How to Win in AI Search: 
Proven Tactics and Research Across ChatGPT, Gemini & Google AI

James Lawrence: Welcome back to the Smarter Marketer podcast. I'm here today with Joe Alder. Joe, welcome back to the pod.

Joe Alder: Hey, James. Good to be here.

James Lawrence: , Looking forward to today's conversation. I'll just do a quick intro, Joe for those of you that might not have, uh, caught one of your , previous episodes. So Joe is Rocket's Head of SEO and GEO, and in my view, one of Australia's leading voices on the evolving AI search landscape.

Over the years, Joe has worked, um, both in Australia and in the UK with some of the, the top, , brands and e-commerce platforms and lead generation businesses in both of those markets, working across SEO strategy as well as implementation of optimization. Over the past three years, Joe has h- basically led Rocket's response to the changing nature of the search landscape, working with Rocket's leading clients to help them navigate the shift from traditional search to AI-driven discovery across platforms like ChatGPT, Google AI Overviews, and Gemini.

And I thought it'd be awesome to have a catch-up with Joe to essentially just talk about- How to do great GEOs. So I guess a l- a look under the hood at what is working at Rocket for our best performing clients who are doing great things in AI search, who are maximizing visibility, who are kinda managing what is in many instances, a transition, kind of away from traditional search towards this landscape where now we've kinda gotta have a foot in each camp and pay attention to both.

So, Joe, I thought just a, a place to kick it off would be just to talk around, I guess, what you're seeing macro. Like, how much has search shifted in the past 12 to 24 months? What's happening with Google v ChatGPT, Google and ChatGPT, other large language models?

Like, what are we seeing from the Rocket side?

Joe Alder: Yeah, sure. So I guess, like, there's a few different categories here to break it down. So first is, like, adoption of AI in Australia, which I think has grown significantly. Like, I think maybe last time we even had a podcast it was high, but it's continually grown.

I think it's basically doubled year over year, um, in terms of search. So I think we have some stats, like 77% of Australian adults are using AI tools at least once a week, um, which is obviously a huge, huge number. And then when it gets to purchasing and searching for products, that has also increased, so 66% of B2C buyers and around 93% of B2B buyers.

So the majority of Australians are using AI, and they're also using it to inform and even make purchases.

James Lawrence: Yeah, it's fascinating, isn't it? I think we, um ... I was catching up with Ash in our marketing team, and we received more inquiries from AI search last month than from any other digital channel, which is kind of insane, right?

When I think 12 months ago was if we got one, I was like, "Oh, this is kinda cute. This is interesting." But it's kind of actual re- It's real now, right? The, the hype has become real.

Joe Alder: Our GEO strategy is working Now what,

James Lawrence: now ... I promise you that wasn't a, uh, you know- ... I was not teeing you up for that shameless plug.

Um, and , for marketers listening, like- ' when we speak to new clients or existing clients, everyone's in a slightly different position with this. I think you've- we've seen these kind of doomsday scenarios or things on LinkedIn where traffic is plummeting to people's websites.

What are you seeing across the basket of Rocket clients? And I think also not just Rocket, but more broadly, like who are the winners? Who are the losers? Is Google shrinking? Is it growing? Is traffic still going to websites normally? Are conversions happening at the same rate they used to? I think 'cause it's particularly for in-house marketers where they're kind of siloed, right?

They're only seeing one Google Analytics instance. They're only across one website. Like what's normal now and what, what has happened typically to websites over the past couple of years?

Joe Alder: Yeah, sure. So, I think about two months ago we did a massive Australian-led study where we looked at two hundred and fifty Australian businesses to basically answer that question, like what has the impact been?

Um, so we looked at all different channels, but even like zooming out Google is actually growing, which I think surprises a lot of people. They did their Q1 report obviously the first quarter of '26, and their revenue in search was up around nineteen percent. From the two hundred and fifty cohort of Australian businesses we looked at, they'd actually seen an increase of forty-three percent in impressions.

So that's a number of times someone was searching for something and that brand was appearing. So more and more people are searching than ever before, which I think is like one side of it. Despite that, we are definitely seeing a drop in traffic and, yeah, it would more be, I would say, a drop in clicks because as I said, more people are searching, but less people are clicking through.

So average site loss from a study we looked at, there was a much broader thousands of websites, was around thirty-four percent drop in traffic, which if your website a few years ago dropped thirty-four percent, that's like mission critical. So I think there is definitely like a reason people are getting alarmed and worried, even from our own clients, but we are performing great.

So we saw forty-three percent increase in impressions. We were down around two point five percent. But I think there is currently no need to panic. Yeah, for my last point, the core reason for this, or at least like I would say attributing to around eighty percent of this drop is AI overviews, and AI overviews, everyone would have seen now, they're basically Google's response to ChatGPT and the LLM platforms.

And they almost have to cannibalize their own search because they would be losing money with people not clicking on ads. But to keep people on the Google platform, Google has been using a lot more AI overviews. So even in '26 versus '25, um, again Q1, there was a hundred and ninety-three percent increase in the number of AI overviews that were appearing.

So it went from around being present on around ten percent of searches to around thirty. And now when that AI overview is present, less people are clicking through, so it's around fifty-eight percent. They might search for, do AI overviews hurt websites? I think is, is the example we used in the webinar.

Google says yes. People then have their answer. They don't click through, uh, at least fifty-eight percent of the time. That in turn then means that traffic has fallen. But that being said, although traffic has dropped because of AI overviews, the majority of this is very much informational content or like answer question-based content.

So if you're a business selling products, service, even software or anything that kind of operates a business online that is not editorial Most of this would be taken off those kind of informational searches versus people actually looking to purchase your product or service

James Lawrence: Because that's it, isn't it?

It, it feels from a, a very top level AI overviews have ramped up dramatically the informational searches. So anything kind of top of the funnel, like if you're a publisher, a news site, a blog, et cetera, traffic potentially has been completely decimated. If you're a business that used to have a content marketing strategy and was a mix of a blog and resources, but also hard-nosed bottom-of-the-funnel commercial searches, you might find that the blog and resourcing gets less traffic than it used to.

But we're finding across our clients that their commercial results themselves should still be continually growing. It's just the, the way to find those commercial pages and commercial, um, calls to action are different, right? Where people are AI overviews, ChatGPT, combination of both, but - the conversion still should be happening on your website.

Joe Alder: Yeah, 100%.

James Lawrence: So yeah, I think if, if you're a marketer out there, you're not alone if your informational pages are losing traffic. I think the stat that you pulled up recently, I think it was in a staff meeting or, or in the webinar, was that our clients had greater visibility through Google than ever before.

If we're looking at Search Console, looking at where they're being returned, um, AI overviews, first page of Google, traffic to sites was I think down 2%, which I think is, is good. It means we're doing a better job than the average 'cause traffic in Australia I think had dropped, closer to 30%. But then your actual commercial results should still be as strong as ever.

Joe Alder: Yeah, exactly.

James Lawrence: Cool. So I think that's, that's good framing for the conversation. Where I really wanna steer it is into what are the similarities, what are the differences? I guess if you can just, like, lift up the hood, right, and talk about how are we doing at Rocket? How are we working with our best-performing clients to help them navigate this changing landscape, and how to put together a GEO strategy.

So I guess can we kick it off from the very beginning? What are we doing for our clients? How are we approaching it? Um, and what's that framework that we're using to guide clients through?

Joe Alder: Yeah, sure. So I think, like, to go back to maybe three years ago when this started and it's like AI is a thing, let's figure out what we do for our clients.

I think there was a lot of information online, and I think even Google has come out and said that good GEO is just good SEO. So like we really wanted to see whether that was true or not, and I think we would've won in any scenario, right? We had SEO clients. We were worried about losing clients because of the fall of SEO, as people were saying, although it's been dying for 20-plus years.

So like the first thing we really wanted to do was to understand if there is in fact an overlap so we basically did a study where we pulled, uh, I think it was 100,000 keywords, and we pulled them across the different AI models as well as ChatGPT, which is still the largest kind of AI search, uh, that drives traffic to Australian businesses.

So what we found is AI overviews being like a product of Google overlap a lot. So if you rank position one to 10 in Google for that search term, around 44% chance you'll be included. If you rank position one, it's around a 61% overlap. But when we look at ChatGPT, it's actually only around a 15% overlap for position one to 10, and that is from pulling the same search term 10 times.

If you only pull it once or twice, the overlap between Google is, like, 2 or 3%. So if SEO is just GEO, you would expect to see much more similar results in what's actually pulling. So that's kind of where we started, and then once we had our hypothesis that there is different ranking factors for both, it was basically about how do we reverse engineer here and then get to what is actually required.

Um, so yeah, I thought that was interesting framing because there is so much advice that it's just SEO, it's just GEO, and I think it is true that SEO sets the base for GEO. We've had clients that have come to us and been like, "Our traffic is dying because we don't have a GEO strategy," and we've recommended they do an SEO strategy because their SEO is terrible.

They've lost loads of keyword rankings 'cause of other issues. So they definitely work hand-in-hand together. But if you want to dominate on both, you need a strategy for both.

James Lawrence: And, and it's so interesting 'cause I feel that, um, I- At Rocket we've always been really non-sensational. Like we, you know, we gen- generally try to put out awesome content through the podcast, awesome content through webinars and in that period, two to three years ago, the more rational, uh, experienced part of the SEO community was...

like the prevailing sentiment there was it's just SEO, it isn't different, and there were lots of kinda cowboys running out, super aggressive. I suspect listeners were getting spammed in their LinkedIn with this, "AI is changing everything and you need a new approach, and it's different and it's different."

And so we're kind of stuck where we didn't wanna be the sensationalist kinda cowboys s- going off and saying everything's changing, but we also felt that we're doing the wrong thing by our clients and by the industry by saying it, it is just SEO, 'cause it's not. And yes, there's heaps of overlap, and yes, as I think as you just framed it, it is this kind of foundation or it is this base for good GEO.

But we found very quickly that if we were just doing the same things we'd always done for our clients, we weren't doing the right thing by them as it related to them driving as much visibility as possible within ChatGPT. I think it was quite a challenging position for us to walk.

Joe Alder: Yeah. Yeah, 100%.

James Lawrence: So maybe if we can, um, maybe let's dig deeper into that.

So what is the Rocket approach? Like, 'cause despite all the change, SEO two years ago , drove 60% of traffic to Australian websites. Paid search, 10 to 15% paid social 10 to 15%, direct 10 to 15%. It still is the biggest lever, right? And even, even though we've got change and we've got traffic dropping from organic and informational, it's still the foundation for a good digital strategy is SEO.

So I guess, how do you blend the two between not throwing the baby out with the bathwater and still ensuring clients are driving traffic and conversions through traditional search, but also then not ignoring what is, kind of a, a big wave coming from behind?

Joe Alder: Yeah. So we actually did a study on that as well when we looked at 250 websites, and SEO is still the biggest driver of traffic.

Around 48% of traffic comes through from organic search, it's still huge. Um, and obviously a part of that is also then from these AI tools. So I guess like how it's different and the approach, at least how I see it, uh, and this was something that we've obviously had to develop, so I tried to come up with a bit of framework.

We came up with something called the STAR framework, . And basically how we have broken it down is the first kind of pillar is structure and retrievability. In more traditional SEO sense, that would be very much technical. And one of the big things with GEO is Google has always been the leading tool by far, like a 95% market share or so, and ChatGPT would be the equivalent, but it's around 68% of market share.

And I think people are a lot more passionate on their tools, like they're either using Claude or ChatGPT, some are using Gemini, some are using both. So one of the, the challenges with, especially from a technical standpoint, is there is around 10 different main AI tools that people could be using.

Previously, we just had to ensure as marketers, can Google access and crawl our content? We now have to look, can Google and then can these 10 other AI agents that are hitting the website. So one of the issues that we faced really early on, and it's still one of the issues I see very common is actually blocking these AI bots from crawling your website.

Because the way your kinda server works, it has security measures in place, and if someone is continually trying to hit different pages, it almost looks like a hack. So they, that in turn was then blocking. Uh, we had a really big client in the legal space, and they basically had security limits in place.

After X number of hits, they would block the IP, which was then blocking these AI tools. We removed that, and in two months they saw, I think it was like around 150% increase in AI traffic. So yeah, the first step is, can these AI bots actually crawl your content, and is it structured in a way that can then be retrieved?

James Lawrence: And that was an interesting one. That was, um, 'cause we won't mention the name for the of, privacy, but it's it's a very, very, very large law firm coming from a really big base, so it's not like we're talking about a percentage increase that is meaningless. A lot of traffic, huge amount of traffic coming out of Google Organic.

And ? It's not like we're dealing with a team that, that are kind of inexperienced, but just to get that structural piece right, so important to then have very quick wins.

Joe Alder: Yeah. Yeah, definitely.

James Lawrence: So if we kind of move on through the framework, so S for structure and reliability. T?

Joe Alder: Yeah, so , this is another interesting one that I feel like goes very, , like quiet in the AI landscape. Like not many people are talking about it, and I find it's one of the most important factors.

So you have kind of visibility, and that is are you ranking when someone asks, "What's the best SEO agency," right? Like, is Rocket appearing? But . we basically did a study where we looked at millions of real user prompts, and we looked to understand how are people actually using ChatGPT and these AI bots.

There's basically no information from any of the tools, so we kinda had to do our own data scrape here to get that. And what we found was one of the really big use cases was business research. So people that have actually maybe further along the funnel and they are looking, like, "Should I use Rocket Agency for SEO?

Is Rocket Agency a good agency? How does Rocket compare to one of our competitors?" And we found a lot of people were searching this way. Like, "Should I use X business? Would you recommend X business?" So that is why the second framework is like your trust, truth, and sentiment. So how are AI tools actually, , objectively understanding who you are as a brand, how you offer your product, would they recommend you in the space?

So we've asked, like we've had some clients where they've come on board and we've literally asked ChatGPT, "Would you recommend this brand?" And it has just responded, "No," and then an answer. So if we know that a huge amount of Australians are going to these platforms to make these purchasing decisions, if you're then asking, "Should I use X brand?"

And it's saying no, that's gonna have a massive impact on your conversions.

James Lawrence: , And it's really challenging, right? Like we started off the conversation talking about traffic. Traffic increasing, traffic decreasing, because that is the metric along with conversions that for the last 25 years we've, as digital marketers working with in-house marketing teams, have obsessed upon.

I think it would, it would be interesting if we can just talk a little bit about how that's kind of flipped now with AI engines, that we can't just rely upon traffic to be the measure of success. Um, the, I think the stat that you talked about in the webinar was ChatGPT's got like a, is it 1.5% click-through rate from a prompt or something like that?

Joe Alder: Yeah, it's super low. Yeah, it's around 1%.

James Lawrence: So , we'll get to measurement and reporting shortly, but I, I think as a marketer, I think it can be challenging to be shaping, reshaping perspective in-house, where for, most marketing environments are used to having a dashboard or monthly reporting, and they hand it up the food chain or deal with different stakeholders, and it's always been if the, you know, if the graph's growing, we're succeeding, happy days kind of thing, where the nature of these AI chatbots, the nature of Gemini, the nature of AI Overviews, ChatGPT, we go back and forth, we go back and forth.

We don't then kinda click out on the, within the actual platforms themselves. Like, the, the user interface within ChatGPT is really not designed to kinda go out onto the open web and browse commercial websites. So... And we'll get to kind of measurement and sentiment, but think just making the point that it's not the volume of traffic coming out of these platforms which is an indicator of success or failure.

Often it's what's being said about you that you need to be measuring. So Joe, I think if we could talk just a little bit, I guess, about how do we measure the sentiment piece?

Joe Alder: Yeah, sure. So we have a few deliverables at least that we do for clients, um, depending on it. So like, in my head at the moment I'm thinking of one of these banks that we have been working on.

I've been working on that quite a bit recently, and basically the way that we have done it obviously you need the tracking piece that we're gonna get into. We have asked a number of prompts, and we've asked them across all the different engines and a number of different times. So basically, breaking it down, like how are they currently perceiving this bank?

And that is per subcategory as well. Like, would you recommend them , to use as your mortgage? Uh, would you recommend them for a savings account? Like going through all of their categories and seeing if they're coming up as recommended. 'Cause again, this is what users are gonna be searching for. The other really interesting one is like then looking at their information that is being cited.

So for this example, for this bank, when I was looking at it, one of their interest rates was coming back around 1% higher than was actually listed because AI was using outdated information. And like something like an interest rate for a mortgage, that is the reason you would or would not go with that provider.

So it can have like catastrophic events, if it's like really incorrect and users are seeing this incorrect information. Interestingly as well, we've been working with a few other clients in the kind of legal financial space, and they've actually received letters from APRA and ASIC, which are the industry bodies basically stating that you are responsible for the information these AI pull about you.

So that's kind of the third bit is the compliance, like making sure the information we have that is being retrieved is compliant and it's something that's really hard to control obviously, but there is ways to reverse engineer where this information is being cited from and then going back and, yeah, counteracting it.

James Lawrence: I think we're finding it's, , it's kind of a throwback to the early days of SEO, right? And, and not in a totally spammy sense, but there, I think there is first mover advantage at the moment. Where back in the day, , SEO used to kind of be, you know, white text, white background then kind of Google caught onto that and then it became the 500 to 700-word blog article with keyword stuffing, but it kind of worked for a period, right?

But at the moment, there is an ability to go in there and shape the content that these large language models are pulling and returning for your brand. Um, what advice would you have as to the best way to I guess get like what are the, you know, the top 10 large language models saying about their brand?

And I hate to kinda keep putting the boot into Optus, but we're kinda using them as the, negative case study in a webinar that we ran last month where there was just like ChatGPT was ruthless and Gemini was ruthless when you were doing comparisons between mobile plans and mobile reliability between the major telcos.

Joe Alder: Yeah, I think basically negative mentions with around 72% of the answers that were being cited when we were asking about Optus. So it shows like the, yeah, the impact that this can have. And I guess my recommendation here is really to have that reporting and measurement piece sorted and making sure you are tracking.

James Lawrence: So just, so just for reference, the STAR framework S for structure and retrievability, T is trust, truth and sentiment. We'll come back to A. Yeah. But this is R, so reporting and measurement

Joe Alder: Yeah.

So reporting and measurement and, um, I guess I'll come back to the sentiment stuff, has been one of the biggest challenge. One of the things I was saying before, right, like there's 10 different AI tools. , Even if we did have this information that's accurate it would be spread out across 10 different places, but we basically do not have any information at all.

Like ChatGPT has no way of getting this. I was hoping with the launch of ChatGPT Ads it would start giving us some information, but it's Yeah, even the ad platform is just so basic. , Google Search Console gives you very light data on AI overviews, but it's not very strong. So the way that we do it at Rocket is using that independent study that we did, we have a framework of how people are searching in ChatGPT and the other tools based on the real user prompts that we've mined.

And then we use, uh, traditional SEO keyword research search volume, and then mine it towards the structure and prompt structure. So if you're a business, I would recommend doing kind of long tail, what is the best X? Could be what is the best SEO agency? What is the best bank for, , my mortgage?

Like those type of terms, like people really search using best a lot. Then comparing to your direct competitors. So it might be Optus versus Telstra for mobile plans, and that you're not looking at being cited, right? You wanna actually look at the answer and see are they recommending Optus over Telstra?

Because again, these are the type of searches people are gonna do. I would then also ask, and this is just kind of on that brand piece, I would also ask "Give me a CIA-level breakdown of everything that you know about all of my brands and entities," and even do that as subcategory level. Like, for Optus mobile plan, tell me everything that's in your training data, and that should give you a bit of an idea of the information it has, at least in its stored data or where it's pulling those sources from and then obviously you wanna move on to the kind of more broad tracking terms.

I'd recommend tracking at least 100 terms, and this is where it gets challenging if you're not using a tool. But because of how much AI is personalized and how much it differs, you kind of need to ask the same question minimum 10 times, but ideally around 100 times per month to get an idea and then get an average of where are the key sources, what is my average visibility for the best mobile plan, et cetera, et cetera.

And then we add it up. We use the data that GA gives us, and that kind of gives us a reporting and structure. So specifically on the brand and sentiment we have created a sentiment score where we kind of look at the number of mentions, how many are positive, how many are neutral, how many negative, and come up with a score.

A lot of these AI tracking platforms have similar, so I would look at that as well as negative mentions. And then the other two key metrics that I see for AI visibility, one is visibility percentage. So out of the prompts that we ask, however many times we ask them, what percent are we actually being retrieved in the answer?

And you wanna be careful here to exclude any branded terms. So those brand terms we spoke about earlier, make sure you're not including them in your visibility score. I've seen quite a lot of the AI platforms do that, and it's obviously gonna be very skewed in your favor. You'll always be cited in those brand ones.

And then the third metric outside of the traditional more Google Analytics stuff is share of voice. So out of all of the prompts and answers, what is your share of voice versus your direct competitors? If you have 100%, are we currently getting around 20% of those searches and maybe our competitor is getting 30?

Um, so yeah, they're the three kind of metrics I'd look at. Mm.

James Lawrence: And it's been fascinating, like at Rocket we generally try to use ourselves as the guinea pig and and I'm involved in Rocket's marketing and how we're perceived out there. ... We get a monthly report of how we're appearing, right, within all the different large language models, and it's kind of Rocket versus a, small number of agencies that we think are a similar competitor set.

Would you recommend Rocket? Uh, what, what do you think Rocket is? What is it not? What's it good at? And it's so fascinating- Hearing from an, from one of the large language models what its perception of us is, and it's allowed us to actually shape that over time, right? - Our head office is in Sydney.

We've got a little satellite office in Melbourne to allow us to serve Melbourne clients. It was kinda saying we're a Melbourne-based agency and Melbourne founded, and we found out that was through one of kind of the aggregator websites that had kind of pulled data and manipulated it incorrectly. So we're able to then go back to that source, change it.

It was pulling some really wacky content around pricing and how we approach pricing for clients, and that, that was from an old blog that I'd written kinda years ago. So we're able to jump in and change the original source of that. And we have a saying, which nothing, you know, more br- broader than the, the conversation we're having now, which is that . That which gets measured gets moved. And so actually starting to have data coming in has allowed us to kinda manipulate and change what the large language models are returning on us. So it's definitely not this, , theoretical impossible to move thing. It can be changed, but it is a, it's a case of actually measuring it and getting that data coming in in the first instance, right, Joe?

Joe Alder: Yeah. Yeah, I actually have another really interesting example of that and, like, where we've actually changed the opinion. We work with, , one of Australia's biggest construction brands, like very much household name, and when we were asking about, , the sentiment, would you recommend them, they have really good reviews on I think Trustpilot is like their kind of main review platform.

They're like 4.8 stars and thousands of reviews. And when we were asking, like, these AI tools, "Would you recommend?" It was coming back there's a lot of negative reviews, um, which like looking at the data I was like, "That's confusing." I looked at it and it was citing from product review where they had, I think, three or four reviews versus thousands, and those three or four were negative.

So we just got the client to direct their review strategy to product review for the next few weeks, got that up to 100, and now basically it's recommending them again. So although they're very smart, like these AI models, because of how they work and where they retrieve the information, it's kind of like they use sometimes the wrong sources- Mm-hmm

and there's no way to control that, so you kinda have to move their eye away from it.

James Lawrence: Can you talk just a little bit more around- What metrics and how in-house marketers should be looking at in terms of how their business is performing in this landscape?

Joe Alder: Yeah, sure. So first is, um, share of voice. So that is your kind of share of voice versus direct competitors.

Um, one interesting one which may be a bit more granular is citation share. So especially for these brand-specific questions, and that is what percent of the time is the AI tool citing information from your site or at least an owned site. So an owned site could be Wikipedia, it could be Yellow Pages, because that's something you have direct control over versus an external source is much harder to control.

Then your kind of overall sentiment score. So out of a score of 100, what is AI sentiment about you? So for Optus it was like 20% the sentiment score. Uh, for Rocket it's some 90% or higher, right? So it's like a huge difference and it's a really good factor to just see how it's being perceived.

Um, the other thing is just looking at your kind of AI traffic and coming into GA4, but I think this is a good like- Not warning, but , I guess something to be aware of. As we discussed before, 1% click-through rate from AI tools. So what you see in GA4 is such a small snippet, and I think it's why it's so important to have a tracking methodology outside of Google Analytics and also not to just listen to Google Analytics.

I've spoken to clients in the past, and they have said, "We don't need a GEO strategy because there's only 1% of our traffic, and then if we go in there, it's not recommending them." Yeah. It obviously is having a much bigger impact.

James Lawrence: . We're using a- an iceberg analogy within the agency when we, we speak to clients or prospective clients, which is traffic is at just the little tip of the iceberg as it relates to the impact that AI search GEO is having on Australian brands.

And depending on the vertical, B2B, B2C, we estimate that now in Australia, 20 to 30% of buying decisions are being influenced by AI, and that's not a number that we've kind of just plucked, you know, out of thin air. We're looking at decreases in traffic, but conversion rates staying the same. Joe and his team have done a lot of work at looking at the impact on branded search and direct search, where when we saw the rise of Facebook coming through kind of 15 years ago, you would see this effect of big investment in Facebook campaigns driving huge amounts of branded search and direct search through Google to, to cl- to websites, and became pretty clear two devices, see people seeing something on one device, clicking onto the next.

If someone's skeptical on this, just get them to use ChatGPT and try to get them to go and purchase a product or click on a link. Sites don't load properly. It's just not a user experience that works. People are using large language models as the research tool, and then they're converting, you know, through desktop, through mobile, using Google So if you're only looking at traffic, you're gonna be - severely underestimating the influence that large language models are having on your business.

So definitely in that - 20 to 30% range, they're real now. They're not peripheral. Trying to build up a measurement framework that isn't just automated. Don't just look at last click data in your analytics. Start speaking to your clients. Start, you know, maybe getting some heard of us data on contact forms, checkout forms, just to start building a bit of a profile 'cause

there's a big gap at the moment between the hard-nosed measurement piece and what the real world activity is. Um, so I think it's really important. Joe, just in terms of practically, like how can listeners measure this stuff? 'Cause you're not getting the data in GA. Google Search Console is not giving much.

ChatGPT is giving nothing. Practical advice on this?

Joe Alder: Yeah. So there's a number of different AI tracking tools available at the moment. They basically all do very much the same thing. Um, some of them are quite expensive as well. , It's because of how many kind of engines they need to search and how they need to search and how often, like there can be like a few of them are a couple of thousand dollars per month.

So there are definitely cheaper options available. There's not many in the Australian space anymore, but the kind of big market leaders here are Peak, Profound Promptwatch. I think Promptwatch is pretty good and reasonably priced, um, so is Peak. I think Profound is quite expensive. Semrush also, I would say unless you're an enterprise, is probably not worth it.

I think they all very much do similar things. But the feedback that we've had, well, we've had a number of clients that kind of already have this in practice. I think the big gap that they, all of these platforms have at the moment is actually using this data. So they will give you all of those metrics and you'll need...

you'll know how you're being perceived, but it's kind of where they stop. There isn't much in terms of recommendations or actions. I'm sure that will change. Um, but yeah, we have had clients that have struggled to actually use the data. Even spoke to someone who owned one of these tracking tools and they said a lot of businesses are kind of setting it up and then not really going in there since.

So if you have an agency, it'd be good to get their advice and work with them on it.

James Lawrence: And , what are we doing for our clients? I mean, I, I know, but , more for the purposes of listeners to kind of understand what we're doing.

Joe Alder: Yeah, so we've built a number of different custom tools where we kind of thread it into our own reporting methodology.

So we use one of these tracks prompt tools, but then we also have a number of different layers on top. So we have our kind of own independent research, so we know how people are searching in different areas. We have a way to then turn that into prompts. Then we have our own like independent tools we've created that look for compliance issues, incorrect information, sentiment, and that is again, just because a lot of these tools don't offer it at the moment.

Like you can do it if you export all the raw responses and go through manually or try and use like Claude to go through and look for it, but it is quite challenging. So we found there's a bit of a gap and we had to create our own kind of tool set here.

James Lawrence: Hmm. And I do- I think we never shill for Rocket on the podcast, but if you are genuinely stuck with measurement and understanding What's happening to your business within AI and large language models.

We are doing a little bit of standalone work. So for clients where we're not doing service layer where in terms of actual optimization ongoing, we do have a, a small product where we are allowing clients to plug in just for their measurement piece, just 'cause it is so difficult and fast-moving, and I do feel like we've, um, we've got a, a really solid offering, which is particularly good for Australian businesses.

I think if you are stuck, feel free to just kind of come through the Rocket website and we can look at how we could potentially just feed in that measurement piece, um, even if you're not engaged with us to do the ongoing optimization . I think that's really good, Joe, from the reporting and measurement piece side.

So now in terms of the STAR framework, A for answers.

Joe Alder: Yeah, sure. So answers and find out is basically the equivalent to your content strategy. So how AI works is if you ask a question, so what is the best mobile phone provider? ChatGPT and the other tools then break that out into 10, even up to like 80 different sub-searches, where they basically run agents to do equivalent of a Google search in their platforms.

So it could be, what is the best mobile phone plan if you live in Sydney? What is the best mobile phone plan with unlimited data? What is the best reviewed plan? So you kind of need to understand what are those questions that these AI tools are asking, and then figure out, do you have content on your site or on external sites that actually answer these questions?

So yeah, very much like the equivalent of keyword research in the SEO days, um, but more prompt research, and then also understanding what content you need. And again, this isn't just the content on your own site. We have no control over where these AI tools are going. So, like, link building has been a huge thing in the kind of SEO landscape, and I don't think it is redundant now.

If anything, it's more important, and it's not for the same reasons of getting a link. But if you get your information cited on these larger websites, like what is the best mobile phone plan? And if you rank number one there, it's gonna encourage AI tools to cite your brand.

James Lawrence: And- in Your view, what are the biggest levers?

. Like what are the things that really move the dial versus the kind of nice-to-haves that probably don't have a huge impact in this space?

Joe Alder: So number one, like technical, if you're blocking AI bots or have any technical issues, you're gonna be like dead from an AI search perspective.

So that has to be right before any of the other items. Outside of that, I think it would be very much external sources and how they're talking about you, where your brand is being listed. I would put that quite high on the list. Then in terms of your content itself is really important, and sometimes you will see the AI first goes to an external website, but then it will validate on your own business.

Like it wants to make sure if it's coming up as Rocket is one of the best SEO agencies, do Rocket even offer SEO? So they almost like self-validate, so you need to make sure you have it. Sometimes AI will go directly to your site as well, so you want content on your own site, content externally, and you want to make sure that anything that is being searched, you have addressed on one of these platforms.

I think that is, yeah, by far the biggest lever. Then outside of that, like kind of what we were speaking about, brand and sentiment, like how are you perceived online is really important, especially because the number of times people are asking best- Mm ... like what determines best? It's reviews, it's um, awards, it's sentiment.

Those things also do help. Definitely not easy to do, but like number of reviews, number of awards that you have.

James Lawrence: I think definitely pointing out to listeners that be very careful- Basing large language models perceptions of your brand on your own queries. There's so much data within your account, within ChatGPT, et cetera, that you're not gonna necessarily get the same returns that an average potential customer in your space will get.

So you need to solve that within your business 'cause you might find that you're getting all these awesome positive feedback, but it, it knows you're an employee there or it knows you work in marketing or it knows you're the CEO. So you've kinda gotta be really careful to not fall into that trap.

We've seen that a few times with kind of providing the initial sentiment surveys to clients. They're like, "No, no, that's not what I'm getting on site." We go, "Yeah, you're what you're seeing is n- not what the, the ordinary Australian is seeing."

Joe Alder: Yeah. ChatGPT loves to please the user and it's led to a few awkward situations.

Uh-

James Lawrence: Great question, James. It really does. Um,

I think, Joe, like as we kinda look to wrap things up, . Any other kind of closing thoughts or things that you, you think in-house marketers need to be thinking of when they're approaching this space?

Joe Alder: Uh, I think from speaking to a number of businesses in this space, often I see it framed as SEO or GEO. , Like we need to cancel SEO and put our money to GEO. We're not doing a GEO product. And I kind of said it before, um, how it really sets a framework, but I think like there's been some studies on it where 77% of users are using both tools together and they really work hand in hand.

Yeah, SEO still drives a huge amount of traffic, so does AI. There's 33% of people start on Google and then they will go to AI tools to basically be like, "Is this correct? Should I use this brand?" 68% would use AI, and kind of your point previously, like the click-through rate is not a great experience.

They will then go to Google to then finalize their searches, and that's why we're seeing brand. So yeah, not thinking of AI as just kind of the GEO strategy, but broader picture. But then I guess to the opposite of that is making sure you have a strategy for both because as part of SEO, we are not looking at your brand sentiment.

We're not looking to see if there's incorrect information. It's not a real core part of the SEO offering. So really make sure you have a very clear offering that's addressing both and using them hand in hand together.

James Lawrence: Yeah. It's so interesting, isn't it? I think like two years ago the narrative - was ChatGPT is growing and Google's dying or going to die.

And I think we've... I'd like to think that most business people, most marketers are now at that point of going, it's not one or the other, it's one and the other. And the last Q1 results showed Google search growing 19%. ChatGPT is still growing, or the other large language models are growing faster than it now, Claude, et cetera.

But it definitely doesn't feel like we're not in this monolith kind of, um, situation like we once were, right? We've got a whole bunch of different players and the, the landscape is more fragmented. Joe thank you so much for coming back onto the podcast. I really hope that for the listeners of this episode, that you've all got at least one practical thing you can take away, hopefully more.

And, um, yeah, look forward to having you back again sometime , in the not too distant future, Joe.

Joe Alder: Cool. Thanks James.

James Lawrence: Thanks mate.