Improve your SEO with video games and exploits - Mark Williams-Cook - brightonSEO April 2025

Did you know that Google personalises many searches by city? And AlsoAsked is the only tool to access live People Also Ask data at city level. Try for free at alsoasked.com.

AlsoAsked.

Tired of SEO guesswork? Quattr's AI agent goes deeper to uncover high-impact optimizations and completes hours of SEO in minutes. Get started now.

I was going to say that SEO with Fat Joe is as peaceful as this, but then, uh, the birds are all having fights. Hello everyone, really glad to be back here talking about some of my favorite things: video games, SEO, and breaking stuff. Are there many people that play games here regularly? This is a safe space for you. Quite a few, excellent. So, um, this is what we're going to explore today, which is I think there's a really interesting paradigm between how to get good at video games and how to be good at SEO. We're also going to look at how people are currently successfully exploiting Google, and some literal Google exploit stuff as well. This is the first game I played on the internet against strangers.

It's called Red Alert.

It's a real-time strategy game. I think I was about 11 or 12 when we finally got the internet and I could go on and play. And I was like, I've played this game, I've read the manual, it's a complex game, but I'm going to go out there and show people how good I am. And I think I lost the first 10 games in a row and cried to my mum. Um, and people were beating me in like really simple ways. And I think there's an interesting paradigm to how people start SEO. You learn about kind of internal linking or schema and all these things, and you feel prepared. You go in, do all the things you think you're meant to do, And then somehow you're still getting absolutely destroyed by other competitors. And I want to talk about systems. So in any system, whether it's a video game or SEO, you have you, you have your competitors, and you are competing in an environment which has fixed rules and parameters. Now, interestingly, I think, is that in competitive video games and when we're doing SEO, there is a third party present. And this third party has their own goals, their own agenda, and more importantly, they have the ability to change the environment at any time they like to their pleasure or your displeasure. So if you haven't worked this out, this is you, this is your competitors, and this is Google. So I want to talk a little bit about understanding system goals. And I was talking actually last night to someone about, uh, Call of Duty, famous online shooty game. And Call of Duty has a system where they try and optimize player fun, right? So they want you to keep playing. If you do things that are counter to this, if you cheat, if you go on voice chat and say rude things about my mum, they have systems that will try and remove you from the game because you're reducing player fun. They do, however, scalp some of this fun in the form of they want to sell you DLC, microtransactions, loot boxes, that kind kind of thing. So your actual kind of experience sits somewhere in the middle. You're having just enough fun to keep you going, but they are hitting their kind of other goals they want to achieve. Same is of course true of Google. Google wants to make a good SERP. That is one of their goals. You are told to produce helpful content. Now I find the interesting thing here is the helpful content is table stakes. This is what you are doing to help Google achieve this goal. Many of you may have experienced the situation where you have produced helpful, good, unique, quality— whatever you like to call it— content, and you do not rank. And this is because, of course, Google has other things they wish to achieve, such as making ungodly amounts of money. Um, we heard allegedly 2 weeks ago an ex-Google exec called sending traffic to publishers a necessary evil, which shows maybe you are not always aligned in your goals. So our ranking reality again falls somewhere in the middle. And by ranking reality, I mean all the things you need to have in place to actually get your website ranking. So the question is then, why is SEO not easy if, if this is true? Why is it this kind of hellish, uh, profession that we torture ourselves with? And it's because Google is doing a really complex job, right? Everything's changing around them— the technology, people's expectations, how we do SEO— and they are constantly course-correcting with Google updates to try and get their kind of target. This is what we consider to be the results that we want. As SEOs, I feel it's our job to know when we are at one of these inflection points. There is usually parts of the algorithm that it's optimal to lean into or just straight out exploit. And this is true in gaming as well. So the, the term in gaming is called the meta, which is basically when there's a really optimal way to play the game. Um, uh, it might be like a strategy or some equipment you use. And generally what will happen is people will gravitate towards playing this way because they want to win. But then what happens when everyone's playing the game the same way? It becomes less fun. So they have to actually change the rules to make the game fun again. And the exact same thing is true of Google, right? So if we all suddenly got super good at ranking our own websites, we're all content goblins, right? So we're going to trash the SERPs for our clients, and Google doesn't want this. So if you want to have a fun time and not necessarily a long time in the SERPs, you can abuse the meta. I was on this stage 3 years ago. Giving examples of sites that went from like 0 to a million visitors in 16 weeks because they were exploiting the meta at the time, getting pretty much brand new domains to rank. We had a whole second renaissance in niche blogging. And then of course we had the helpful content update, which you can see running there, which essentially reversed things. So the meta at the moment, if you talk to a lot of SEOs, is around parasite hosting. So it's around you have to have your content on these bigger better-known sites if you want it to rank. And we've all seen examples of that. What really interests me is when these dials get turned, that they start overriding things that maybe we have thought previously have been more important. So an interesting one I like to think about is the concept of topical authority. So that's this idea that, well, if you want to be— if you want to rank for this topic, you need to produce lots of content around it so you become a topical authority, which makes sense. But then you're like, how is the Vatican website a topical authority on best CBD gummies? So, if you missed this, uh, the site got hacked and someone put this page on the site, and for whatever reason, because the site's very trusted, it was allowed to rank. And actually, there's lots of examples of this currently. Uh, so Martin McGarry works in iGaming space, showed me this example. I'm talking about it because it has since been outed and dinged, but it's ranking for best casino sites and it's Abercorn School is the website. And when I looked into it, it actually in 2 weeks went from 0 to 40,000 visitors with one page for casino terms. And by casino terms, I mean thousands of casino terms. It was ranking in the top 3 for thousands. How did they do this? Well, they bought this expired domain, this old school which had loads of great links, gov links, all the kind of stuff, you know, link builders tell you is great, and they stuck this big affiliate page on the homepage. But surely Google has a system where you can't just buy old domains and put anything you want on them. And I think how that's currently working is that functions on the basis that if you change all the content to something different, But what this site did was just kind of snip the homepage off and leave all the other school pages, the 600 pages, in the index as a school. So when you look at the whole site from a Googlebot point of view, it's still 99.9% a school website. It just so happens to have a gambling casino page, which is fine, I guess. Um, so this is just an example of how I think these different tunings have opened up these kind of things that people are taking advantage of. But for those that want to stay within the guidelines, honestly, I say this— the content is still not as important as you think. Because we have clients come to us sometimes where they've had traffic loss and they're like, you know, what do we need to do? We've been slaving over this content. But you look at it and I genuinely think from every way we can measure their content, the engagement, the expertise, it seems fine. We had a nice segue with Dungeons and Dragons. We had a Dungeons and Dragons website we built at the beginning of 2023 as like a little side hustle, uh, fun project for me. I think the content was good on the basis we had random people emailing us just thanking us because the content was so good, because we did put a lot of effort into it. Google seemed to agree. It got to about 20,000 visitors with not a huge amount of stuff on it, um, quite quickly. And then we had the helpful content update, obviously popped by, 99.9% of traffic absolutely decimated. So is it because the content is bad? Well, we can test it, can't we? So we copy and paste our content that doesn't rank onto LinkedIn Pulse, and 72 hours later, we are number 1 in AI overviews, and we are ranking number 1 organically. I think it's still in the top 3. So it's hard to balance these things of how can we say it's the content when we put the content somewhere else, it's fine. So all your content is delivered in context to the site, to the vessel that it's on, and I think some clients spend too long looking at the actual content rather than the important things that are measured around that. So how do we get better at SEO if, you know, reading the manual isn't working? Much like gaming, we have brilliant communities in SEO that kind of summarise what's working, different people's experiences. One word of warning with any community is occasionally you get dickheads that claim they're doing things that they're not, and you can protect yourself from this, and you can actually verify how things are working by trying to understand understand the underlying mechanisms. And you can do this with inside information. People mine games for data to find out more about how they work, and you are actually spoiled for choice now. You've got stuff like the DOJ documents, you've got patents you can read, there's a whole load of leaks we've had over the last 10 years. Googlers themselves, in my opinion, are eminently helpful. Googlers are not Google. Google, uh, here's a patent called that talks about the initially inverse response transition, which is when you do something where Google's like, yeah, I'm gonna rank you better, sometimes it's just gonna rank you worse first just to see what you're gonna do. So stuff like that, not that helpful, but, uh, Googlers can be very helpful. We actually have an SEO Patent Podcast, which is like AI 10-minute summaries of SEO patents. If you, if like me, you struggle kind of reading the, the detail. I do want to touch back on, um, as it was kindly mentioned in the intro, this Google exploit that we reported last year. So for those that hadn't heard about this, we found around 2,000 kind of properties which were conditionally appearing on Google Search, which were things like classifiers for queries, they were different metrics for site scoring. Uh, and thinking about this more, we had a leak around 2019, uh, talking about Twiddlers. Who here's heard of Twiddlers? We know not, not many. So the documents that were leaked about this Twiddler framework essentially described it as a set of kind of self-contained programs that can do things like re-rank search results after ranking has been completed. So Google goes through this long kind of ranking process. And then on the front end, they can readjust things as necessary. Once we combine all those bits of inside information, it becomes a lot clearer when Googlers tell us things like, we don't use CTR directly in ranking, because they're using it in this front-end Twiddler framework. This was backed up by more data we found in the exploit that seemed to show there was predicted click-through rates for everything in the SERP. And actually, Googlers for almost a decade have been saying to us, This is 2016, looking for changes in click patterns in the SERP. So if we have a predicted click-through rate and something is wildly different from that, it's a good kind of hint to them that they need to be doing something about it. We actually have specific run-through tutorial details about this in the DOJ exhibit. So that's one of the actual, believe it or not, slides from Google, and they describe a situation in which they are re-ranking results kind of on the fly. Um, a lot of this, obviously, this documentation is quite old. I think now where we are, as maybe a few speakers have said, a lot of the important inputs are based around models that are built on user behavior, and these are some of the most important signals now. My shocked Pikachu moment for our Google exploit was finding something called site quality, which sounds really exciting as an SEO when you find something called that. Um, there is a site quality score patent which describes, um, essentially how many times people are specifically searching for a particular website and when they specifically choose that site, which kind of goes in line with, again, a lot of the concepts people are talking about, whether you want to call it brand or navigational queries, whatever. These scores as well, interestingly, were assigned on a site basis, so a subdomain basis. So if there was a subdomain, uh, like a forum subdomain or a help subdomain, it had a separate score to the rest of the website, which again I thought from a technical point of view was interesting. The distribution of scores was a normal distribution. It looked like this, so very, very few at either end. You'll notice there, eagle-eyed, that it starts at 0.4, and that's because the data that we got was specific actually looking at a, um, a specific search feature on Google Search. So we kind of accidentally discovered that if the site quality was below 0.4, you were not eligible for these search features. By search features, I mean things like People Also Ask boxes, uh, feature snippets, that kind of thing. So it really made me realize sometimes you can be optimizing for like a race you're not even in. So, imagine you have not hit one of these kind of score thresholds that you need, and you're like, well, we're optimizing, we need to write our text like this to get some kind of snippet. But actually, Google's not even considering you. So, this is what I mean when we're saying thinking outside of just the content. And Google's talked before as well about the concept of consensus. We did find what looks like scores where Google's trying to work out if content kind of goes with the consensus or against it or is neutral, which I think is really interesting information again in terms of your thinking, in terms of why you are or why you are not ranking. I'll give you some examples. So there was another query classifier called is debunking query, which just got me down this path of thinking of if there's a search term like is the Earth flat, Google gives you a very definitive answer there, and actually all of the SERPs tend to be the science-based with the consensus. Bad news, it's a sphere, right? So I think depending on the query is depending what kind of consensus they're looking for. So if it's something like this, maybe like a medical thing, they can go towards the consensus with the idea that that's going to be correct. Interestingly, if it's something maybe like you're Googling something political and there's no objectively correct answer A good SERP might look like you have some things that are with the consensus, some things that are against it, some things that are neutral, which then puts you in the position where if you're like, well, why is my content not ranking? Because we're definitely better than number 3 and 4. But we'll say your content is with the consensus. Google already has that in position 1 and 2, and it wants in position 3 and 4 something neutral or against the consensus. So it's not that your content is not good. It's not just the right flavor for the syrup they're trying to make. So again, I think we can be quite one-dimensional in our thinking of what is good and why am I not ranking, when sometimes, you know, they already have those ingredients for what they're trying to bake. I found this tool on SEO Notebook by Steve Toth. Uh, it's kind of an LLM-based approach to calculating consensus. I think it's a really nice use for an LLM because naturally the way they function, it's going to give you a good idea of if something is with the consensus or not. Lastly, we had— we've done a bit more data work on this. So we found there were 8, uh, what Google called RQ semantic query classes that most queries seem to sit inside. And they pretty much are what they say on the tin. They were spread out in this kind of distribution. So quite a lot of short facts, kind of bull, um, types of categorization. And what interested me here was I was, I was wondering, well, is this going to impact which SERP features are generated or the likelihood we're going to get an AI overview? And to make this useful, we need another tool. So we built, um, well, fine-tuned a DistilBERT model. So we trained a model on 4. 6, 4.8 million queries that were tagged. So now you can use this tool for free at rqpredictor.streamlit.app. You can put in your queries and it will make a prediction on how Google would classify them. Why would you want to know that? Um, I'm just finishing a piece of work with this lovely chap, uh, called Tomek who runs ZipTie. ZipTie is a fantastic platform to monitor AI overviews. And we've taken, uh, all of these millions of queries that have been labeled into different query classes and then stacked up the likelihood of an AI overview appearing. So now you have the ability, you have a tool to go there and say, for this content, this is how we think Google's categorizing this. And then we have the statistical data on this is the likelihood of an AI overview appearing or not. Um, I have about a minute left. No time, many graphs. We're going to send these out all for free because there's a lot of data there. So if you go to Core Updates, you'll get an email, uh, in about a week's time that'll have this data. So the idea is now we've leveraged this information, put it public, and you can help with your kind of planning if you're going to be impacted by AI overviews. So I'll summarize with I think it's really important to align your SEO strategy with system goals, whether it's Google or any other platform that you're on. There's a saying that I hear Lily-Rae say quite a lot, which is, it works until it doesn't. And this is what this applies to, essentially. There's always ways to exploit and trick things. Sometimes it's good to know where those soft spots are just to adapt your tactics. But if you lean in too heavily, at some point, um, you will be adjusted. Or if you are an affiliate and you just wanna make money, you can just do it. Gary's not a cop, he can't stop you. Um, inside information can verify your understanding. I see loads of really amazing, brilliant theories in the kind of SEO community. And what I always try and encourage like our team to do is if you find those theories interesting, go and try and find out how it's actually working. Because if you can't find any evidence, no one's said anything about it, there's not a single patent, nothing referenced it in a leak, is it worth spending your time on? And I do believe now many important things, uh, we can't directly influence in terms of SEO. And I think I won't labor that because I think a lot of people have made this point about brand and user behavior. Lastly, play more video games. I think they're great brain food, um, and they improve how you think about big complex problems like trying to rank in Google. I'm Mark. I'm on Blue Sky LinkedIn. Feel free to Google me. I'm the only Mark Williams-Cook, so I should be easy to find. Thank you. Thank you.