'll look at the pricing charts and like, all right, okay, this does content, this does social media, this does a little bit of SEO, this does link building." Boom. We're, we're happy with that. And then the other story, which is very similar to what Malte was saying about the burger one. I remember I was in Bali three years ago, whenever it was. There was this ice cream stand and they had, like, literally a shack, right? Obviously went and got my ice cream and they had one of those, like, little... The ones that you tap your phone on and it leaves them a review, like, it redirects you and stuff like that. But what they had done, which was really interesting, is every time you tapped it, it would auto-generate the review for you and it would have the keywords there. I'm like, how has this shack got such good technology? And they had, like, 9,000 reviews. It was insane.
Kasra Dash: Yeah. One more thing that, uh, you shared, Kasra, regarding the directory in the sources. Just today I found out we were sending a report to a client and list of all the directories they need to be on, and I'm starting to see a good trend of just what we call, you know, EMDs, PMDs, satellite sites. Same with directories as well. Like brands have made their own directories in their sub-niches, AI, fully AI generated. And this is my guess and I can be wrong. One of the reasons I see these EMDs, PMD satellite sites getting cited a lot consistently, that their publishing date is updated regularly. That's one common factor I've seen. Every single time we retrieve, their date is always the latest date. And yeah, that's a test that would love to test out as well, creating our own. Is that, is that by any chance... The EMDs and PMDs, is that by any chance on ChatGPT? I've seen a similar trend where ChatGPT seems to like EMDs and PMDs, and Google just doesn't anymore after the spam update. It seems like it just culled, like, 60% of EMDs and PMDs and stuff. We've seen many sites that were... They're not even indexed. Only one page is indexed on Google, but they're fully indexed on Bing. And more than EMDs, PMDs, what... Like, calling them as satellite sites because they are, like, hub sites fully focused around one topic. And they seem to be ranking really well. And by ranking I mean being in the citation section a lot.
James Dooley: Guys, thank you so much for your time. I will be sharing your socials anyways. Uh, besides social, do you guys run any newsletters or any other platform or any other Skool community? I know Kasra is doing a Skool community. Something that you would like to share, I'll add that to the description as well.
Malte: For me, it's just my LinkedIn. It's where I share what I know.
Kasra Dash: Probably for, for me, probably YouTube and then Skool. That, that's, that's, that's where I'm most active on.
James Dooley: For me, definitely it's my YouTube. At present, I'm going to be running a lot of AEO case studies or GEO case studies, whatever you want to call it. I'm doing it one with regards to visibility of, like, the discovery part, and I'm doing one with regards to the decision engine optimisation part and showing people and monetary value of us winning work because of AI visibility at the 11th hour. I want to try and start showcasing this a lot more because I think a lot of marketers are obsessed with vanity metrics with regards to clicks, impressions and stuff like that. I want to, as a business guy, want to show physical results. If I didn't rank for this or didn't get cited for this, I would have lost this £50,000 job. If I didn't get cited for this, I would have lost this half-a-million-pound job. And these are real. I'm actually giving the facility of where we got the order from and the company name. I'm showing some real-life examples and they're going to be going on my YouTube channel.
Jabez: Thank you so much, gentlemen. This was very insightful. I hope we can do this again.
Malte: Thank you.
Kasra Dash: Thank you, Jabez. Cheers. Love you, guys. Thanks.