Semantic SEO in the AI Era with Koray Tugberk GÜBÜR

I believe the majority of them already already left SEO. That's why actually the industry right now has a better focus for engineering and they have a higher way higher respect. Welcome to revenue optimization with Stats Drone. This affiliate marketing podcast is brought to you by Affiliate Track, which is an affiliate link research tool and affiliate finder and also brought to you by Stats Drone, the best stats and analytics tool for affiliates, agencies, affiliate networks, and affiliate programs.

Let's start with what we were just chatting about like two seconds ago, which is like, you know, I I'm building a uh a product that uh scrapes the internet of just affiliate sites only. It has the affiliate link linking to a product, whether it's like casino, SAS, uh SEO, affiliate network, and we're also downloading the page titles. you were just telling me that like the the capabilities of these page titles is immense because the the basic search is one thing but then there's the semantic search

which is like what is the meaning behind it. So I'm looking at actually vectorizing our own data which is going to be a lot of work but I think it's worth it and I'm assuming that's worth it. it. it. >> I believe uh actually every SEO should be doing it like around four years ago I created a GitHub repo. It's still actually in my profile in GitHub.

they can check the username which is KTG1. uh I called it actually s beat which means search engine result page beating which means that actually I just because in that script I am checking actually search results every second every second I search it then I take the data every second second second >> the main trick there is you realize that after a point even every second the rankings are keep actually changing and there is a pattern in these ranking changes and after getting let's say 60,000

snapshots for that data I was able to actually create the animation of these ranking changes by seeing that every certain site is changing their place with another certain site. It's not randomly changing. There is a pattern of these ranking changes from second to second. Another thing is that when I get all these data from the SER, I was also scraping the results to see actually why these websites are changing with each other.

For instance, the second site suddenly becomes seventh. seventh becomes the second and why actually they are replacing each other based on this let's say data and I was checking let's say whether they have a number in the first sentence whether they have let's say a certain predicate in the title pack or whether they have an HTML form element converter or any type of input area if you're able to basically run certain type of let's say similarity checks or certain type of uh let's say entity attribute

relations uh based on these uh let's data that you have. You can actually run these data with the historical changes of Google and you can see what websites are actually increasing in the rankings and what websites actually are not performing that well which will be a great actually snapshot of everything. Other way you have to spend like your three months click here, click there, check this and check that and if you have the data already with your tool it will be helping actually greatly for the

plate marketers. Okay, that was called search B. search B. >> Uh, search B. Search beats. Let me just find it. find it. find it. >> Search. >> Search. >> Search. >> Yeah, maybe I have. >> I think I'll be able to find it. >> Yeah, it's in my basic GitHub. It's directly there. And I created it like there is also a webinar that I did earlier with Ranksense from Hamlet Batista.

Batista. Batista. >> Okay. >> Okay. >> Okay. >> He's not around with us anymore unfortunately due to the co uh situation. uh but he he was also doing great things here and I just wanted to show him to this and he wanted to do the webinar as well. >> Okay. So yeah, like I said this uh I I went down a rabbit hole a couple months ago spending more time doing um uh building products with AI tools and then I'm reading a book and the book

is called um semantic AI search with vector databases by I don't know if I'm pronouncing the name properly but it's Emlit B H O T E. uh it came out recently and uh I'm doing this because of that project I mentioned before. So I'm like okay I need to start analyzing databases and getting better with like um you know just doing search. So what you just said uh a couple minutes ago.

So would you say that people should take the content of their own site and also vectorize it and maybe produce a better search engine within their own site? So uh if you have programmatic SEO site let's say you are ranking for 7,000 setrains like in the CBD niche in that case managing that content is going to be really hard. If you have 40,000 products in your ecom site again it's going to be very hard for you to actually manage everything.

If the project site is big I will definitely vectorize it and I would actually also standardize my sentence structures. For instance, when I create a templated content to decrease the repetitiveness and increase the originality, I use certain type of synonymizations or I change word orders and I try to add also uh let's say different criterias because in Google's language system they have four main actually principle. One of them is factual content, opinionated content, structured content and unsstructured content.

Which means some sections of the content should be tables and the lists which is structured. Some sections should be paragraph and some sections especially should be reflecting opinionated angles uh perspectives and the experiences as happens in the Reddit or qu or any type of social media site you should include these angles there. And when I do these type of vectorizations, I'm able to see the percentage of factual content or percentage of opinionated content which actually helps me to correlate it with the affiliated especially for the

affiliate industry is important because since 2023 the the content that is factual with certain perspectives also reflected in the language is actually favored by Google especially user generated content heavily has been promoted already even Google's API leak. If you go to the Google's content API leak for content warehouse API league and if you search there UGC especially UGC change frequency you will see that there are lots of metals there to see actually how frequently the user generated content is changing there.

If you go to the Google's quality guideline, you will see human effort. Total human effort is a really important signal and you can see also content effort as a property in API lead and they say that content effort is measured with large language models. So in short, vectorization will be very smart especially changing thousands of pages programmatically while also doing that try to decrease repetitiveness to signal originality as well.

Okay, that was pretty heavy. Uh we we were just kind of like joking earlier that, you know, as I was studying this stuff and reading this books, I'm like, "Wait a minute, you were talking about this stuff like four years ago or longer." And it's like, you know, I I thought I understood semantic SEO, but now that I'm going deeper down the rabbit hole of studying like how Google works, um vectorizing of content, um the databases, um I realize like it's a lot deeper.

And I think you said earlier that you believe that a lot of people didn't understand semantic SEO back then and maybe a lot of them still won't today. today. I believe the the majority of them already already left SEO. That's why actually the industry right now has a better focus for engineering and they have a higher way higher respect. And I already know that actually when I did the interview with Matt Deidi years ago, I know that actually some people were saying that oh topical maps

or topical authority it will be just for seven days thing people will forget that and these people right now they are trying to sell topical maps and they are trying to actually try to own even the concept of topical authority and yes years ago when we were explaining these things some people were claiming that we are trying to make the SEO harder. Now actually we were trying to let you know that these are coming for your your way and people who saw it they joined the

community and they basically do a great work right now. We published as I said ear to like around 400 success shares in the community publicly and the ones that they didn't follow I believe they left the industry already and or either they are trying to read adapt from a certain level but I can tell that especially if you started SEO around 2012 13 uh I usually see two type of people like there for instance in our mastermind in Turkey there was a person I can name

Luke Bustin he is from UK and he's a great person and He actually improved himself greatly. He studied the course. He even went to the actually information retrieval conferences and he said to me that uh professors in the area of information retrieval they actually hate SEOs because whenever we go to a conference we try to monetize their knowledge and they hate from that situation. So Luke Baston here is a good example.

He changed he started on that era and he changed developed or developed new skills and understanding and majority of them the people who started SEO around 2010s I believe they still have old habits and majority of them left SEO by saying SEO is that already I can tell that the year that people claim that SEO is that was the year that actually we did the most money already.

So I'm very happy that it is keep dying. keep dying. keep dying. >> Yeah. >> Yeah. >> Yeah. A and the irony would you say this is true or maybe not true that with the new AI tools that are available that it's actually making semantic SEO a lot easier to get into like a lot easier to understand maybe even the research itself. I actually agree with that out of three four reasons because when I was starting these things around 2017 to just read a single patent

I was spending my heart because you need to keep everything in your memory and you need to remember okay I I remember this engineer and he mentions this then there is another pattern that mentions the same thing and what is the relation between these two it was like a journalism I could analyze even I don't know Satam Hussein's actually government in the same way it it was a heavy work actually but Today you just actually give a patent to CH GBT and tell explain it to

me and he actually explains everything to you and then also he it tells you what other patterns exist from the same engineers. It makes things very way very way very easier. I can tell sometimes of course it it is hallucinating by giving some fancy random verse that doesn't have a proper meaning on that section. You should of course cow shoes but I believe it made everything way easier because you can give a pro typos a paragraph to GBT and ask the GBT basles here entities the

attributes or what are the sentence let's say number of unique words in this area or what are the let's say hypermes or the hyponyms and they are passed there and you can see the anotomy of semantics in every text unit easily with the LLM so it makes it easier in my opinion 17 opinion 17 Then would you say that can like are we here today where I can go into Gemini perplexity and be okay I need you to act like you're Karai and I want you

to be my my SEO advisor um analyze my site what should I be doing read all of Karay's uh material uh download all the podcasts through what is it notebook LLM and and and >> they do it already my community already do that and they keep sharing actually my core box they call it And they it also works too because for instance I don't remember every podcast that I did like through the years but AI remembers it sites these things very in a very good way

and finds all the connections and suggest in a very structured way I can site that actually it already works and many people in our comments already actually is using them. >> Okay. I I need to look up some krybots but uh I did try to try a couple examples of that before recording so I was like I'm sure I could do this right now. Yeah, definitely. I believe so.

>> Okay. So, over the last couple years, I've interviewed so many different SEO people and there's always been constant talk about both AI and AI with content uh generation. And I've been personally scared about using AI because I've seen some people brag about how they use AI to make all this money and then they went quiet because their site got destroyed. And I've seen a lot of destroyed websites.

So is there a right way of do of not only using AI for content research but actually creating content itself? I believe there is because we created actually AI websites in 2021 way before CH DBT and I explained these actually websites in my Saigon speech in 2023 and I remembered actually how people were shocked on that speech too because they saw it already because they were already discussing this like is it safe or not and when I showed them the websites that actually already are hit

even two years earlier than GBT era they were understanding okay this is already actually covered by Google but the main thing there is they usually cover it from your momentum because when I created a topical authority concept I say that there are three principles you cover it actually wider you cover it deeper and you cover it faster the fast section I called it momentum if you cover a topic let's say you publish 8,000 uh strange pages on one day initially it ranks really greatly it indexes

very fast but after like around average six weeks later your website is going to be hit and it will be dropping in a very great way So that's why I usually suggest actually for momentum stay in the human range. Don't publish everything at the same time but just distribute it at a time. The second part is that the humanization is a big part because we always have a blacklisted blacklisted nouns, blacklisted adverbs, blacklisted adjectives because if you look at the AI, it is actually called statistical

semantics because AI distributes certain type of verbs together with certain co occurrences and this usually called actually styometric. It has a style already. When you see that style, when you go beyond that actually let's say ordinary people's angle and when you see actually the anatomy of the text that it creates you're able to actually hide the signatures in a better way because I'm able to say that never use this word always replace it with one of these actually synonyms you're able to create a decision tree

for your own industry by giving an SEO dictionary sentence structures and order of these sentences and you can always actually create unique content thanks to that but you should have very strict system instructions and a very let's say a good uh blacklisted word uh component to actually write it in a proper way but it works already. works already. works already. >> So maybe a quick example of this would be let's say if you want to do 8,000 affiliate program reviews you do it all at once

you're risking a problem but in order to get through 8,000 human you know reviews of 500 words that's going to take a long time. So maybe as an idea that you could kind of take like a like create mini databases where you know someone's manually going through the current site um they're trying to document what they see maybe even what's been written on the internet about them and then kind of like do a mini summary feed that into AI and then use that for >> there

is example site for it actually I always give that website as a bad example and as a good example though it is topfires reviews.com it should be as much as I can remember it's a good example because it's also programmatic. It always uses actually Amazon reviews. >> In my Sophia speech in 2024, I explained actually how we use existing reviews from Amazon, Reddit or other places to scrape all of them and combine them with our large language model in a very unique way in a way

that actually Google sees it user generated content. Okay, it's originality. It also does the same thing. It is also a better example because if you check the way back data for the era that they are losing the rankings you will see that once they get the rankings for firearms they also start to review hotels bicycles bags etc but so it was diluting the topicality and decreasing the site focus score as the API actually says and they were hit once they removed this unnecessary content they came

back one more thing is about the images Google doesn't index text every image and when they go to the Amazon reviews or products to the page fifth, six, seven, you will always see unique user generated images there and that unique images on your pages also contribute to the uniqueness and the originality will sees that image first time on your page as well. So these type of many other angles actually can be used as well.

>> Okay. Well, I'll give you an example of something I've done but I haven't published it yet but now I need to publish it. So I took uh I was doing a presentation at a Sigma conference um last year and the topic was why online casinos close. So I analyzed Ask Gamblers database, lcb.org database. We also have a database of closed affiliate programs and I said what's the signatures behind it?

So, of course, I went scraped a lot of data, ran a lot of Python, and then I started looking at, well, what were the signatures? And one of the signatures that came to the top was uh most of these sites had really bad terms and conditions for players that said, uh, you can't cash out more than like $2,000 per day or 5,000 per week or 10,000 per month.

It was pretty bad. And that was actually most of the the the co-work. So, I think by publishing that data could be useful. And I think another example uh just to use your example would be taking casino complaints which is ask gamblers lcb.org for casino guru. I mean I don't understand why people don't do this. Like if I was an affiliate today I would create an aggregator site and say I'll tell you everything you need to know about all the complaints not just one site.

That's true actually and in one actually s for instance in the online dating industry it is the main competitor of Tinder but I can't give the name due to the NDI on that company for instance when it comes to the subject of senior dating we added some design elements to the page. These design elements actually have some polls which means that actually let's say we ask a question like what is the best place to meet with seniors over 60 years old on that answer we don't

answer it we just give let's say options on you choose the answer on that poll we show the percentages and AI is automatically writes the paragraph based on the user inputs which also adds actually user generated content angle with also uniqueness also shows engagement And at the most bottom part, we also add the most asked questions directly as user generated content. And then we use directly an entity person from the Google knowledge graph on these topics especially to answer these questions as well.

But this also ask basically human effort one more time too. So many people still try to rank with ordinary blog layouts. So try to always include these polls or Q&As or some forum components as much as possible to avoid some filterings of the Google system today. Yeah, but that's also a great content generation idea because I do polls on LinkedIn and sometimes the answers I get surprise me.

I'm like, I had no idea this is what I was going to get for an answer. And because I get those answers, it changes changes a lot of things for me. So I think I think this would be a total content playbook here. >> I hope one day people will be implementing these more. >> I I'm going to find clever ways of doing it and AI tools make it uh really easy.

I'm actually going to experiment with the the podcast website we have now. So I built it in I'm pretty good with code but I actually went to lovable and it's just super easy. >> All is really surprisingly good actually. And it is able to integrate everything with GitHub, WordPress or whatever CMS that you need. It's really good strength there. >> Yeah, I I'm definitely going to do that there because I've already um I've used the RSS feed to autogenerate each podcast episode and then now I'm

going to like you know we're talking geek talk. I'm going to go into nen download all the audio convert it into text and then I'm going to do a whole content playbook which is a it is another question I've got later on uh for this. Okay, so I want to talk about I I was actually using a lot of AI tools and I was starting to build some Google search console um uh dashboards and as I was doing I was like wait a minute is this

going to be a waste of time like is Google search console going to be totally different in a year from now? So where do you think a tool like search console is going to be as we change from like the regular Google to you know the AI version? So the first thing there is I believe over three years search console is already not giving maybe 40% of the real data or the value I can say was the first thing is there there are so many new

surf features we even didn't name them as SEO industry unfortunately I will I have my own dictionary there some of the names that I use might not be that clear but for instance when you search for the uh let's say the best uh SEO tools let's Okay, they usually give all the brands at the top. It is actually in the patterns they call it generated ranked entity list.

It's the official name in the patterns. My naming there is basically hyperser I call it because it's like a SER inside another SER. That's why I call it like hypersurer. And when you click to the one of these brand bubbles, it opens another accordion and gives you the three top results for that index as well. Let's say you search for best SEO tools. I don't want to advertise anything there but let's say you do one thing thing thing >> and then actually shows you three results

from that specific index which means actually it connects an index to the another one >> and technically right now none of the SEO tools actually track what entities appear on that list and inside that actually bubbles what it ranks it's not also not known if you go to the what people say surf feature which usually LinkedIn Tik Tok Instagram Reddit type of places are ranking there we still don't know actually which brands social media or which brands influencer is ranking on that part nobody cares nobody

checks and I believe Google search console also doesn't give this data on purpose because if they start to give that data SEOs will realize it is important then we will be asking even more so they also want to keep us in that dark zone the one key difference is that at the 8th December 2025 they announced that they are integ rating social media channels into the search console and I believe it was the biggest change of entire year in my opinion because uh in 2023 they

announced helpful content system by saying that helpful information for people on social media and forum sites. This was the exact phrase and still they keep doing it because Google's moto is very clear for SEO or for search. It is basically people for people in the search. They want people to create content and they want people to find people content on social media sites and the forums are the main principle right now for them.

And that's why they are integrating your YouTube channel, your Reddit will be the next one and other social media channels to do your search console today. If your YouTube channel basically gets more clicks from Google, they will be telling it to you and it helps them for two reasons. One thing is that the search console will be topic based which means that in the future rather than telling you what query brings you the most clicks they might be saying you what topic is bringing more clicks

as well. The second thing is that they want you to also they want you to create more content by showing your face in a video in a podcast and they want you to create more content for their integrated platforms like YouTube or the Reddit. If I add one more thing there is if your website gets fewer clicks due to the AI search features when they give you the social media clicks in your search console they are actually telling that yes your website gets fewer clicks but

your socials are getting more so focus on there they are trying to tell and in 2019 in one case that I said that if you have a YouTube channel there's a higher chance that you can actually save yourself from these core updates because ownership of a active YouTube channel pro directly proves that there is a brand attribution and there is a business behind actually everything that you do as well.

So I believe the search console will be very much more integrated with AI ser features. So you will be using all these filtering filters with AI or LLM. It will be it will be a kind of conversational data analyzis tool not for the queries mainly for the topics and it won't be only for your site. It will be also for your web entity which involves all of your alo social media channels and maybe even beyond as well.

Okay, that was pretty heavy. But um that's actually something uh I was actually planning to do on that initial project where we we scrape and we find all the affiliate sites. Like we're actually scraping all their social contacts like do they have YouTube? Do they have Facebook or X or Reddit and contact forms and addresses and we I believe there's a correlation there, but we won't know what the correlation is until we actually start saying here's affiliate sites, here's all the links they have on all

their pages and here's everything. So I think there's going to be some like that's obvious I think um but I think a lot of people they need to see the data before they they have that aha moment. aha moment. aha moment. >> Yeah. If I would think about maybe one more functionality for that tool, if you allow people to integrate their search console data with your technology and if they also give you the log files, you can see actually whether social media links help Google to

crawl more URLs from their actually let's say uh website overall because comparing some data like this let's say there is a URL it is being crawled by Google bot higher frequency during actually social media shares because if you go to the search console use your inspection tool sometimes you will see that actually Tik Tok or Instagram comes in the referred page section they find your URLs from socials and then they start actually scrolling there you can decrease the indexation delay and you can increase number of

keywords that you rank through actually getting these links from socials I'm not talking about link juice or page rank here it's different I'm just talking about exploration of new content and forcing Google to crawl it more frequently as well. That type of maybe check on would be helpful too. >> That's actually a great idea to ask to say, "Hey, do you want to connect your search console data?" But it goes without saying that the user needs a benefit from it.

But I do think like I've actually built some of these console uh reports where it's like okay, you know, you know how your data is only what 16 months of console data before it disappears. And I do believe that the daily console data is useful, but Google doesn't make it super easy for you to export just the daily. So you have to pull the data once a day.

And I'm basically, you've probably already done this before, but uh I'm basically building this stuff. So maybe there could be some connectivity where it's like, how do we make this useful for someone to connect their search console? Maybe if we give them this new um console reader or console report for free, that might be the intent of the big, hey, I have a reason to actually upload it.

And maybe there's some sort of social not social but maybe there's some sort of big data sharing where the data is not exposed but it's put into a a giant database. Many other things can be done with the search console data. I believe for instance you can always create custom click curves every seven days. is you can tell them actually what pages are increasing in impressions from what queries or you can give them number of URLs that give that takes impressions for the same queries which

actually shows micro cannibalizations. You can always actually uh easily export your data to the bigquery automatically as well and from BigQuery you can always try to actually see for instance there is a seasonal event every year for some certain certain let's say types of situations and you can see actually how your CTR is changing during this seasonal events or also you can import your log files and you can see actually your traffic data and the crawl data how they are correlating with each other many other

things actually can be done there uh actually >> it's a lot of data >> especially your actually vector database let's say you vectorize my content and you have my search console data as well you can sometimes uh because in semantics actually if you want to do semantic SEO the first thing that you should be doing is decrease the dimensionality dimensionality because I can't keep adding new data new correlations new causations many other things I my script in the GitHub I have 2,000 columns.

I I have taken everything there. But actually seeing the most important thing there is you don't need to check 2,000 columns to see how this website is ranking or why it is not ranking. If you decrease the dimensionality of the data you can just check some simple things there. For instance, in a simple way just compare actually your crawl frequency to the actual content updateness let's say or your update frequency.

Compare let's say your number of embeddings when you vectorize the content. Compare your number of embeddings to the number of queries that you are ranking for or compare your let's say number of heings to the number of queries or percentage of this impression increases as well. All these actually will be more helpful for you to see what I should be increasing most in that area. If you start to deep dive into the every punctuation, every word, every in today's reviews, it's not humanly possible.

Uh on that level, I can say, >> yeah, but I actually think that would be pretty cool to give someone the power to be like, hey, here's your page. You know, the data technically is available in search console, but it requires I think a little bit of data manipulation to see here's your pages and here's what you're uh getting traffic for, both in terms of impressions and clicks.

And I think if people had easier access to this, this would actually educate them on the actual basics of semantic SEO to be like, "Oh, I didn't realize this was an important term." Or, I mean, I don't know if this is possible. Do you think it there's a future where the search comes in from something really obscure, so it's not a lot of clicks, but maybe the value of the click is worth more?

So, we have a tool that that has dynamic variables. So, it's click ID level tracking. So, we can actually map each individual click. Now, I don't know if Google Search Console has already switched to by the minute or real time data because if you actually know what click came from where, maybe you might be able to match the search query to the actual click going to the affiliate program and actually getting the revenue associated with that click.

>> It is uh it's hard to do because the thing is that they're anonymizing or K anonymizing the data. Uh if you manage to do that, they will be very against it. They might sue you. if possibly or hide it basically from them at least >> or or acquire me. >> Yeah, that's possibly I mean you can try to run maybe timestamp similarity. There is an organic click on a certain time stamp and if you are able to also see the session time stamp and if

they are close to each other uh closer than let's say 1 second probably it's the same user and from there you can assign basically this user ID below to this clicks and from there you can actually track them but legally I'm not sure how it will be and and and >> it goes without saying you have to start with the user saying do you agree to these terms and conditions and cookies and you start there with permission Once you have their permission, I mean that would

be phenomenal to be able to match the revenue. >> True. There is one small tool actually like I guess it was a keyword hero. uh it is a partner of Google Analytics for and they are doing it but they also do it in a like fro not like for every click like a level of clicks they are trying to give you uh but they are having a special permission from Google Analytics as a partner and even it doesn't do more than maybe 25 person I can

say can say can say >> okay I I've got some homework I'm gonna give it a try because I'm actually building another building another uh small project where it's basically like it's in I gaming so obviously you have recurring revenue which is valuable to an affiliate. So the idea is that when you have a newsletter you're usually sending one tracking link to your entire database. So all users the revenue ends up in a single bucket.

You can't you can't separate it. But with click ID level tracking what you do is you basically are able to assign a unique click ID to each email. So, if you signed up to one of my newsletters that I've built on the side, it starts with permission and then once you filter the flow, I I sign a unique ID associated to you, that ID is put inside the tracking link and that tracking link shows up into my revenue reports.

So, I I could actually see your last deposit. Um, so I know if you stop playing at the casino whether the casino removed you from the database because they want to steal from me or you actually generally stop playing. So um I think there's >> great actually surveillance in that area. It might be closing a very good good cap gap actually for monetization as well. It's it's a nice nice design.

>> Yeah. And then of course uh the thing that you could build is taking the user going back to the website as long as they're logged in then you can start putting that parameter in all the pages and going what pages and content are they consuming and as they send emails to you. So rather than sending a newsletter like email, you actually send a conversational one like, "Hey Karai, how's it going?

I'm curious to know what games you like." And as you reply in text, we convert the text, put it into a database if and when it needs to go into a database. And then u and then I think you can see why I went down the rabbit hole of like, okay, let's start studying rag and vectorizing and all these databases. And databases. And databases. And >> so >> so >> so >> yeah, definitely >> that's where I'm at.

>> It's a nice design. I believe you can if you can basically create it. I believe it will be worth really good amount of millions of dollars. >> Yeah. And I've actually built another component of it which is it starts by taking the user's email and then basically it goes through naden. So it's like you get one of four messages and it's all like AB split testing and then um based on your response like if you if you view the page or if you click on

the link um that all goes back into the database. So I built a database. I used a Postgress to build it because I thought it might be better when it comes to this AI rag vectorization. Um, of course you can use almost anything but um that's where I'm at right now. >> That's also a good plan as well. And also you can always defend yourself against the casino or whoever basic you are affiliate with.

So it will be a very strict proof actually. Yeah, I can. >> And then this is where it can get more uh interesting. So, not all programs give you click ID level tracking. So, there's postbacks over here, dynamic variables over here, and then there's a lot of programs that don't have either of them. Now, if you have a player that's worth a lot of money, what you could do is you can go into the affiliate program, create 100 packing links in advance, and say, I only

want to assign these links to players that I know are actually worth something. So, you're you're getting the equivalent of a click ID level click, but you're doing it manually, but it's still worth it. So, if you got a player that's uh spending, let's say, €10,000 per month, um it's worth giving them a custom link. You know, it's worth actually going customer customer >> VIP customers deserve VIP IDs basically.

>> Yeah. That actually uh segments into a question I have about personalization, which is the newsletter is a personalization concept. Um, but do you think that there's a a new world of personalization in terms of like creating the content and going how do we hyperpersonalize this to users with and without them logging in if that makes sense? sense? sense? >> So when you mean the personalization do you mean personalization of the content on the site or personalization of the content on just a newsletter or >>

more on the website? one. So that part is a little bit actually bringing me to some of my latest SOPs because I started to differentiate domains from each other because some domains and technically of course as I said earlier too what you do for AI features of Google and for the regular traditional search it's not that different to rank in the AI you have to rank your document if you rank your document your passages are going to be processed and ranked and retrieved if your passages

are ranked in the feature snippets or the People also ask also the retrieved and generated and augmented passage also is going to be citing you as well. But for these money sites usually you also have to balance some sections. Let's say I'm searching for car accident attorney Houston for Filipinos or Filipino victims because Filipinos sometimes actually just want a Filipino lawyer for themselves. And if you keep adding every nationality to your web document then actually you are going to be decreasing the relevance while increasing what

Google calls gore. So the personalization that you do on certain valuable domains rather than making it very explicitly you have to annotate it basic by adding the most important ones and annotate the rest from that class basically. And there is a second group of domains that actually I call and I started to call them either permanent EMDs or temporary EMDs. If it's a permanent exact match domain, you can check one of them for instance otoex.com otoex.com and we ranked it like around just two months and

right now it's around 6,000 clicks. Sorry, 60 days 6,000 clicks already exact mesh domain and it's converting really well on that one for instance. It's a permanent exact mesh domain that we want to also manipulate every type of surf feature in these ones too. Actually, I try to be closer to a certain type of vector with your with my brand attribution and the content is adjusted for that.

There is also wasteful domains like I call for the wasteful domains. For instance, I can actually open any page for any long-term question or whatever. And that is actually not even a permanent domain that much because all of my social media accounts or social media official accounts or nonofficial accounts and my wasteful domains that I call them, they are all actually adjusted for creating a consensus around the web by actually creating a surrounding sound around my brand and also around my topic.

It is actually called surround sound campaigns as well. And we try to manipulate LLMs and reach out to these people with hyperfocus and hyper personalization. Even if that domain doesn't rank for the long term or even if that domain doesn't rank for valuable query terms, it's okay because we target maybe 20 words of queries or sentences or instructions. So these are usually uh mainly for actually LLM focus domains.

And if you want to do this level of personalization, usually this is what we do. Beyond that, if I need to do personalization for every landed user to do my web page on that segment, I need to usually use either some JS redirections which Google doesn't like that much because they call it sneaky redirection which might be causing some cannibalization and also canonicalization problems. If I use uh some let's say different content based on the user agents and user agent IP address or user agent location

I need to accept the risk of being diagn diagnosed as cloaking source because if I'm showing a different something for every user including also Google user agents it will be a way of actually cloaking and if I change my URL for some of them but I if I keep it same for some others Again actually it will be seen a bit dangerous by Google bot too. To increase your chance of success there you can try to see what location Google bot is crawling you and from

that location you can always show a certain type of content and for other locations you can show some different content as well but eventually in my experience they will catch catch catch >> both. So if the website is money site I usually try to not do that that much unless they are logged in. If they locked in, you are safe. But if you are serving different things to the not locked in users, I usually prefer wasteful domains, which means domains that I can waste easily >>

to do these type of things. That's challenging. I don't think I've got an answer to to that because like I'll give you the example like we have our website stats drone. We're a tool for affiliates, but we have a lot of different uh personas on our website. We have affiliates, affiliate agencies, affiliate networks, we have SEO affiliates, we have PPC affiliates, uh affiliate managers, operators, and the software.

And it's kind of like trying to create all trying to keep all of them happy is not easy. But I have some ideas of what pages some of them like to go on. Uh but there's a lot of the pages where it's actually suitable for everyone. But it's like I'm trying to go, okay, how do I create content and keep everyone happy? Like do you think that there's a way where we could have like a instead of a geo bar that says you're from Canada, it

could actually be like what is your ICP? >> Might be also that also in a way that's also working too. But I believe for this issue there is not a 100% perfect solution. There might be maybe just one thing that also might be harming less while helping for personalization. It's something I usually started to do when these LLM starts to be more important. I still do it for some pages.

So technically uh I just choose some zones on the web page and I just do personalization for that section and the percentage of content that I change with the personalization is not over 10% of the text. That's the first thing. I never personalize the content or customize the content from the important headings or important sections of the HTML. The third thing is that if I do personalization, I use data no index snippet which means I tell Googlebot if you come to this page do not index

this section of the HTML which decreases its prominence for indexation or rank of or ranking purposes. By using that approach, I am able to also basically constraint the personalization to the non-important sections and I also tell the search engine it's not for you. Don't index it as well. So this is the least or let's say most possible healthy version there. Still not perfect but it might or at least for LLM angle it is working because for large linkage model sometimes I'm adding very big chunk of

text to the bottom but I tell Google bot it's not for you. Don't index it. don't check it, don't care about it. >> So this type of maybe differentiation with some certain data snippets instructions actually is helpful. >> That that might be a good solution. It's kind of like, you know, let Google know that it's like, hey, I've got five pieces of content. They're all personalized and it's like one of five personas and it's kind of like, hey, if you're this persona, click here and I'll

tell you something you need to know about this page. >> Yeah. And it's not for Google bot basically because technically if they see something no index to save the cost they don't even judge it. They even don't check it. They have a kinds of basically pattern behavior there. If for instance if the page is no index they even don't render the JavaScript >> to decrease the cost.

So if you say data no snippet uh so data no index for a certain part of HTML they wouldn't run a proper algorithm to check what is actually going on in that area too. >> Okay. I've never heard of that before, but you know what? As we're scraping all these affiliate sites, that's going to be one thing we're going to add to the list. We'll be like, who which affiliates do this?

Because my personal take is if it works really well, then we might see some of the best um SEO focused affiliate sites ranking at the top and if they start having these signatures on their code, uh we'll find it and all that. I believe you can check also Google's own documentation. They have some of these uh these type of things there as well. If you create some different URLs uh with JavaScript redirection also try to use there is one instruction that you can use for Google

bot it tells until what time it should be crawled. be crawled. be crawled. >> Okay. >> Okay. >> Okay. >> For instance you create a alternative version and you redirect only that user to the air that zone and you can use base like shorter than two weeks. They are very tolerant for two weeks time zone for AB testing or for these type of sneaky redirects. They don't see it's sneaky if it is shorter than two weeks.

If you create an automated URL change system, every two weeks you can change it to another URL and you can again tell after this way it's not needed anymore. So don't try to index it or don't try to canonicalize it. It's just for some small portion of users for limited number of days. Basically these type of signals actually can uh let's say help you to personalize the content without being flagged as cloaking source in that area too.

area too. area too. >> So in other words SEO is dead in 2026. >> Yeah. I mean I hope it keeps dying. As I said whenever it dies we make more money. So money. So >> no I believe that to be true. I mean it's kind of weird that I left the affiliate space and I I long time ago I did SEO consulting and I just be like wow if I actually stuck in this space I mean that there's so much money to be made.

Yeah, definitely >> especially in affiliate industry I can tell for let's say Scandinavian countries countries in I gaming every country like for instance even if you go to the Nigeria especially Nigeria actually they gamble a lot but let's say even Somali even the countries with civil war they keep gambling and gambling and gambling and >> and the the internet on these countries are virgin and you can easily manipulate things there actually with hack domains expired domains exact mesh domains and many things there so you can

use technically. technically. technically. >> Yeah, >> Yeah, >> Yeah, >> it's infinite. I mean, we we have so many different types of uh affiliate customers. I mean, we we don't see their data, but I mean, we we a lot of them talk to us and they tell us what they do and it's um yeah, people keep asking us like they're like, "So, SEO is dying, so therefore our business at risk.

I'm at risk, we'd actually be dropping and we keep getting uh we we keep gaining more customers, but we actually are increasing our our our proof like all the time. It's just like because we have more advanced features and they keep using them and it's like look affiliates is not dying. >> True. I can also tell maybe one definitive thing about whether SEO is going to be dead one day or not.

In my opinion, I say this earlier too and I'm saying it actually with very high confidence. If you want to kill search engine optimization, first you have to change the economy model. If you bring communism one day, yes, SEO will be dead with many other things. But as long as you have free markets and the free will, people are going to be searching. For searching, there will be a search engine.

For search engine, there will be optimization. It's not going to change. Maybe you will be searching from your narons. Maybe you will be searching direct in your brain. You won't be looking at the screen. Maybe you'll be using your brain directly with narrow link or whatever. But even then there will be search engine optimization even for that too. Eventually it's going to be always working. So you need to change economy model to kill it to be honest.

Well, hopefully we don't get the communism version where it's like post on social media can get you in jail. Will you surges get you in jail? >> Many other many other things there actually I can't say. But I believe that's the only way of killing that. Other than that, I don't think that it's going to be ever. It will just keep changing, but it will be also keep existing.

existing. existing. >> All right. I'm confident in that. All right. So, I'm going to wrap up with the two more quick questions. Um, so with this podcast, like I've been doing this for two and a half years. I haven't actually took a lot of the content from it from audio to text because I used to use descript which is a lot of work in editing but I know that the the tools now are like even easier.

So I'm about to build an edit and automation convert it but I don't want to just publish the transcript blindly. I want to do lots with this. So what would you advise whether it's me as a B2B podcast or anyone who's doing a BTOC podcast as an affiliate marketer? Well, great question actually and again it is touching to the my latest SOP because we created an SOP for law firms and in that SOP we turn every lawyer first into an entity in a knowledge graph then

we open official site for them then we create all the social media accounts for them including IMDb professional account then we add every podcast by turning every podcast into entities in IMDb and assigning that lawyer as an actor in the podcast. podcast. podcast. >> Okay. And then to do official site we also keep adding every podcast episode and sometimes we also create a separate site for the podcast itself too.

While doing that we turn every podcast episode to the text because we know that when we keep adding this podcast to the Apple podcast or Spotify or whatever else all of them basic being consolidated by large language models and Google and then they also actually change the large language model answers directly as well. So the text there is actually used by LLMs in a very good way.

We distribute to the official side of the entity and to the IMDb and to the many other social media accounts. And then we also use certain GPTs or N8N as well to turn this transcript or transcripted text to the tweet floats or to the Reddit threads or to the qu basically post and every prompt basically takes the same text and turns it into something else. Then it doesn't end there.

We also take that episode and we chunk it to the smaller pieces because episode is designed in that way and then we also add a kind of uh sound way to do that that thing. We turn it into a video then we post it to the YouTube and then also we basically make it YouTube shorts and Tik Tok shorts, Facebook short videos and many other things as well.

So it is a kind of content reformatting, repurposing and syndicating with different angles. Then you will see that actually you are creating so much content from alternative sources with every type of format like text content, audio content, video content and then eventually you start to change the answers of LLMs as well because the moment that you're an entity and the moment that you have too many documents and the pages that are mentioning you, LM also start to site you and since we have too many law

firms as well, we also say things like this a law site says that basically owner of this law firm from California states that we keep citing you. We keep mentioning you from other low sites as well which also creates one more time entitization and also trust around that specific entity and his or her works as well. So this will be one of the things that people can do.

>> Okay. So basically if you weren't doing a podcast before this is a great way to build authority plus content. >> Definitely. >> Definitely. >> Definitely. >> Okay. So I did the right thing years ago without realizing what I was doing. I was like, "Let me just do this and figure it out and try to get some customers and it's paid off in more ways than one." than one." than one." >> Definitely.

>> Definitely. >> Definitely. >> Second question is uh Yeah. So, we have uh over a few thousand affiliate program reviews. We have a site um that we compete with. It's not a direct product compare uh competition, but it is uh get lasso.co. So, they have affiliate program categories and reviews of affiliate programs. I do think their content is paper thin. Doesn't matter. They rank really well. uh what should we be doing to improve our rankings and authority for we want to be like the number one

affiliate marketing company like when it comes to the B2B side so in that area actually I can tell that many people focus on actually the text mainly around 2023 around 2023 even the half of the 2024 it was still the main focus but right now Google is mainly focusing on your layout okay >> they are able to understand whether you are AI generated or not from the or visual semantics.

I will give two names. One of them is Alexander Nayor. The other one is a Michael Benderki. If you check these two people, you will see they are reshaping the goals ranking algorithm right now and they heavily focus on your layout. If you're able to use annotations and specific visual semantics by putting the same data in a better structured visual way, you can again keep increasing your rankings in a better way as well.

I will be focusing on a little bit in layout there. The second thing is that without bloating the page too much. I will try to add more information with the same data. Even if the data is same, you can interpret the data in a different way as well. You can add different comparisons, different declarations or different let's say uh clarifications around them. One more thing is that defining a concept is always making you an authority whenever that concept exists in the queries.

If you want to be an authority on B2B affilate defining B2B marketing and also affilate marketing and connecting these two each other and distributing all these links and authority back to your review pages will be a good approach as well in the period. Okay, that was >> pretty good. And just something that I wanted to share that because we have this this uh research tool for affiliates linking to operators, we want to try to put that data back into the reviews to say what are some

of the affiliates linking to this this brand. Um and then also the increase or decrease of links because we think that's actually a signal where it's like what happens when the five biggest gaming affiliates remove a brand. That's a pretty strong signal there. We don't know what's behind it, but it could be it. It's just not good. Let's put it that way. One more thing is that by the way you can also try to defend these user generated content signals by adding submit a question feature

or by asking people would you vote this let's say a late program let's say for this this this criteria and then if you go to the Google open the visa AI overview future by using let's say USA VPN or whatever and just try to search for this give me top 20 a programs for these three criteria whenever you change these three criterias you they're going to see that the answer is changing and it means that actually they are able to judge basically for these criterias in

a very clear way. If you add these criterias and comparison for these criterias, definitions for these criterias and also if you allow users also review these programs or products or services for these criterias that will be a very good approach there as well. For many sus we use this de top five up to 20 for this with criteria results. We secrete these results. Then we try to see which criteria is associated with which product, service or brand.

And from there we try to actually exchange the consensus with the wasteful domains as I explained earlier. explained earlier. >> Okay, I got lots of homework to do. I got so many pages of notes. So let's end it there. Uh Karai, thanks for for doing this. I want to pass it back to you. How can people get a hold of you and tell me what's new on your site?

>> Thank you so much. And I am active in Facebook, LinkedIn, Twitter and in other places. And if you want to join to our coach topical authority or you can join to our community as well. And if you want to see me in person, you can come to the Kushasa Turkey during the September always. We have a mastermind for seven days and probably we'll be focusing on a launch of a new course and a new sus in 2027.

We are actually planning for mainly semantics. We are starting the design and also coding it as well. But I won't I'm not going to be announced officially until I I trust the sauce as well. And that's it. Awesome. Uh, I got to make my way to Turkey to hang out at your event. So, I'm gonna put it on the calendar and if another event has to move, uh, we'll see what we can do.

can do. can do. >> Thank you so much. >> Thank you. Thank you as well. Bye. Bye. >> Damn, that was an intense episode. Um, I told Karai that, you know, this would probably be a 20 to 30 minute chat and we went close to an hour. Uh we chatted a little bit before and after the podcast recording and yeah we just talked about a lot of things in the SEO space.

I think uh Karai is possibly one of the more misunderstood SEOs in the world. Um misunderstood because I think he attacks uh SEO from a different angle and I think he's actually been doing it in an an advanced way for many years. I think it's actually now because we've got uh the world of AI and people are starting to talk about LLMs and rag and you know vectorizing databases that all the stuff Karai was working on years ago.

Um it's almost like it came true. Um so I I got a lot of really good ideas out of this podcast. I hope you did too. Um I'm personally planning to listen to this a couple more times. So the next flight I go on I'm going to listen to this episode and I'm going to take my pen and paper and just write some notes. So very grateful to have a this chat with Karai.

Um still buzzing from doing the recording and he was one of the more uh downloaded episodes we've had in the last couple years and I think this one is probably going to beat it and it's going to be just new content that's relevant for today. And for those that have kind of uh maybe thought that you know Cry was just talking too much about semantic SEO. Well, semantic SEO is a real thing, but it goes deeper beyond just kind of, you know, doing eat and uh

topical authority. It's you can take it to another level. And I think in the world we live in today with AI, there's definitely a lot of opportunities here.