Conversation with Mark Williams-Cook | Does Answering People Also Ask Questions Improve Rankings?

Hello and welcome to this week's episode of Search with Kander. I will do an intro in a minute because actually it's Mark and I in the studio this week. I just want to preempt some technical issues we have with the audio. About 37 minutes into this recording, our audio just stops. So the podcast will just kind of stop at the end and then you'll see me again in this little outro sequence talking again and kind of giving the explanation of what we were talking about at the end there. We don't know why it happens. We're going to fix it. Don't worry. I apologize in advance for Mark just being cut off mid-sentence towards the end of the show. There was only a couple of minutes left of the podcast anyway. I think we have like 4 or 5 more minutes left, but yeah, it does just kind of stop at the end. I apologize for that in advance. I think the content is still great. It's still a really interesting conversation with Mark for the next 37 minutes, but it will just kind of end abruptly, shall we say, because the audio cuts out. Like I said, I apologize in advance. I hope you still enjoy the episode. And yeah, please do enjoy and still let me know what you think of the conversations we're having as well. And I will see you at the end of the episode. Hello and welcome to this week's episode of Search with Kander. I'm back in the Candid studio and I am Jack Chambers-Ward and I'm joined by Mark Williams-Cook.

Hello, sir.

Welcome back, Mark.

Thank you. It's been a while. We always say that, but it has been like, like, you know, when you're going out and you say, I'm going out, out. It's been a while, while.

I feel like we, we said we would do these like once a month and it's, and it was like, oh, sometimes it was 6 weeks and then sometimes it was 8 weeks and it's like, oh yeah, we haven't done this in a while. And some topics we wanted to talk about are now like old news and all that kind of stuff. But since then I've had Jessica Bettis, one of the other directors here at Kander in the studio, had Annie Mae Hodge in the studio as well. So yeah, the seats have been kept moving.

I couldn't believe how long ago it was, uh, you, you had Annie Mae Hodge.

4 and a half years.

Yeah.

Yeah. It was crazy. Uh, I'd only started hosting the podcast like that January and she came in in July, I think it was. It's like, yeah, crazy. But yeah, we are here. We are back in the Kander studio. We've got some topics to talk about. Including how much does asking, or sorry, how much does answering also asked questions and people also ask questions affect the rankings? That's a study we've done with Chris Green and lots of interesting also asked data. We're also gonna be talking about the kind of core updatey kind of stuff that happened that affected a lot of self-promotional listicles and perhaps some other interesting things we found out about what might affect rankings and things like that. This episode is sponsored by Trendos, and you'll be hearing from them later on in the episode. But let's get stuck in, shall we, and talk about Chris Green's case study, an article that he did with us over at Also Asked, shall we?

Yes. So Chris Green, if you haven't heard of him, he's been on the podcast 4 times, I think.

I think he's a record holder, like number of appearances or something like that.

Can't get enough of Chris. So he's the technical director at Torque, if you don't know him. Um, and has been doing SEO like 15 years plus years. Yeah.

I've said that intro a few times. Okay.

So, you know, yeah, I knew it was a long time as well, but, um, always have fun talking to Chris and I'll point out this study he did, um, was kind of his idea he was talking to me about and it was done completely independently of AllSawUs. So We published his study on, on our site in the insights section. And I had to add a kind of disclaimer after a while because, um, the results kind of, if you like, were good for us as a company. So some people are like, whoa, this all seems a bit sus. And I'm like, no, he—

What a coincidence.

Yeah.

You say, oh, it turns out what we say is good.

So, um, yeah, we had just given him API credits basically to go and do the study and I didn't have any input into the methodology. And like, I was just like, just go away and do your thing and then talk to me about it if it's interesting. Um, and the headline was essentially, and I'll go into a bit more detail about it, that with some of the top ranking sites, there was a strong correlation for if they are answering more of the people also ask questions that are related to the query that has been done. they tend, tend to rank higher in Google. Now that's really cool for a couple of reasons. One, hopefully if you've been doing SEO a while, that shouldn't come as a massive surprise to you, right? Because this is the whole topical coverage thing and being helpful to users.

Part of the genesis of Also Asked in general, right? It was like, well, you were already using that data.

Exactly.

In your content strategies.

Exactly. So I would hope that people aren't like, This is so unexpected. But it was cool because why I was excited about Chris doing this study was I hadn't actually seen anyone verify that.

Yeah. I remember us talking about this a little while ago. You were thinking about doing this sort of thing where, again, like an independent kind of case study, I think is particularly interesting for something like Also Asked. Because like you said, it seems intuitive, but actually having real data for me as a, you know, evidence-based, science-based, database kind of person, That makes all the difference because every tool is going to say, yeah, obviously it works. Yeah, but actually having something that can back that up. And I think that's been a, a really interesting thing of what we're doing with all the insights and also AST is we're working with people like Chris and Arnoud Hellemans and Doron Sharpaow and Alice Rowan and all these cool people across the SEO industry in different facets as well, right? It's like, here's how you can use it in a technical way. Here's how you can use it in a contenty way, all that kind of stuff. But bringing external people in to do that I think helps that kind of scientific validation, the evidence-based validation, right, as well.

Yeah, absolutely. So I just wanted to run through, for if you can't be bothered to read it, reading is not your jam, uh, Chris's methodology and some of his key takeaways. So his methodology was armed with Also Asked API, SERP API to scrape search results, and OpenAI's API. It's a bit of a mouthful. Uh, he was looking at the correlation between ranking and PAA questions. And basically step 1 was he started with a list of seed keywords and he would fetch the top 20 results for those keywords in Google. And then for each of those keywords, he would get the people also ask questions back from AlsoAsked. So that would normally be about 25 questions per, uh, keyword. And then for each organic result in that top 20, he would extract the main readable content, so scrape the website. And then all of that information is sent to ChatGPT asking, does this page fully answer this question, right? Yep, fully, partially, or not at all. And then based on that, he would score how the page is like coverage of of that topic. Um, and essentially that was allowing him to see the relationship between how fully pages answered the related people also ask questions versus where they were ranking. And his key takeaways were this. So across the top 5 positions, so the top 5 in Google, pages that fully answer more PAA questions tend to rank higher. And there was quite a strong correlation there, which was particularly interesting for me because there's like a million different things that are impacting rankings, right? So even if something is, as you could call it, definitely a ranking factor, it's quite rare that you get, you know, very strong correlations because, um, there are too many other things that can impact that.

Yeah.

Interestingly, he found commercial and transactional queries showed the strongest correlations. Navigational and information queries were weaker. Navigational, I can understand. Yeah, I was quite surprised by—

That was my biggest, one of my biggest takeaways because I had a conversation with Chris. There is a video on the Also Asked YouTube channel. I'll put a link for that in the show notes where Chris and I go into depth and talk about his processes and stuff as well. And that was something I pointed out was like, you would think that the questions being an informational thing would lead to higher correlation with informational queries, right? I was also surprised because it's something I mentioned. I've got it was SEO in 2024. I think it was my first one where I talked about like using also asked, I didn't say also asked data, but like—

PAA data.

Yeah, PAA data.

I love that you're branding PAA data just as also asked data.

Exactly right. Yeah, yeah, yeah. Just totally on brand. But no, it was generally like, If you, if your customers have a question in your e-commerce site, you should probably be the one providing that answer and not leaving that up to forums and other stuff and all that thing. I know we had a, the example I always use was a weird example was an old e-commerce client of mine where a lot of the questions around like Reddit and stuff like that was like, are they legit? And I remember having a conversation with them like, should we be answering that on our own website? I'm like, Uh, I mean, if you weren't legit, you would just say yes. And if you're not, you would also say yes. So, uh, you can't really win here. But in general, asking, answering questions, even from an e-commerce standpoint, like that post-purchase journey as well of like, oh, I might need to come and get, you know, how to use this thing or other accessories that work with this. Or if I need to upgrade at some point, how do I do that? And all this kind of stuff. that I think often gets neglected in the traditional funnel kind of thing. Right. And I think that ties perfectly into this of like, oh, right. Yeah. Commercial and people actually looking to do transactions are asking a lot of questions.

Mm-hmm.

And it took me, yeah, like you said, it took me a few seconds. Like, surely information would make the most sense to correlate, but I'm guessing, like, you know, trying to think about how I would buy things, you ask a lot of questions before buying things. So that's kind of my, my thought process of like, okay, yeah, that does kind of make sense eventually once you kind of follow the train of thought.

I think it's, yeah, there's, I'll talk about it in a second. There's like a parallel also I used to explore this, but that was the conclusion I was kind of orbiting, which was like you say, if you're making a very specific informational query or question, your kind of goal is to answer that question as quickly as possible.

Yeah.

And I think it's rare that you're then gonna go onto like a subsequent sort of chain of, oh, I also wanna know this, know this. But yeah, you are completely right. If you are in that, do I buy from this company or this thing? There's immediately several different types of questions.

Mm-hmm.

That you answer. And that I'm, I'm interested as well now as a further study, how this is reflected in things like Gemini and, uh, fanout queries. So if you're asking specific informational queries, what the size of the fanout is versus if you ask more commercially focused, like about a brand type query, how the kind of breadth of questions that it's going to infer that you need to do that because there's like subtasks there almost like, can I trust this brand? What do other people say? Is the price the best or immediately things you ask basically about every company you buy from, right? And the, yeah, the other case study, and it's something that I won't spoil it, but Arnoud in Helmers is going to be talking about at Search Knowledge XL, is using the PAA data to get brand insight by regionality.

Yep.

which I hadn't thought about in too much depth before. I thought about it through the lens of when I was just generally doing query keyword research. I understood that, you know, when you deal with places like America, people in different states across that very big country ask very different questions about the same thing.

We've talked about it before in terms of state law. I know we had a client that was dealing with a specific legal thing and I make up the stuff. I don't remember, which helps. I can anonymize the data accidentally. But like Texas state law is so different to like New York state law that they're like, if they're a solicitor or whatever, they're a lawyer or doing insurance or whatever, that fundamentally changes what you are allowed to say about your products.

Yeah.

And you literally like cannot say, oh, this product does A in Texas. you have to say it does B, where in New York you can say what you like and it doesn't matter. It's like, but the website is the website. I'm not doing like a New York page and like, like you said, like a per state page, it turns into this whole thing of, oh God, how the hell do I personalize that? And I, funny enough, I had a call with Arnoud a couple of days ago talking about this and it got me so excited for his talk because I know Arnoud always has brilliant insights. He's one of those people. I'll put a link for Arnoud's episode of the podcast where he's just full of brilliant insights. I feel like he's one of those people I learn something. I'm like, I'd never thought of that. He's like, yeah, of course you do this, this, this. I'm like, oh my God, yeah, of course. That's brilliant. And he mentioned, because obviously we launched the city-based selector on Also Asked a little while ago, didn't we? When we sponsored Brighton SEO and all that kind of stuff last year. And there was a real kind of thing. I think a lot of people were saying like, yeah, how much does it matter per city? And then Arnoud told me a couple of examples of clients he'd worked with. Like, yeah, I was working with this client in this city in the US and they also had a branch in this city in the US. Completely different stuff.

But you know that feature came from Arno.

And yes, and he, he was in because I was like, yeah, I remember I said that. I was like, yeah, I remember that. He was like, yeah, I talked to Mark about that. I'm like, oh, right, right. Okay. Yeah. Yeah.

He, he made a very convincing case as to why we, we need to do that. And I know how few SEOs understand how important specific geography is versus country geography because I get the help requests of SEO saying, oh no, I just want to do a national search in the US. And I'm like, there's no such thing. Like every time one of your customers does a search within the US, that information on specifically where they are is going to Google. And the results are customized based on this.

And they're getting more and more personalized as well.

And they're not, they're not all necessarily customized. So my point to them is if there is no, um, specific geography, uh, importance around your question, it will just be ignored anyway. So it fundamentally is then a national search. But my point is, you're trying to understand your visibility and the questions your customers have. You have to try and be that customer. So—

And I think a lot of people focus on it on a query level, right? If you, if you put like best SEO agency Norwich, then you get the ones in Norwich. But if you were in Norwich and you have your location turned on and you're signed into Google and you put best SEO agency, It's going to give you local results.

Exactly.

And whether you think of it as a local query or not, and like the amount of personalization is only increasing with how they're doing— Google is tying in like personal intelligence stuff. So like your Google Photos, your Gmail, all this other kind of stuff of like, oh, you've searched for this particular kind of restaurant and booked this on your Gmail before. That now influences your search results. That's a whole other level and a whole other conversation. But yeah, it's like an unavoidable thing. And I think you're right when people have that, I want to search nationally. It's like, that's not how that works. Sure, you might be serving a national audience, but your customers don't live in an entire country. I live in Norwich and that matters when you're searching, right? Yeah. Yeah.

I think it's probably now a good segue to talk about best, best things, best companies, best agencies.

Exactly. Yeah. Yeah. So before we go into that, Like I said, full conversation with Chris. If you want to go and check out that, I'll link the video so you can get more details and link the article and that kind of stuff. If you do want to read, do want to listen, we'll get into more details about that with my conversation with Chris. If a potential customer typed into ChatGPT, what's the best marketing agency or what's the best SEO tool to use, would you and your brand show up? And if it did, what would AI actually say about you? Most marketers don't know the answer to these questions. And honestly, that's a big problem right now. Here's what's happening, whether you're paying attention or not. ChatGPT, Gemini, AI Mode, and AI Overviews are already answering your customers' questions. They're recommending tools, ranking brands, and shaping the buying decisions of customers quietly at scale every single day. And unlike Google, there's no ranking page to check. There's no keyword report, there's no visibility dashboard, there's no Google Search Console. Traditional SEO tools were built for a world that's already changing, and this means most brands are flying blind right now. They have no idea if they're winning or losing because until now there was no way to see that information. That's exactly what today's sponsor, Trendos, was built to solve. Trendos is an AI search visibility platform, a tool built specifically for the way people are searching today. It tracks your brand across ChatGPT, Gemini, Perplexity, Google AI Reviews, and all the other major AI models, and then shows you the data in one clean dashboard. Where are you being mentioned? How does each AI model describe your brand? Is your visibility going up or down? Who are the competitors that are showing up instead of you and why? TrendOS can answer all of these questions clearly and in real time. And here's what makes TrendOS genuinely different from anything else out there. First, there's no demo wall. You don't need to book a call or sit through a sales pitch. You type in your brand name and you see your AI visibility data immediately. Second, you get historical data from day one. Most tools only start tracking from the moment you sign up. Trendos gives you that past data right away so you can see how your AI presence has been evolving even before you knew where to look. Third, it tracks more than just mentions. Trendos tracks brand sentiment so you know not just if AI is talking about you, but how AI is talking about you. It also tracks source attribution so you know which articles and websites are shaping what AI is saying about your brand. It gives you competitive benchmarks so you can see exactly where you stand against your competitors. It gives you a full picture, not just a couple of data points. Pricing is also really straightforward. There is a free plan that actually means something. The basic plan costs nothing, gives you 100 custom prompts and 3,000 analyzed AI answers every month, across unlimited countries and for up to 10 brands. There's no credit card, there's no catch. It's just a genuine look about how AI models mention you and your brand. If you're ready to go deeper, the starter plan is $169 a month. That gives you 250 prompts, 7,500 analyzed answers, and unlimited brands. This is built for teams taking AI search visibility seriously. And for teams scaling a full AI visibility strategy, The Pro plan is just $339 a month. 550 prompts, 16,500 answers analyzed per month. It's the full suite of tools. And all plans include free AI answer datasets and unlimited countries. There's no hidden tiers. There's no contact us for pricing. Just pick the plan that fits and start straight away. AI search isn't the future anymore. It's what's happening right now in your market with your potential customers. Don't let AI rewrite your brand story without you in the room. Click the link in the video description or the show notes down below, or you can go directly to trendos.com. That's trendos.com to check your brand for free and see exactly what AI thinks about you. It takes just 10 seconds and there's no sign up required. That's trendos.com and sign up. Today, there was a big self-promo listicle crushing across the SERPs, it seems like.

I don't think this was like an official Google update.

No, I was trying to— we were talking about this earlier. I was trying to remember if it was tied because there was the December core update.

Yeah.

And then the core update we've just had in March was the first of this year.

Yeah.

So I was like, I don't think there was anything specific. There was a couple of spam updates and stuff, but there was that one that took like a day. But that was after all this happened. So yeah.

Yes. I think this happened January 20-something.

Yes. Yeah.

Because we had, um, 3 or 4 external, and by external, I mean not current clients.

Yeah. We had some inquiries.

Yeah.

Yeah.

Yeah. Being like, oh, uh, we're getting big traffic loss. Can you help?

Yeah.

Um, and funny, it sounds like this episode is sponsored by Arnoud because I had a phone call with him the next day. Just, he phoned me, which is very rare, and he, but he just phoned me. and we had a long conversation with it. And essentially he had clients that were experiencing exactly the same thing. And we had a talk about, um, the types of things that we were seeing. And then it was probably, I believe it was Lily Ray, um, published some data on it as well. And it was essentially very clear what was going on, which was that sites which had a lot of content, which I would define as self-promotional listicles were losing traffic, like hand over fist.

So what we mean by that is if us here at Kander, we wrote top 10 SEO agencies and surprise, surprise, Kander is at number 1. Like you see this all the time. I was speaking to another marketing manager at another agency a little while ago and they had posted like a result. I can't remember which. AI search which LLM it was, but they searched for it and they appeared like straight away of like best SEO agency and they popped up. I was like, I did that search, completely different results. And they had a self-promotional listicle that featured themselves. And ironically enough, when I did it, the LLM cited their article but listed all of the other people and not them. And I can't remember exactly the time. It might even be like before this update. And then I checked it after the update or something like that. I was like, I can see you being cited a few times, but you're not mentioned in these. It's like, it's your position 2 and your position 5 are being mentioned more for whatever reason. And honestly, these kinds of articles have annoyed me for a while. Like I hate that kind of thing. And I know you did a, uh, you and a couple of other guys did a like—

Top SEOs list.

Yeah. Top SEO LinkedIn Pulse, like troll thing basically. It was just like, See, you can just say you're the best and you're the best and just manipulate the rankings quite easily.

So I think that was, so I posted that on LinkedIn Pulse. A couple of other people did it on theirs. There was a few websites that went up. So it was the same list. It was something like top SEOs in 2025.

Sure. Yeah.

But that list persists. And I know when someone's generated a list of SEOs now, because it happened literally last week, someone was like, here are the top SEOs you need to follow on LinkedIn. And it was the identical list in the same order as the one that was in my LinkedIn Pulse article that we'd republished because yeah, that's how the system works. It's not very complicated, but this is actually a really interesting point, which is this approach has been recommended by some big companies involved in the AI/geo space as a tactic that people should be doing to increase their AI visibility. Now, arguably it works in that it was working. That's always true, right?

It works until it doesn't. It works until it doesn't. It works is the phrase you've said many times before.

But, you know, the argument I made was that, yeah, okay, it works if you, if the metric is, okay, we got the citation or we got the click, right? I do question whether someone searching like for best SEO agency and then they land on the Kander website and then they see Kander as number 1, it's a bit kind of like, it feels to me like quite flat face emoji. Like, well, you've kind of lost some credibility with me.

Yes, I agree. Yeah, that's why they annoy me.

Yeah, because, you know, we, I think we're a good SEO agency. We do a bunch of good stuff. Are we the best SEO agency? Probably not. I don't know who's better than us, but I'm, It seems just statistically unlikely that we are the best, right?

And I'm fine. And like, how do you define the best?

And like, I'm kind of like, fine, you know, I'm fine with that. Like, whatever.

We do good work. We work with good clients.

Yeah. And we're nice to get on with, blah, blah, blah. There's a whole bunch of reasons to work with us or someone else. So immediately just positioning yourself for that, it's like, that's my ick list already. It's a bit like, especially then when you've got 50 lists and you're the best content agency, the best video agency, PPC, SEO, geo, blah, blah, blah, blah. Right. So my premise was, even if it works, at what cost? Because what's good— what good is it if people come to your site, they see your brand, and they're a bit like, well, yeah, but I'm sure you do get some business from that because people are like, we— I want to be number one. But we know that those AI results are obviously powered under the hood by the kind of fan-out queries and stuff that happen on traditional search engines. And we know really that's against the spirit of what they want you to do, right?

Yeah.

And so yeah, this update happened and we had some companies that— and you would know, so if you're listening to this, you would 100% know their names.

Yep.

And they have websites driving significant amounts of traffic and they were coming to me saying—

It was like hundreds of thousands of clicks a month.

Yeah. And they're like, we've lost 40% of our traffic in the last 4 weeks. And it's like, oh, and then when we look, um, they have a lot of that kind of content, i.e., best X, blah, blah, blah. And they put themselves at the top.

And it would be a variety of different like tool variations or product variations or service variations. Like you said, you'd be best PPC and best SEO and best technical SEO and best GEO and all that kind of stuff.

Right. Exactly. Um, so is that, and sometimes a mix of like programmatic type content where it's, um, so we had one, not a current client, but again, a big SaaS player come to us and they had produced a lot of programmatic content over the last about 18 months, like thousands of pages, which is, which was comparing them versus competitors. And then each of their features versus competitors, loads of different pages. And they're, They're kind of arguing because they were posting about LinkedIn the whole time they're doing this, being like, aren't we awesome? This is going great. And we put all the information in a spreadsheet, so it's real data, and then we get AI to write it and churn it out. But they're number one all the time. Again, like 50, 50% traffic loss. And the interesting thing as well was there's 2 interesting things. One was some of those pages, and by some I mean like a handful, Survived, and they were all the pages where it was the bullseye of actually what that company does.

Oh, interesting.

So the main product things seemed to survive.

So they didn't stray too far from their topical authority, essentially.

But everything that was even actually quite close, just nuked. But the non- programmatically non-listicle type pages were also being suppressed in most of the cases. So to me, it looked like Google's made a judgment, maybe like you would have a person being like, yeah, you're talking a lot of shit.

Yeah.

So I'm kind of not gonna believe the other stuff you're saying. So their actual money pages, if you like, had, had dropped. And we're talking, you know, when you go from top 3 to top 10, that's very painful.

Yeah.

Um, because they had a massive drop in signups as well.

It really ties into, again, something we talked about a lot when EEAT, when they added the extra E a couple of years ago, talking all about, you know, firsthand experience with your products, being honest, having legit pros and cons of your thing. So you know that Cander is really good at A, but we're actually not good at B. And I know we do this all the time and I, I've mentioned it on the podcast before, like, you know, you would say a lot of agencies will take any client they can for whatever reason, make as much money as you can. Times are hard, all that kind of stuff. Not judging people, you know, trying to, you know, make money and survive and all that kind of stuff. But I've always think it's credit to what you and the directors do here where we say no to clients, or we say, actually, no, you'd be better off with a freelancer, or I think you'd be better off working with this company instead. Because We're honest about what we're good at, what we're not good at. Yeah, we don't do PPC anymore. We probably could take on more PPC clients if we really wanted and hire a team and blah, blah, blah. We don't want to do that anymore. So we recommend this freelancer or this company or whatever. And I think that not only, like you said, builds up rapport and like goodwill with companies in the long run. And I know we've had examples of people coming back around to us and being like, oh, remember you had a conversation with me about this? 2 years ago. Now we're looking for a website migration. We'd like to work with you. Like, aha, now you're in our wheelhouse. Let's come and work together. And that's also true for humans in the same way that it is for search engines. And I think there's so much of what we've been talking about in terms of that, like you said, bigging yourself up, just being like, I'm the best, I'm the best, I'm the best. That annoys the humans. And eventually, you know, the search engine, the algorithm is going to catch up with that and reflect human behavior. You know, we also talk a lot about like how often companies talk about themselves. Like you said, the brand visibility stuff we talk so often in LLMs and the whole who are you being talked about by and what are they saying about you? And if it's you're talking great about yourself, but there's a bunch of, I don't know, forums or message boards or social media posts or whatever, they're like, oh yeah, but Jack didn't do this and he's a terrible podcast host and blah, blah, blah. But I'm like, I'm the best podcast host.

You saw that.

Exactly. Oh, that was you. I wonder who Mark SEO was. Um, but yeah, there's that element of like, you know, just that, yeah, bigging yourself up too much. I find that stuff really annoying as a person and it makes sense that it's reflected in that, I think.

I think that Google has, this has been added to the like system prompt of Gemini because I don't know if you've seen now when, um, AI mode returns these kinds of articles, it literally has a warning now.

Oh, really?

That says, like, um, you know, be aware lots of these types of contents are self-promotional, in like a little warning.

I didn't know that. That's cool.

Yeah, yeah. So it obviously— the LLM surface is a bit more spongy than the search surface. So the search surface is very easy for them to algorithmically say, nah, don't show this. obviously whatever's baked into the model is kind of in the model. And I think by the nature of how the token generation is happening, it's almost impossible before the tokens are generated to be like, don't do this.

Yeah, yeah.

But they can add to the system prompt, obviously, if this is the type of content, and it's easy to describe that type of content, you should warn the user about the intent. So I found that interesting that Google's been like, Obviously we notice people are doing this to exploit like AIOs, to exploit AI mode.

Yeah.

And it's working. So we need some kind of— I mean, I've, I've— the whole— every time system prompts get leaked, it's just a circus to me because, because of the nature of how those systems work, it's very difficult for them to build in like— well, they can't build in like explicit programmatic rules about what is shown or not. They just have to have these like panicky prompts like before the content's served where they're like, whoa, don't say that! Whoa, don't give advice on that! And be professional, uh, don't use 12 exclamation marks, and, um, don't refer to the user by their, by their first name even if you know it.

That's weird.

Like, there's, there's a Gemini one where it was like, there's all these really specific instructions about not referencing like the question that people are basically desperately trying to make it sound like a normal person. It was like you, you brought an alien into a conversation and you're desperately trying to coach them just before they have this conversation about how to act.

Yeah. Um, it was the classic thing of like, I think it was, was it a Claude ad or a Perplexity ad? I think it was Claude that used the ChatGPT thing of like, you ask it a question, it's like, great question. I think you're really smart and cool. And I think I really appreciate you asking that question. question. Aren't you brilliant? Anyway, here's some answers. It's like, all right, mate, like, chill out. Like, I don't need this, like, self-aggrandizing weird stuff from my LLM.

Claude is a bit more of the Gordon Ramsay approach of, of feeding.

You're an idiot sandwich.

Um, so on, on the subject of query fan-outs and how LLMs work, we were having a discussion about being a not Award-winning agency.

It's literally on our website. Yeah. That's one of our taglines.

Yeah. So this, this is really interesting subject for me. So generally as like a thing, Kanda since I think 2016, 2017. So about 10 years.

2016 is when it turned into Kanda. Yeah.

Yeah. So basically since we've become Yeah. We just made a decision that we don't want to enter awards.

Yeah.

Basically, I've got nothing against the awards business.

Yeah.

Um, it was just kind of like the way we communicate what we do and the way we plan to build the business wasn't going to be reliant on like external accolades.

Yeah.

And like, fair play to the people that do awards, they're a lot of effort.

Oh yeah. Um, and like some companies have like specific people who go through that process. Like you have members of the team that are specialized in that.

Exactly. We've got Gemini now. But yeah. And also, I mean, yeah, like back then, um, probably didn't have enough money to do like loads of award entries because they were expensive.

So we were just like, thousands, often like 10 grand at a time, right? Yeah.

So we're just like, nah, I'm not going to do that. And it was a bit like, um, so if you're outside the UK, I don't know if you know this, there's an insurance company called Direct Line and like their big thing for a while was probably still is, like, you won't find us on comparison sites.

Right.

Um, they were just like, we're not going to play that game. Like, we do good deals, come to us if you want it, blah, blah, blah. So it was similar to that. I was just kind of like, look, we do good work, here it is. We don't do awards, just come to us if you want to speak to us. And that worked fine. And it became a bit of a, like, fun thing because occasionally people mention it almost as like a meme that, oh look, this agency's award-winning because it's just like the tagline of so many agencies because there are so many awards now. And again, you almost need a guide to which awards are good because I know there's a couple. Um, I think for instance, I've got a lot of friends, we've got a lot of friends that judge in like the European Search Awards, which is a Don't Panic event. And that looks to me like pretty decent. I see the conversations they're having and they put genuine work into deciding.

I've spoken to, like you said, a couple of the judges about the process and stuff. There is a lot of effort that goes into that. And like my experience from the podcasting side of things, I've been a judge at podcast awards. I'm a board member of the Independent Podcast Awards. Like there's a lot of work that goes into it behind the scenes as well. Right. And there can be totally legit ways of doing it. But yeah, it's an interesting thing.

I mean, we've won awards we haven't entered. I've, I've, I've got emails basically saying like we've been selected as top web agency and blah, blah. And, uh, we just need to pay like €200 to get the, like, they're going to send us a trophy. And it's like this really cheap trophy, but you just buy it for €200.

Yeah.

It's literally like, well, you don't know anything about us and you've just given us this. And they're obviously just emailing loads of people, but it's just like a badge we could put on the site. So I was always a bit like, do clients know the difference between good award and bad award?

Yeah.

So that was the thing.

However.

However. This is what has, um, boggled me slightly now. So we were building some more tooling internally, playing around different tools, and, um, when recently ChatGPT moved all of their fan-out stuff out of the browser, you could only get it through the API. Although now in ChatGPT 5.4, it's back in the browser.

Hey.

So you can use Chrome extensions again. But in that short gap, I was like, oh, we probably be able to see those fan-out queries. So made a little tool that used the ChatGPT responses API. So you put in a query, really nice little tool, and it says, here's all the fan-out stuff. So it's live currently as well on queryfan.com if you do want to use it, it's free. And one of the, the questions that I tested on this tool was best SEO agency UK. So I was like, what happens if someone types this into ChatGPT? It whirred for a bit and then it listed its like 20 fan-out queries it did. And it was like SEO agency Drum Awards 2026, SEO agency. And it was doing like specific award brand searches.

Wow.

To find agencies that were on shortlists for well-known awards. And I was a bit like, ah, so I was like, we're 100% not on any of these. And then I was And my immediate reaction, obviously my immediate biased reaction was like, well, this is bullshit. This is terrible. And then I was like, well, I guess if you had no other external information, you had no industry knowledge, what would be an okay method to judge, again, tricky word, best agency? Yeah, I guess who's been winning awards might be something you look at alongside reviews.

Especially like you said, I think a lot of people don't understand a lot of the background backend type stuff that happens. Like, oh, you have to pay for, to enter this thing or whatever. Like, or the, like you said, the people that just send you a random email. It's like, here's a little trophy. But like, if you don't know and you're not like in that industry, like take me out of SEO or podcasting, whatever, and say like, oh, you want to go and buy some other thing? Like, sure. It won the best I don't know, Hybrid Vehicle Awards. I'm like, what? I don't know what that is, but sounds good. Sure. I have no idea. And who has time to go and like, is that a good, is this a good, like almost like the question we were saying earlier with the e-commerce kind of like, is this a legit awards show? And like, where would you find that information? Like, yeah, that's interesting. And the fact that it's, I didn't know this until you just said that, like it was doing the specific branded things like European Search Awards, Drum, all that kind of stuff. It's like, okay, it's pulling out. specific examples.

That's really interesting. ChatGPT 5.4, different to 5, yeah, um, has been doing a lot of fan-out queries with the site operator. So specifically knowing, okay, for the— so it does it if you search like for a bunch of things about SEO best practices, it will do a site:google.com search to get their dev document documentation. Um, I haven't done this search on 5.4 yet, but my assumption will be it will be using site searches for a few of these award sites.

Interesting.

Um, as well. So yeah, it did leave me, you know, because then you get the knock, the, the statements made. So if you win a European Search Award, you will say, oh, we're the best in Europe for this. Yeah. When obviously, like, you're the best out of the 20, 30 people that applied for the award, not the 50,000 agencies in Europe. But having those statements then on the site obviously gets fed into LLMs, and again, they're, they're pretty much just going to take what they read, and likely if there is the link back to the site, like European Search, that is that kind of verification process, right? Kind of like a language verification. So I'm very much now on the fence of, well, I don't wanna be on the fence is the thing, 'cause I think the fence is the worst place to be in marketing. You need to be on one side.

So yeah, I wasn't joking about that abrupt ending, right? The audio just absolutely cut out. I have no idea why. I'm gonna go into the studio in a couple of weeks and try and diagnose that. But if you are still here, Mark and I and the team are going to be in Brighton next week. We're doing a live podcast on Wednesday the 29th of April. If you do want to come and check us out and you're listening and watching this on the Monday that it comes out, of course, I'll be teaming up with the SEO Mindset duo of Sarah and Tasmin. So you can come and check us out at a brand new venue in Brighton. If you're in Brighton, if you're attending Brighton SEO, come along, come and say hi. It is usually a fantastic time. And I'm sure You'll have a brilliant time. We'll have drinks, we'll have pizza, we'll have a great conversation. And yeah, come and join us for the live podcast in Brighton. In the meantime, you can get tickets for Search Norwich XL as well. That's on the 24th of September, 2026. And yeah, tickets are on sale now. Go to searchnorwich.org and go and check them out there. And apart from that, I think that's everything. And, uh, yeah, thank you so much for watching. Thank you so much for listening. Sorry again for the audio weirdness. And, uh, yeah, I will see you next week. Thanks for listening. Thanks for watching. See you next week.