From People Also Ask to AI Search — Reading the Signals That Matter | Mark Williams-Cook

Hi, I'm Gianluca Fiorelli and welcome back to the Sales Session. Today we are going to talk with a very well-known person in our SEO search industry. Probably you know him for alsoask.com, probably one of the most used tools by all of us. But our guest is also, let's say, it's also the digital marketing director of Kandoo, a digital marketing agency in the UK, in Norwich. And he's also very well known for being quite vocal on LinkedIn, especially with really, really cool, let's call them so, zero-click marketing stuff. Which are the Google Daily Unsolicited SEO Tips. And he also has a wonderful newsletter that I really recommend you to subscribe to, which is Core Updates SEO. And he loves video games, as I discovered in listening to him at the last Brighton SEO. And this person is Mark Williams-Cook. Hi, Mark. How are you doing?

Hello. Very well. Thank you for such a kind, generous introduction and pleasure to have time to sit down and have hopefully quite a fun chat with you about SEO.

I'm sure, I'm sure it will be fun and not only fun. I think that, uh, from our conversation and especially thanks to you, starting from myself, I will start, we will all start learning something new. Or to have maybe a different perspective to something that we think to know. As I always ask to my guests, SEO guests, let's do this classic question of mine. How is SEO treating you lately?

Pretty good, I think, to be honest. So there is obviously a lot going on. The agency, as you said, Kanda is still very busy, which is great. Also Asked, which you also kindly mentioned, I think is probably slightly busier than usual in that we're getting more new customers. And I think that's partly to do with the discussions that are happening around AI search and kind of the other half of SEO has woken up about things not just being about keywords, especially with like all the query fanout stuff happening as well. And I guess for me personally, 2 things, which is one, I do honestly feel like my head is very full at the moment and my subconscious is kind of working through all of this new information, trying to figure out what the best next steps are for medium, long-term future, which is Firstly, I'd say really exciting. I don't think that happens that much in SEO where we have such a big change. And I guess closer to home in the trenches, I, we're making at the moment at Kander a series of short videos that we're sending just to our clients to talk to them about AI, AI search, geo, AIO, whatever you want to call it, because I think a lot of our clients are being actively approached now by People kind of saying to them, hey, you know, SEO is old hat. You need to be doing geo or AI search now. So just communicating with them a bit more proactively about the research we're doing and showing them it's not a whole new thing, at least not in my opinion. So that's something we've had to start doing that I haven't had to do before.

Your idea of shooting video, it's very smart because It's something that you can also recycle for other things, but it's also useful because so you shoot a video and it goes to all the clients you have. So you don't have to repeat always the same thing and waste energy. And yes, I think it's needed because maybe a video is impacting more. I totally agree with you. I don't think that the AI bros, as I call them, are— I mean, yes, AI is entering search. AI is its own characteristic, but like every other phrase that is used by search, because we didn't invent something like TTEO for saying TikTok Engine Optimization. We just call it SEO for TikTok. AI search is so easy because it's combining really human Easy to understand. And yes, we know that at the end of the day, many of the things that these AI bros are suggesting, I mean, is SEO 101.

Yeah. And firstly, I want to give some credit to Will Reynolds from SEER Interactive. So he's a little bit of a low-key kind of hero of mine. I really look up to him and I saw he had posted kind of a very candid video, just like selfie video of himself talking about AI search. And I was like, this is really good. I should probably be doing this. And that's what kind of kicked me to start doing that with the clients. We had as well, I think it was yesterday or day before this week at the Search Central Live. I mean, Gary Ish saying from Google's point of view, at least for Google in terms of AI mode, AIOs, search, it's basically just good SEO practice. Obviously, that's confined to Google. But there are things that we are changing in the agency, you know, it's not like nothing has changed. And what I mean by that is, obviously, the way people interact with search services is different in terms of they're doing these, like multi-dimensional searches. With a traditional search engine, you can pretty much only search well for like one thing, whereas with AI search, you, you have this more conversational nature. And I think it's pretty obvious and it started to come out in data that people are saying, oh, this is surprising. The traffic we're getting, say from ChatGPT converts higher than traditional search, which again, isn't necessarily surprising because a lot of that funnel is happening inside, inside that. So there's certainly different things we need to be thinking about in terms of How people interact with the information that we put out there, how we measure it, how we track visibility, and maybe tweaking even the kind of the weightings on the things we are doing for SEO. But like you said, all the actual core activities that we are doing, and I can foresee us doing, are still the same things that we've been doing for SEO, you know, about Whether it's making sure the site is technically clean, whether it's about content, whether it's about getting coverage, they're the same things, but maybe weighted slightly differently.

Yes, totally, totally. And a friend of mine once said, what has changed in terms of work is obviously what you said, how the search behavior of ourselves, because sometimes we talk about people and searches as if we We aren't people and searchers. So we also, I noticed on myself that I changed quite a lot the way I search online. But he was saying to me something that I've always considered interesting as a perspective. What has changed is the time of a session. Once search was, especially before with zero click, SERP phenomenon. So before that, Google was already trying to gate people inside search. The search was, I want to know something because you did the search before, let's say the clarifying part of the search before in your head or using offline media, for instance. And then you were going directly and searching for a specific question. And now it's people searching inside. And yeah, yes, what you said about ChatGPT, better increased value of a click, considering that the clicks are very small on ChatGPT, with respect to Google, still, or I think it may always be so. But yes, it's because people have done a long conversation before the click. And if they are lucky that the link they click is not hallucinated, which is a new task that I was never thinking to add to the task that I gave to my clients, to doing hallucinated broken links redirection, because at least 1/5 of the links to my clients in ChatGPT, I don't know yours, are going to a 404. So, but yes, I totally agree with what you said. And, and I think that maybe you already had this sort of perception of this longer session, totally because you created a tool like also us. So you probably saw that these expanded queries, these new potential queries and new potential path in the search session, we're leading to this direction. How do you think it's going to be the future for a tool like yours, but especially for a feature like People Also Ask in a context where Google seems that wants to bring more and more AI inside classic SERP?

So I've been using people also ask data for over 10 years now. And I remember quite clearly the first time I showed what we were doing with that data on stage was 2016, which was a conference local to me called Talking Tech. And I was talking about SEO. And at the time we had like a command line tool version of Also Asked that did some extraction and and such. And I was quite surprised that after doing that talk, I had quite a few people come up to me being like, oh, this is a really interesting tool. Can you share it with us? So I was like, yep, fine. And I shared it. It was originally in Python, this script we were running. And then I started getting lots of kind of support requests of people. There's a barrier to entry there, right? Of installing the right version of Python and making sure you've got the right modules installed and stuff. So that was when I was like, okay, there's definitely space to make this as a, like, online tool. And the reason I was using the data was, I think it was 2014 or 2015, I saw an SMX West talk with Paul Haar on it. And he was talking about how Google internally measures the success of their search results, which I was very interested in, right? Because I was like, if we, if I can understand more about how Google judges its success, I can maybe predict which kind of direction they're going, what kind of things they're looking for. And one thing he spoke about was TTR, so time to result, which was, I believe their definition was roughly, they're trying to measure the amount of time between a searcher having some kind of intent that they want fulfilled and how quickly Google can achieve that for them. So my original logic was the people also ask questions that are surfaced are the closest, what I'd call intent proximity. So it's like if someone asks this question, this is most likely going to be one of the next questions they ask. And it's actually quite detailed in terms of you have your first 4 questions and when you click on those questions, it opens up like a concertina of more questions, right? But the questions you get different if you, instead of clicking on an answer question, if you just re-Google that question, and that's because it represents a different starting point in terms of the searcher's knowledge, which I found really interesting as well. And there's actually quite a few patents about this in terms of Google and trying to work out kind of the skill of the searcher in terms of how technical the information should be or how in-depth it should be based on their previous searches and the types of searches they're using and the behavior of when specific words are used. So there's a lot of nuance in that data. And to me, that was the very, that was like the manual version of AI search in terms of, okay, someone's researching this topic. So they are probably going to want to know these things to complete that journey. And that's what's happening now, I think, with, in some ways with AI search and this query fan out, which is, okay, someone's asked me this. So Google automatically in the background then is saying, well, these are the other relevant bits of information that I need to find. And I'm going to tie it all together for them to give them this better experience, which again, I imagine if they measure it on a time-to-result basis, is going to be better for the user, right? And that's definitely a trend I've seen in all technology for many years, which is if you reduce friction, you give people this, you know, a comparable outcome for less effort, that will be the thing that wins out in the long term. And in terms of your question, how do I think it will last? That's a super interesting question to me because I have in my head this kind of 2-stage thing, which is at the moment we have LLMs, whether we're talking about Google AI Mode or ChatGPT, right? And we know, we absolutely know Google does. And I've been told from people I trust very much that companies like ChatGPT are buying up link graph data to help to help their initial kind of pre-training and weighting what data they're listening to. Google obviously has 20-odd years of experience of working out who they should listen to about specific topics. Now, all of this works on the basis that there is a link graph, firstly, which is an expression of trust and popularity on the web. People like this, they link to it. And from Google's point of view, certainly we know now without a shadow of a doubt, a lot of this relies on their click data on their SERP. However, you know, without going into specifics, however you want to kind of think about that, a lot of this relies on looking at how people interact with their results and how satisfied they seem. So what happens if lots of people do move to AI search, so they're doing I don't know, 80% of their searches within one of these models. So they're not visiting websites. So they're not clicking on traditional search results, which have a lot of parity. They're nowhere near as personalized as AI or AI answers. And they're not linking to websites anymore. So all of the traditional metrics, which we know, you can argue the toss about this is spammy, this isn't, but they generally work pretty well at scale, those signals. They're essentially going to decay over time and become less useful. And we're left with, I guess, from a model point of view, instead of a link graph, we've got like a language graph of how codified language fits together. And the next obvious move I've seen is obviously some of these LLM companies are talking about releasing their own browsers, which I can only imagine is because they want some kind of way To measure the interaction data. So at this stage, I think the way in which the kind of questions that People Also Ask data surfaces will probably change because that data is generated from basic traditional searches. My kind of last piece of that puzzle is I think that's still quite far away because I don't think anyone has a working solution for how You get an LLM at scale to give reasonably decent answers without all of the metrics and link graph information of the rest of the web. Yeah, I've got more to say on that, but I don't want to keep on that. That's where I think we're going with that. So I think the data is going to be useful for, I think, quite a few years to come. And as I've mentioned before, I think you highlighted it in one of your emails to me. The interesting thing as well is like, people aren't just using LLMs as search engines. They're using them like mini tools to do like sums for them.

Yes.

Like one of the top uses we saw from the Harvard Business Review for ChatGPT was, agree with it or not, was people like getting therapy from it and talking to it, which is wild to me. So this is a whole different— there's not a question, you know, if you're doing SEO for someone that provides therapy, it's no longer a question about how many clicks did we get? Or how can we drive traffic? Or what content should we do? It's like, you know, that's replaced you for good. I won't argue whether that's a good idea or bad idea, but factually, for some people maybe that would've had therapy, they're just doing it on ChatGPT now. So the whole notion of a website just goes out the window. And I think there's gonna be a few uses for AI that just snatch up entire slices of an industry that's a bigger wider issue for me than did we get a click, did we not get a click. Well, yes.

I mean, people using Google Doctor were existing also before. And usually when I think about people using ChatGPT as a therapist, I think that my concern, how many hypochondriacs, hypochondriac people are going to surge in a world which is already affected by a lot of mental health Problem. But yes, when I was thinking in, um, people also ask what could be something similar. Don't you think that maybe the follow-up question could be something similar to, we already see them in AI mode and they were experimented in SGE last year. Don't you think that that could be the way Google can integrate the concept? Let's say the concept of People Also Ask inside the AI surface.

Yeah. So I think one of the things I learned from the DOJ documents was that features like Google's People Also Ask function as a feedback and validation loop for their search results, meaning if you have a search a broad kind of search where the intent isn't necessarily clear, whether someone's searching for like GM, do they mean General Motors or do they mean genetically modified? You haven't got any more information. One of the things that people also ask boxes do is help Google understand where the majority of that query intent lies. So if everyone's clicking on things about General Motors, then they can start adapting the system automatically being like, okay, This is actually more what this query is about. In terms of the follow-up questions in AI MyDF, for sure, that could absolutely be a way they adapt this as well. I think they'll have less issues with identifying intent because of the conversational nature of those systems. Because we're still, when we do traditional searches, We are still taking a mental step ourselves in that we're translating the thing we want to find out into a search term because everyone goes through that process of people's Google fu, if you like, is different, right? Some people are really good at finding out specific answers and other people aren't so great. And that's through learning how to work with the search engine. That again kind of goes out the window with conversational search because it's our natural way for most people of communicating, right? And you can be a lot more specific though. That's what they allow you to do. Knowing the right questions to ask is, I think, a really interesting concept in terms of information foraging and finding things out. So, if I'm, for instance, moving house for the first time and I start doing searches around the solicitors and conveyancers that can help me, I don't know about that space. So actually it's helpful to me as a searcher to be like, this is what other people have asked. At the moment, I would think, and I have a test lined up for this, that the PAA data probably quite closely aligns with both follow-up kind of questions and query fan-out. So I've got someone doing an independent study on something else with PAA data, which has been super exciting. I don't want to share just yet, but that's certainly on my list of things to properly interrogate. Like, what's the overlap between these 2 data sources?

Yes. And in fact, this is also one thing that was popping up into my mind when you were talking before and talking about where query fan out, because, and I think this is a defect that all of, many of us, I include myself, sometimes we have. When we speak about things, we try to synthesize too much the things we want to say. And for instance, in the case of a query fanout, everybody's talking about the query fanout as such. But I see very few people saying, okay, but when we see the query fanout, for instance, using the Mike Ivor-Ragg tool of Queryphoria, you see that the query fanout is as a taxonomy. So an expansion query, a related query, a personalized query, and nobody talks about this. So when you were saying about, yes, we putting in relation, people also ask, we query fan out, maybe the follow-up question are including the concept of a very generic concept of knowledge base of Google. Are they pre-query fan out that they know? that people, especially in the related searches, that people are going to ask after. So maybe studying these kind of things, we can make a better use of query funnel, people also ask, follow-up question, because, and I think you agree with me, a lot of people have really used bad a tool like also ask and the people also ask. question, just as a means to discover some new topic to, to feed the content machine.

Yeah. So one thing I'll say as well, so I've got the domainqueryfan.com and I'm hoping next month to release a tool that's aiming to be essentially an open community database of AI, real AI prompts, so not synthetic ones, and the related query fanout that happens. So this is an experiment. It will require people's participation. I think I've worked out a fairly good way to do that at scale. But the idea is you'll have an online searchable database where you can search through queries for specific things. And then you can get an overview of the types of query fanouts that were generated from them, because I think that's one of the easy things we can do in terms of AI search to figure out what we need to do, i.e., if we know these prompts regularly trigger search actions and we know there is some consistency in what the LLM is searching for, then we know these are what we need to optimize these these pages for, to show up for, to appear in the, in the AI search. And you are absolutely right in terms of People Also Ask data. So I think a lot of people got snapped up during the helpful content update in terms of, I saw lots of sites doing programmatic SEO. I use the term very loosely with People Also Ask data in that they'd take any topic. And they'd scrape 50 people also ask questions and just in a QA format, like list the answers and generate tens of thousands of pages. And at the time it worked really, really, really well. It was actually what my 2022, I want to say Brighton SEO talk was about. And I was showing sites that went from zero to like a million visitors in 4 months. on brand new domains using this tactic. I think it's one of the imprints that's made its way into the model that was deployed with HCU in terms of if there's lots of pages with this kind of pattern, then it's probably a signal it's not a great site. So the way I've always used People Also Ask data is, you know, you can use it for the old FAQ. But the way I think about it is if someone has done this search, they have this question. These are the things ideally that you should include in the answer. And of course, you need a human in the loop with all of these things to give context and to understand that, but it's been eminently helpful with clients just showing them this is the majority of the questions that people have around this topic. And whenever you're inside a company, you develop your own biases, your own blinkers, your own terminology. And it's been quite surprising for some clients to see that. And it's actually triggered us in some cases to go on and get them to actually speak to users because they, in some cases, they didn't believe people were actually asking these things. And it turned out they were when they actually, shock horror, decided to speak to the people that they wanted to use their website.

Yes. Well, now I'm talking as a user, I will ask, and personally, I also, in the very first times that I was using it, before understanding how things were really working, using for classic FAQ, it was a classic output for when you were using People Also Ask. But then I understand one thing. People Also Ask can be considered as a sort of conversational expanded query. And then I started to see Google using expanded query everywhere. So what I started to do is, okay, let's use the information also that people also ask in combination with the people also search. And I consider people also ask as deepening into a topic and people also search widening the topic. And then topic filters, then when sometimes because they are using less the image search tags. Also, the image search tags are expanded query. So combining all this information for me started to be the true keyword research instead of going to whatever beloved tool. And we use a lot of tools, everybody. Instead of getting crazy and asking to a tool, to a database of a tool to do the keyword research for me, I prefer to do that kind of keyword research by myself. Initially quite manually and really painfully, but now more and more tools are combining all this stuff. I was able also to create something with the help of a friend. I'm not, I'm totally an imbecile developer. It's something that I cannot really make mine, even if I try. I mean, if I was vibing code, the world would collapse. Brilliant. And but using, because I think, I mean, it's Google already suggesting us where people are actually searching because they know what people search. So let's concentrate on that and create and see what all of this information is giving us about a mapping of a topic and subtopic and subtopic. That's why I use a lot, people also ask of that reason, in order to inform better the topical content cluster that I want to create for my clients. And now I'm combining the query for out mostly also for this, because for instance, just a case for a client of mine, it is fun because we were talking about moving to a new apartment. One of my clients is one of these that is offering service for moving contracts with electricity, internet, renting, all these kinds of things for people moving from an apartment to a new one. And he was saying, we are here, we have this content, it's perfectly done, we have a good profile, link profile, but yet we cannot win against our competitors. It's the classic super keyword that is Very, very difficult to find a solution. And then in this case, we found out, which is for me, maybe the best use of a query shoutout is saying, okay, let's see the question that Google offers us. The question, the query that Google expands this problem. And it was almost immediately visible that, okay, you're not ranking better than your competitors because you and your competitors substantially, if we quit the template of a website, we are seeing the same thing. So let's try to find a way to make your content being still meaningful for all these things, but saying something that people are asking and nobody is answering to. And in this case was classic. It was about, you know, customer care phone numbers and all the content were about these are the customer care for this, that for that, this is for Madrid, this is for Barcelona. And one of it, at least 2 or 3 queries found out were about the time I have to wait for someone to answer me or about the robocall, you know, when you have to say 1 or 2.

Press 1, yeah.

Yeah. So now we are going to work to find something like this. So because this can be difficult because obviously you have to talk also with the customers because you have to ask the customer, what is your experience? So to aggregate quite significant statistical volume of data in order to also give correctly this information. But we are going to do this. And let's say if, because sometimes there is these things about the consensus, which is ingrained also, if I remember well, in the algorithm. And I think that if the consensus is fine in some cases, But probably too much consensus is going to make you weaker. Also in terms of visibility and usefulness, because if someone, your brand is less known and you're saying the same thing of a brand which is better known, why people should trust or use your brand website instead of the one they know better? So try to be different. So I think these are the Many, many things that people also ask, query for now. All these features that we have on Google and offered by Google can be more useful than the classic do this, this, and that.

You mentioned, um, because I know you were really interested in them and I never actually learned the name of what that feature was precisely called, which was the little like bubbles at the top, the filters of the search. Are they still, are Are they still live? Because I don't see them anymore in the UK.

No, they're still live. I see, I think it's the classic feature that Google is sometimes quitting and sometimes putting. It's less present. It's the topic filter, but it's still present. I see, for instance, in Spain, in Italy, it is less in the US, less in the UK, but I still, I see them. And, um, another feature which is not so famous. like many others, was the Things to Know.

Yeah.

Things to Know was fantastic because when it was appearing, it was giving substantially the taxonomy that you should have to use for creating something, because it's based on taxonomy, the Things to Know. That is something that somehow disappeared, but with the latest June, July core update, I saw for a client, for instance, increasing a lot the visibility as if Google was resurfacing this feature again. So let's see, sometimes we know that features sometimes tend to be, to have a short life, but Google still has not decommissioned the page about topic filters.

Haven't reached that great conclusion yet.

On the official website. So let's see. When we are going to quit it and putting, or putting the feature is deprecated, is not going to be until—

I mean, outside of, outside of traffic, I think they're always super useful, those kinds of Google features, because they give you insight into like how Google is thinking about things, how Google's categorizing things as well. So I, I'm always all over those features from a data point of view. Before I look at how external tools like try and classify searches and try and classify intent.

Yes, totally. And in fact, I mean, what I do usually after having collected all of this information, then I go with tools like our friends of Keywords Insights in order to have a faster cluster. I can do also clustering let's say, not relying on tools, but sometimes I like to have the two, my work cluster and work of clustering and someone, the clustering backup by a tool specific for that. So I can see where also I can improve or where I saw some differences that the tool is not seeing or vice versa.

I really like that about your approach, what you said as well about using like You know, although we may also ask, we use a whole range of different keyword tools as well, because it's exactly like you said as well. Like Google suggests data, for instance, is useful in my opinion at like a different hierarchical level on a website because it's, it's more like a topical level. Whereas PAA data is, is more the in the weeds. This is what this page should deal with. We, we're a Keyword Insights customer as well. So yeah, I'm a big fan of anyone who's kind of, and like you said, doing things manually as well. I think there's a huge, huge value that. So yeah, I'm very much a fan of cross-comparing different data from tools and not just being like, yeah, this is the clustering tool. This is the intent or keyword tool.

Yeah. Yeah. Yes. And I mean, sometimes it can be quite expensive for a freelance. I know, right? But it's, it's fine. I mean, it's something that I think it's, it's needed to do. And I think that my manual approach is also something that I bring back from the early days of SEO when almost everything was manual. I mean, we were using Xenu for doing crawl. So I think—

I remember the Link Sleuth.

Yeah. And so I think that it's a way for me to learn by doing. And learn. It's like when I say don't look at the surface, but no, not just look superficially, just put your eyes on the surface, start clicking as if we're making, you know, a user experience test. So you can discover things that are there and nobody's paying attention. Like for instance, the same function of a search menu, it's something that we give for granted. And it was— I was surprised when I started to say that depending on the order of the voices that are presented in the search menu, you can understand what is the perceived search intent that Google has. Because obviously we read from left to right in the Western countries. So obviously Google is putting on the left, apart AI mode where it is investigating blocked on the left because they want to push it. But usually it's all and then the one that they think is closer to the search intent. It can be shopping, can be news, can be— sometimes I even saw web, classic web. People for this thing just want links.

Blow the dust off the web.

Yes. And talking about Another thing that is really scratching my head, and I want to talk with you about this to hear your opinion, is I'm more and more questioning why us as a SEO, as a marketer, why do we need to know the prompts that are used by people in AI search? I mean, I don't care because, and it happened to me many times, I can prompt the same thing. The LLMs, also AI or AirMod, when I tested AirMod, always give me a different answer because it depends on the moment I do the search. It depends if it's my son, for instance, I use it as a tester in his own PC, is doing the same conversational query using ChatGPT, but it only computer. So with a different story behind, a different conversational story or memory on ChatGPT. So why should we know and base all our data-driven marketing on understanding the visibility for the prompts? What do you think?

Yeah. So I understand why people want to do it because they want to try and understand their visibility. As you rightly point out, one of the challenges is the output a lot of the time from LLMs, if you're just interacting with them on the web, is not deterministic. Even if you could ask the same question from the same person at the same time, you may get a different answer. And I've seen tools that are measuring visibility do all kinds of weird things. Like they'll do the same prompt 100 times and then look up who the most commonly cited people were to try and figure out that bell curve of token probability. So I understand why people want to do it. I'm not sold yet on it. So we haven't got like a partner or a specific tool that we use to do this yet because I'm not sure whether it's a good long-term bet. In terms of, I was having this discussion with a SaaS provider this week, which was they now have a big database of prompts, millions of prompts they use in ChatGPT, and they can tell you if you are cited or not. And I was asking them where they get the prompts from. Are they real? Are they synthetic? And they get them from their search database of queries that people are just doing in Google, right? I said to them, obviously people aren't just doing the same queries in something like ChatGPT as they are in Google. Are you telling me that the results though are very similar? And their answer was an unqualified yes, that it doesn't matter if you wrap it in a more conversational way, you will get similar citations if you just put in like a basic kind of search term. So I'm, firstly, I'm waiting to see the data on that because I'm aware it's in their interest for that to be true. So I kind of want to see the data on that. I think that being said, there is, that's one, again, piece of research that I mentioned earlier that I'm interested in, which is that for things especially that trigger retrieval augmented generation, How different are the queries that go back to the search engine is one thing I don't know yet, but I would like to know. So I think as an exercise and as an industry, it's worth trying to understand these things because we need to figure out if they are useful. And the other question nobody can answer, I don't believe at the moment, is how much personalization does actually happen? I did an experiment live today on the, on the Office Hours SEO, right? Which was, they were talking about ranking in ChatGPT and I had a little conversation with ChatGPT and mentioned that I was vegan just as part of the conversation. I'm not, but I mentioned I was. And then later I said to ChatGPT, oh, hey, I want to start running. What are some good running shoes? And it immediately said to me, oh, Brooks are a great brand because they don't use animal products in your shoes, in their shoes. And I was like, that's interesting to me that obviously it's got that memory context and it's strung together that that would affect my purchase preference. So the searches or the brands it was at least surfacing, and that was from inside the model, it's done that. And a friend of mine, Myriam Yesier, has done some really good work and research around this, that if you Google brand quadrants, like Miriam Yesier, you'll find it. And it's about what people understand about your brand, how that is echoed within like AI search, and how you can improve that and make sure those models understand you and your brand and the details about it. Because it doesn't worry me as well as a concept, because The same amount of people, taking my trainers example, I don't know why I always use trainers, but the same amount of people are going to be looking on the internet to buy trainers before or after AI, and the same amount of money is going to be spent. And yes, I may get fewer clicks on my website that specializes in vegan running shoes, but it's because I'm not being sent all the people that wouldn't buy them now. So I'm open to people doing the research. I'm not integrating it into like workflows and I'm not making decisions on it yet because I think there's too many unknowns to be able to have this confident, maybe like you said earlier, AI bro approach of you need to know the prompts so you can do this. Because when you start, when I've had conversations with these people and started picking it apart, they don't know either. So I'm not prepared to go to clients and kind of say, you need to put resource and stake things on this yet.

I totally understand you because I am somehow in the same situation with mine. So I want to go quite quickly, but meaningfully to the big discovery that you did during this year about the Google exploit discovery. I want to ask you just one question. What is the biggest thing you learned, the most relevant thing you learned from this discovery? information that you discover. The one that you made, okay, this is clear. If I have to choose one, this is the one that really makes me understand something that I was not understanding well.

For me, it changed how I view SEO a little bit in terms of it became very clear looking through that data that there are like qualifying rounds you need to go through in Google to even be eligible to rank for something. So I think the most important thing that we discovered in that exploit was the site quality score. I actually spoke this year to an ex-Googler who had worked on that. And he said he was shocked when he saw that kind of been leaked out because it was, he said, really top secret stuff, but the way at which that score was very brutally being like, if you score below this, you are not appearing here, here, here. And, you know, in these features or whatever. The type of things where featured snippets were generated to me from the data looked like once you had made that qualification run of sites, the things that were being done were fairly standard, basic, nothing mega smart information retrieval, just for kind of matching to the query. So what I took away from that is I've experienced lots of SEOs being like, oh, you know, we haven't, say, got this ranking or this featured snippet. And they immediately focus on the content. I learned from that, that your content is especially now absolutely delivered in context to the vessel, to your website and you are measured on a bunch of stuff on that website, on that vessel. And it doesn't matter how good your content is and how many experts you have. If you are not scored correctly, you are not even in the race to rank. So that's something I look at now with clients when they come to us and they say, we've lost ranking, we've done this, or we're not ranking for this. And we've tried so hard on the content. It's a lot of the time, it's not the content.

Don't make it somehow an Assassin's Creed jump of faith. Cannot we somehow find— I know they are very, very distant between one each other, but all these signals that you say that it's not really just about your content, but all the environments wrapping the content. So substantially the website and the signs that the website is bringing is not substantially the engineer way of explaining to a machine what is E-A-T, because E-A-T, I always thought that it's not a checklist. E-A-T is a good website usually has this sign. So that's why I always talk about semiotics, because a sign like newspaper always have, usually have the name of the author, a good e-commerce Always present this kind of information for about payments, about whatever. So cannot be these things somehow the two faces of a same coin. AAT, the concept explained to, let's say, business owners, because usually these kind of things are not said to the developer google.com web subdomain, but usually in the google.com, the keyword official website. And what you have discovered are substantially, because you cannot transform, translate as such these signals, these concepts, you have to find association, find things that can be representative of these concepts for the machine. Can it be something similar? What is happening? Well, I'm just brainstorming.

Yeah. I mean, the So take Site Quality Score as an example, right?

Yeah.

There was a fairly clear patent called Site Quality Score, and there was a description on how it was calculated and estimated because it was calculated primarily with user click data. And if they had no user click data, they would build what they called a phrase model of the content. And compare that to phrase models of known scores and then make an estimation based off that, which is, I think, why actually they had such a problem when AI text generation became public, because by its nature, AI text generation versus old school text generation of, you know, Markov chains and Mad Lib article spinning didn't statistically look like good content from a phrase model point of view, whereas AI content did. The patent made it clear that how Site Quality Score was scored, that was also reflected in what we saw because we had the actual scores. So for instance, the score for the www.amazon.com was actually like a fairly average score. It was in the kind of bell curve of this is just an ordinary website. Amazon's subdomains like Flex, And their logistics subdomain, which are very specific, scored higher. And some of the top 0.1% of Site Quality Score sites were things like the FAQ section on the library subdomain of a university. And that's because the way those sites were only really found through search is people were specifically searching for them. that was a high percentage of their kind of search traffic. And I think that's overlooked in terms of if you're ranking in Google and for whatever reason you lose some visibility, if Google sees that nobody is actually actively searching for you. So if you don't appear in search, you just don't exist, then your site is dead in the water. Which is why people talk about, I think, brand now and brand search, right? Brand search has, or brand has become a much bigger topic, but I think that's a hugely important signal now. And I used to have this discussion again, like 10 years ago, because this used to be a thing that seemed to happen in Google, which is with a new website, you would rank like 40th, whatever. And then Google would kind of bounce you to rank really nicely for maybe like 2 weeks.

Like today still.

Yeah. And then it would bounce you out. And then I think based on that behavior, I don't see it work quite like that anymore. But that certainly to me feels like a basic like site quality score of, are we getting the click-through rate we expected? Are we getting the, you know, SERP interaction signals we expected? If we remove it from search, do people try and find it anymore? Or do they just not care? So Those things, how do you, but in terms of E-A-T, how do you tell people to do that from a Google point of view? And that's where I think E-A-T comes in. And I agree, and Google's told us it's not like a checklist. It's not like an E-A-T score. It's a combination of a whole myriad of things. And it's a mindset where you are building something that the algorithm is aiming for, you are not aiming at the algorithm, which is again, I think an easy trap to fall into when you start SEO, which is I want to understand how the algorithm works. And then I'm going to base my strategy to aim it at the algorithm before you appreciate that the algorithm is constantly moving. So you're aiming at a target that's moving. You need to aim ahead because you can get really good results doing that. Right. If you work out, okay, this works particularly well at the moment. So everyone does that thing, but then you're just setting yourself up because inevitably, if it works too well, at some point it will stop working. And then everyone's kind of like, shrug. I wish we'd actually, like you said, been working on the quality, on building trust, on building brand, all those things that Google can't take away from you. I think a great question to ask outside of things like technical SEO, where you are doing things for bots. Is if search engines didn't exist, would I still be doing this thing? And it's a great way to know you're building equity. So you're making your site faster, for instance.

Yeah.

Like, I don't care about Core Web Vitals or that it's a ranking factor. Even if Google didn't exist, I would still want my site to be faster. Therefore, it's a good thing to be working on.

Yes. And returning to the click and as a signal. So to user signals. Not so many remember, it was more than one year ago when the SGE patents came out and also it is present in the AI mode, AI overview and all these kinds of things, documentation that user signals are extremely important also for AI search for Google. Because I remember very clearly the one about SGE, but I think that many things are very patterns when were translated and migrated to the AI overview one. And it was saying that it was not only talking about clicks. We know that clicks are not so big, but it was telling about how far in the answer person were scrolling down and really looking at. So it's Chrome, Chrome 100%. And then if people were stopping on a piece of the answer, if people were going with the mouse, the pointer of the mouse, where is the part of the answer that we're looking at, which is something that we do instinctively, unconsciously. So all this to understand, maybe not for the, in terms of this is the website by Master Ranga, about this is the part of the answer that matters the most or not. So we can recalculate the answer and maybe there is a consequence obviously for the website from where the chunk has been collected. So user signals surely are the most important thing and that's why I think going really high in terms of politics of the internet, That's why the biggest pointed question is Google will be able to own still Chrome or not in the future because—

Oh yeah.

The browser wars is going to be the most relevant wars of this internet moment. So 1 hour, 1 hour, it's a time runs fast. So let's stop talking about SEO. Let's stop. Let's start talking about it.

Okay.

Just a quick question in order for people to know you more. Let's talk to you about your love for the dogs.

The dogs.

It seems to me that you are more a dog person from the things that I've read, from the things that I've seen, also from your side business, which is—

Yeah, it's a hard, I'll be honest, it's a hard thing for me to talk about right now. So I lost my dog of 10 years.

I can understand that.

A week before last. So that's been super tough. I'm dog sitting at the moment, a little girl who's down here called Pippi, which is my business partner's dog. He's just gone to Australia for a wedding. So I've got her for a month, which is great. But yeah, so we had Snoop, who we rescued from like a dog shelter when he was 5, and I had him for 10 years. He had been given back, I think it was 2 or 3 times to that shelter. Because he needed some work, but we gave him the work. He's kind of made his way into all kinds of Canva branding around the office because one of his kind of quirks was he couldn't really be left alone. So I essentially had 10 years of this guy at home with me, coming to work with me, sleeping on my bed, sometimes coming to the gym with me. So yeah, that's been a big, big shift for me now he's not here.

Yes, I totally understand you. It happened to me a few years ago with— I'm a cat guy more than a dog guy, and I lived the same situation. It was something that maybe who don't have a pet cannot even understand. But let's talk about some, again, return on technological stuff, but in this case, video games. What is the video game, the first video game you played?

Oh, the first video game I played would be on the Spectrum ZX, which our family had. And it was probably something like Manic Miner, which probably means nothing to, to most people, but it was like a super basic platformer game. You put a cassette in, it takes like 20 minutes to load the game. And in those days, there's no such thing as like a save game. You just play, you get like your 3 lives and it doesn't matter if you're an hour into the game, when you're dead, you're done and you have to start again. So there was, I think, definitely some lessons that early video games taught me about, you know, patience and perseverance and grit and trying again.

Yeah, I remember that period. I also had that ZX Spectrum. Spectrum. Yes, it's 32K memory. Yes, it was a joke, right? And, and now talking about, yeah, you're like, uh, let's say, um, kind of role-playing game, video games or something like this.

Yeah.

If you have to choose, uh, let's say, because video gaming, what is fantastic is because It's of the worldbuilding, especially. So let's say, let's imagine there is the world of Prince of Persia or world of Warhammer 40K that you know that I like a lot, or there is, let's say, I don't know, the world of FIFA soccer or the world of Cyberpunk 2077. I don't remember exactly the name. What is the kind of universe that Let's say doing some sort of flipping from reality to virtuality, virtual world, you would like to experiment at least once.

So, I mean, about, so about video games, right? My, I like video games for, I think, weird reasons. Like, so I most commonly play like simulation type games, super nerdy ones like RimWorld and Dwarf Fortress, which are basically these Big simulators where you control small things, but they're very, they're very large complex systems. And this is why I kind of integrated it into my SEO talk in that it reminds me of kind of the work that I do around search engines and Google in that there's a big complex system that I have some understanding of. And it's just always satisfying to me to work out how to kind of try and influence those things and there's like an optimization outcome for me. And yeah, I made that whole kind of joke around, I don't get why people play games like Connect Four because they're solved games. And I know if you start in the middle column, it's like 42 or something moves until you force a victory, you know, on a basic version, same like noughts and crosses, they're solved games. So I really like these complex, like, system games, or if I was playing a board game with you, it would unfortunately be one of those board games where we have to spend like an hour reading the manual and then the game takes like 5 hours to play.

Oh yeah. Yeah. Yes. Well, uh, you have to know that when I was much younger, I was used to, to play war games with my uncle and his friends. And they, they were, you know, the classic, the war game rebuilding the Battle of Waterloo.

Yeah, yeah.

Always trying to make Napoleon win for once, at least fictionally. And yes, I love also the simulation games and something that I passed to one of my sons who can literally spend— he doesn't play things like, you know, the classic, or like sometimes, yes, for friends, to play with friends, but he's spending a lot of time in something like, you know, Art of Iron 4, for instance. So a really deep simulation of the Second World War with a huge community, with a huge amount of mods. So it's this kind of games that make you think are really cool. And I totally hear you that there are also a way for us to escape, but yes, still training our brain is something that happened to me when I paint. Because painting is not just slopping colors on a figure, at least if you don't want to be the Pollock of mini painting. But it's really, you need a methodology, you need to understand it, you need to know what is volumetric, what is, how to put light, et cetera, et cetera. Okay. I think that was Maybe time to let you go and return to your work or to whatever you want to do. Because today when we are recording is Friday, so it's, I mean, we are already ready to go.

Yes, it's quarter to 5 here on Friday, so I might just make my exit after this.

Fantastic. I wish you a wonderful weekend and it was a pleasure. Maybe Let's promise ourselves in the next future, 6 months, 1 year, to discuss and have a new conversation and see what has, what did SEO did to us in 6, 12 months.

Let's do it.

Thank you, Mark.

Let's do it. No problem. Thank you so much. I really enjoyed that.

And thank you to all of you, dear watchers and listeners. Let me do the classic thing. Remember to subscribe, ring on the bell, and give it a like to this video because this video with Mark really deserves a like. Thank you and see you in the next episode.

Thank you.