Episode 53: From SEO to GEO: Winning the Shift to AI-Default Search

Are AI search citations really driving business outcomes, or are they just a vanity metric?

In this episode of Digital Clinic, Aaron Burnett sits down with Taylor Hurff, Director of Digital Strategy at Wheelhouse DMG. Together, they unpack Generative Engine Optimization (GEO) in healthcare and medtech, moving beyond simple citation counts to focus on brand sentiment and real visibility. Taylor shares practical ways to diagnose traffic drops, build compliant AI workflows, and adapt as search engines shift to AI-default modes.

Listen & Subscribe:

Introducing Taylor Hurff

Aaron Burnett: Ask most people optimizing for AI search what success looks like, and they’ll likely point to citations showing up in that little list of sources under an AI answer. Today’s guest thinks that’s pretty close to irrelevant: he’d rather have zero citations and a brand mention people actually remember than a link nobody clicks.

Taylor Hurff, Director of Digital Strategy here at Wheelhouse Digital Marketing Group, has spent the last eighteen months figuring out what actually moves people from an AI answer to a website. In this conversation, Taylor breaks down why brand sentiment beats citation count, and what that means for how you should be spending your GEO budget.
I think you’ll really enjoy this conversation. Let’s get to it.

This podcast is sponsored by Wheelhouse Digital Marketing Group. Wheelhouse provides exceptional performance marketing for healthcare and medical device manufacturers. Every Wheelhouse client saw record performance in 2025, even after implementing HIPAA-compliant data solutions. Find out more at wheelhousedmg.com.

Why AI Search Matters

Aaron Burnett: All right, so we’ll talk about the topic that’s been in the zeitgeist every day, every week, every month for the last three years, and only more so today, tomorrow, and the next day. Let’s talk about LLMs, AI optimization, GEO, choose your acronym. We’ll get there, but first, maybe you could give us a thumbnail sketch of your background in SEO, and then how you’ve come to be focused, and considerably expert, in the AI aspect of optimization.

Taylor Hurff: Happy to.

From Content Writer to SEO Strategist

Taylor Hurff: At the start of my career (this was a brief beginning), I actually started as a content writer out of college, writing content for search. The more I did that, the more I was dying to know why I was doing it. What was the impact? What was the reason behind it?

So I started learning more and more about SEO, and I liked that a lot more than actually writing content. Pretty quickly into that job, I started down the SEO path (learning, asking questions) and ultimately found myself at another agency with a focused role on SEO strategy, and I loved that.

I’ve always been a math person, so I really liked the numbers behind it, playing the algorithm, things like that. But I also just liked having an impact and helping people at the end of the day. I’ve been in SEO for nine years now, and it’s never a dull moment. There are always algorithm updates, changes in the SERPs, and the launch of all these AI platforms is no different.

I’m honestly having a lot of fun unpacking how we show up in these AI results, how they impact businesses, and how we can do something that’s genuinely good for people while also doing what we need to do to show up in AI results. That’s been my passion for the last few years, and it’s been a lot of fun testing new things out. I don’t like when things get boring, and especially with the launch of AI, it is never a boring day.

Building Wheelhouse’s AI Workflows and Monitoring Tools

Aaron Burnett: And here you play a multifaceted role. You’re Director of Digital Strategy, but you’ve also led development of a lot of our AI workflows and tool sets, which gives you an interesting, and more technical, perspective on AI generally, and then the application of AI in search. Tell me about the work you’ve done to develop AI for Wheelhouse.

Taylor Hurff: Certainly. There’s a lot we use AI for, and really, we’re not trying to replace the things we do with AI; we’re not trying to replace any expertise. We’re trying to get data into the hands of our strategists quicker, so they can be smart, capable people and make good decisions.

A lot of what we’re doing is connecting different APIs and MCPs to all the data sources our strategists would go to anyway. Instead of them having to log in, go to reports and dashboards and filters and manually pull data, we’re able to get that into their hands incredibly quickly, and then they can interact with it: be smart, intelligent people, and make good decisions for their clients.

One I really enjoy: we have a landing page monitor that’s constantly looking at performance data across all these different dimensions and segments, helping us identify where the weak links are, where we have to optimize. If I were doing that manually, it would take hours to drill down into every facet of that performance data. We’re able to do it almost instantly with AI, and immediately flag: okay, these campaigns and segments aren’t matching with the landing page: let’s go do something about it.

But these always-on monitoring systems are the ones I like the most, because it’s not just doing work we would have done quicker; it’s actually letting us do way more than we ever could, and go deeper, without taking expansive amounts of time.

Aaron Burnett: I think that’s a great example.

A Privacy-First AI Setup for Healthcare Compliance

Aaron Burnett: I think one important point to make here is that we’re in healthcare, where data privacy is sacrosanct, really critical. So the way we’ve implemented and use AI workflows is different than might be conventional. We’re doing things that go far beyond chatting with Claude or chatting with ChatGPT.

We’ve built workflows (you’ve helped build workflows) where we have specific skills employed in specific projects, and we’ve connected the AI applications, the LLMs we’re using, to our data warehouse and other data sources in a way that lets us both orchestrate activity automatically, and enable, as you said, our expert strategists to query through those workflows and get access to rigorous analysis, informed by data in our data warehouse and other platforms, much more quickly and efficiently than they could individually.

Taylor Hurff: Oh, yeah, 100 percent. One of the things you touched on: we never feed data back to the LLMs when we’re using it for our jobs. We have our own private instance of Claude that we lean on pretty heavily for these AI workflows, and we also have clean data in our data warehouse that we’re able to connect into our Claude, so we can use data that’s fully compliant.

Whereas a lot of agencies are just doing very fancy prompt engineering, most of our AI systems don’t require a complex prompt. It’s very detailed instructions and context files that tell it exactly what data to get and exactly how to parse it. It’s not prompt engineering; that is so far from what we do. And it’s all compliant, because we only use clean data connected to our own private instance. There’s no sharing back with platforms, no PHI being shared. It’s very clean, the way we do it.

Aaron Burnett: That’s great.

The GEO Landscape in Healthcare and MedTech

Aaron Burnett: All right, so that’s a good foundation to begin talking about generative engine optimization, or any other three-letter acronym you want to apply to AI. Again, we’re in healthcare, med tech: there are specific constraints, interesting dynamics around search generally, and even more so around the introduction and centrality of AI in search. Give us a sense of the landscape of GEO in healthcare, med tech, and other privacy-first industries.

Taylor Hurff: It definitely depends on your actual product or services, but in a lot of healthcare, trust is still paramount, and trust is important for LLMs, but it’s also really important for users at the end of the day. People aren’t quick to make a decision about their healthcare, or a medical device, without doing deeper research. AI platforms are heavily used in those kinds of decisions. You’re not just going to sign up for some critical medical device on a whim. It’s not a one-search, one-prompt scenario: you really have to be there as they’re doing their research, because these AIs are research tools at the end of the day.

We hear a lot of businesses say they want to show up in AI, but what does that really mean? We see a lot of people talking about citations: ‘oh, you want to be cited by AI sources.’ That’s good, but a lot of this research spans days, and people aren’t clicking all the links mentioned in a citation. It’s good that they’re referencing your content, but that’s not going to drive people to your business. A more impactful metric is a brand mention: getting your brand mentioned when they’re researching is way more impactful than having a link in the sources section of ChatGPT. And again, these are still leading metrics: what you really want is for people to ultimately make it to your website, learn about your products, fill out a form, or purchase something.

But a lot of the people I hear talking about GEO treat citations as the holy grail. Really, I’d rather have no citations but have my brand mentioned, and mentioned positively, so that people, as they continue their research, go Google our brand and come to our website.

Another fallacy is that clicks and visits from AI sources are the most important thing you can have. What we see more often is that when we’re getting brand mentions, and it’s very positive about our product and our brand, we actually start seeing direct traffic go up. People visiting the site directly, or our branded search traffic, goes up, and the quality of that traffic goes up too, because people have already done their research. So citations are great, LLMs coming to your site to train is great, but really you have to be looking at whether you’re being mentioned and recommended as a brand, whether that’s positive, and whether it’s better than how it’s talking about your competitors. Because that’s really what’s going to separate you: that’s what’s going to stick in the mind of a consumer, and what’s going to get them to your site to learn more.

Is GEO Just SEO With Extra Steps?

Aaron Burnett: All right, let me ask the big question that’s debated daily on the internet: is GEO, AIO, is it just SEO with a couple more bells and whistles, or is this a fundamentally different discipline from SEO?

Taylor Hurff: That’s a great question, because there’s a lot of overlap between SEO and GEO; I wouldn’t say it’s exactly the same. There are different facets of GEO. One is: are the models training on your data? Are they able to find your information, and are they using it in their training? One thing we do here to measure that is server log analysis, because these platforms use different bots to crawl and train on your information. Looking at that, to make sure your content is accessible and can be picked up by these LLMs, is very important, and that’s not necessarily an SEO thing. They don’t necessarily abide by the same systems Google Bot does when it’s crawling and indexing your content, so you have to look deeply to see if they’re able to find your information.

It’s also important that it’s not just about what’s on your website. SEO is very heavily about what you can control on your website to get more links, more clicks. LLMs aren’t just training on your website: they’re looking at third parties, aggregators, review sites, even Reddit and what people are talking about there. For GEO, it’s much more about how you show up on the web holistically: how people are talking about you on Reddit, the sentiment they have, how other sites are referencing you. It’s much broader than traditional SEO. It touches into the outreach that was so important when backlinks were critical in SEO, but it’s not about backlinks: it’s about sentiment, mentions, how you’re showing up online holistically.

Aaron Burnett: Yeah, the outreach associated with GEO has much more of the flavor of digital PR than anything like link building.

Taylor Hurff: Exactly. The SEO outreach was almost entirely link building; there’s some, you know, a listicle that shows up for a lot of people that you want to be part of, but it’s much more about digital PR essentially.

Diagnosing Whether AI Is Behind a Traffic Drop

Aaron Burnett: All right, so you mentioned server log analysis, which is a good point of transition: we did another episode on server log analysis, and we’ll link to that in the show notes. Let’s talk about how you diagnose what’s actually happening related to AI overviews and AI in search results. Site owners, webmasters, digital marketers are noting what they feel are precipitous drops in organic traffic, or shifts in traffic, in their own analytics. And the worry is, ‘that’s AI eating my lunch: my educational content is now being cited in LLMs or incorporated into AI overviews, so people aren’t coming to my site, they’re not consuming my content.’ So how do you diagnose what’s going on, and are we actually seeing that?

Taylor Hurff: Diagnosis is a good place to start: it makes you think about what data sources you have and what you’re tracking. Just like any drop in organic, even before AI, there could be a number of causes: a ranking issue, an algorithm update, competitors with genuinely better content. The first thing I’d do if I’m seeing a decline that I think could be AI is try to pinpoint when it started. First, check whether there was a major algorithm update. You should also keep a close eye on the competitive landscape: are new SERP features showing up, new AI overviews, or are competitors just creating better content and getting more clicks than you are? If there’s a clear pinpoint (a single day, or a drastic drop over the course of a week), it’s likely an algorithm update or broad AI overviews now showing. So you really have to look at the SERPs over time.

Aaron Burnett: All right, so let’s say you’ve looked at what’s happening in the search landscape, and there’s no clear moment where everything dropped: it didn’t drop twenty-five or thirty percent from Monday to Tuesday. You want to identify, or prove or disprove, that AI is now cannibalizing your traffic. How do you go about that analysis?

Taylor Hurff: What I’d look at first is impressions over time, because it could easily be seasonality, or it could be that people are slowly migrating to AI tools. You want to check whether impressions are steady but clicks are going down and rankings are going down, or there’s a new AI overview. Those indicate something’s happening in the SERPs. Or, if things are pretty steady but volume is steadily decreasing over time, that could indicate people are shifting over to things like ChatGPT or Claude.

I’d also look at your server logs. If you’re an authority in your space, the big LLM platforms are coming to your site: they’re training, but they’re also requesting information when people prompt. They have separate bots that come when a user asks a query, it searches results, and it comes to your site to pull in content. If you’re seeing that go up as your organic impressions go down, it’s likely a shift to AI: consumers using AI more frequently, and it’s more important that you show up there. If you’re not seeing those server logs change (they’re not hitting you anymore), it could be that you’re not well optimized, they don’t trust you, they don’t want to come to your site. But it could also just be that people are searching for your products and services less.

If you’re not showing up much, if they’re not coming to you for information: we know AI platforms aren’t slowing down, so if you don’t have an AI GEO strategy today, you should have one, and at least start getting that data in. Start being a source if and when people go to LLMs. That’s only going to give you more data to make confident decisions about whether this is a trend people are shifting toward, and whether you need to focus more there.

People still use Google almost as much as they did before: I believe Google just reported having more active users than ever this past month. So Google isn’t going away, but they’re incorporating AI more than ever, increasingly pushing people to AI Mode, and Google AI Overviews are showing up more than ever. You have to look at how you’re showing up for queries that show AI Overviews. In Google Search Console, if the impact you’re seeing is on ‘how to’ and ‘what is’ queries, things that clearly show an AI overview, that’s important, and it tells you that you really have to be showing up in that answer.

It also begs a strategy shift. We saw this when featured snippets started rolling out: it was good to own a featured snippet, but a lot of times that answered the user’s question and they’d go do another search, or end their search entirely. I’d rather have a number-two-ranking page with no featured snippet than a featured snippet for something that totally answers a question and they don’t need to click in. So if your whole strategy was built around early-upper-funnel, easy question-and-answer content, that’s not really the strategy that’s going to win in AI. It’ll get them through that step quicker, but that’s why you have to be there when they start searching for brands and solutions, and make sure you’re showing up in a positive light.

Aaron Burnett: Got it, okay.

Building a Monitoring Program: Sentiment, Citations, and Prompt Volume

Aaron Burnett: So maybe that’s the answer to this next question: let’s say you’ve gone through the diagnosis, you see the presence of both flavors of bots, the training bot and the bot that answers questions live, and you have a significant drop in organic traffic. What should a digital marketer’s response be? Should you think, ‘I need to recapture that traffic and bring people back to my site,’ or lean into the shift and become more effective in the LLMs, and think of your site differently?

Taylor Hurff: It’s both, but your first step should be figuring out what people are searching in these AI platforms, and what those answers are. Too many people I see are flying blind, following different frameworks and best practices for how to optimize for AI without really knowing what consumers are searching and what the AI platforms are showing as answers.

The first step should be: we set up very comprehensive monitoring programs where we’re monitoring tens of thousands of prompts across all these platforms, so we have a really good idea of what’s getting shown when consumers search an AI platform for our products, services, or even our brand by name, and where that information is getting pulled from. You can’t optimize what you can’t measure. So your first step should be to get high-fidelity measurement of the AI space. Then you can start changing your content strategy to fill the gaps, or change the sentiment. But I wouldn’t just fly blind and stand up an llms.txt file and hope that solves all your issues.

And I also wouldn’t ignore Google Search entirely. You might want to pivot away from some simple question-and-answer queries, so that you’re not just answering a question and having them leave, but you have to think holistically about your strategy moving forward, because these LLMs aren’t going away. It’s also more difficult to get that deep audience insight about what they’re searching and what the AI platforms are showing as the answer. But that’s the new normal: figuring out what your audience is using AI for, what you should be showing up for, and then having a good understanding of what the answer is, why, and how you can influence it to better show off your brand and get people to learn about you on your website.

Aaron Burnett: You mentioned a couple of important things there: that we have significant monitoring programs for LLMs, and that digital marketers should figure out what their prospects are searching for and what they should be showing up for. Talk about how we approach monitoring, how we figure that out, and how a digital marketer beginning to grapple with this should set up or obtain that kind of monitoring for themselves.

Taylor Hurff: I’ll get into how we identify the prompts, but we’re doing sentiment analysis, citation analysis, and a number of other things: really tracking over time how you’re showing up, what brands are showing up, what content is being cited. It’s important to do it that way, and not through one-off ChatGPT searches asking, ‘why am I not showing up?’ Because, as everyone listening knows, the AI answers are different every time you prompt. They’re similar (we do see a lot of similarity over time), but a one-off question and answer is likely to miss the mark of what people are actually searching, and it’s only going to reflect your answer in that moment in time. Things evolve, the AI answers shift slightly, even the citations change frequently, so you need a larger-scale program.

As for what prompts we select: there are a few ways we do that. Prompts are getting longer, more conversational, more question-based, instead of just Googling ‘healthcare systems near me.’ They’re asking, ‘what are the most trusted healthcare systems near me?’ or ‘what are the best systems for X, Y, and Z kind of service?’ And all of that is showing up in Google Search Console. People are using Google similarly to how they use these LLM platforms, because Google is turning into one itself. So I’d start there: it’s easily accessible data for anyone who’s done SEO, and you can start seeing the types of prompts people are using in Google for the AI Overview or Google AI Mode. That’s one way to start getting at the answers.

The other is classic audience research, still one of the most valuable things any brand can do: speak directly to your consumers, talk to them, do panel studies, do actual surveys. You can get a lot of information that way. We also, through the platform we monitor on, get direct data from ChatGPT itself, so we can do query investigation: essentially keyword research using actual ChatGPT data, so we can see prompt volumes over time and see what people are actually asking, and what we should be monitoring closely. It’s not difficult to get a confident list of questions: if they search this, or something similar, we want to know what the answer is. Is it mentioning us? Is it framing us in a positive light? That’s clearly the first step before you actually change anything on your site.

Aaron Burnett: I know the landscape for diagnostic data, the things you just mentioned, Google Search Console or other monitoring platforms, changes almost as quickly as the AI landscape itself. There have been rapid evolutions, particularly over the last few months. Do you foresee an era, a point in time, where we’ve got super clean, comprehensive data related to the LLM search landscape, the queries behind it, who’s being served, similar to what we’ve got for SEO?

Taylor Hurff: That’s interesting, because one, I’d say we already have a ton of data: it’s about collecting it and making it usable. The server logs, all the query data, the questions in Search Console, it’s out there. It’s about making it usable and bringing it together. But at the same time, because of the nature of using ChatGPT or Claude or another platform, everyone’s search query is different, so I don’t know that you’ll ever have super clean data. Google, for the longest time, has done recommended and suggested searches, taking these unclean prompts and funneling them toward something it thinks you’re trying to ask. We’re not seeing much of that with ChatGPT, Claude, and these other platforms: it feels very unclean, very chaotic, for most people right now. You can pull this data together in a much more usable way today, but I don’t know that it’ll ever be as clean as Google Search Console: that’s a very systematic, aggregated way to look at things. I don’t know that AI will ever get there, but I do know that if you harness the data available now, you can already get very actionable information. It’s just never going to be as clean as Google Search Console, in my opinion, but who knows, ChatGPT could launch a ‘ChatGPT Search Console’ and change things overnight. It all moves so fast.

Aaron Burnett: Yeah. I think even with Search Console (although it’s ostensibly clean data, presented in a relatively user-friendly fashion), it still suffers from being curated by Google, from Google deciding what data you get access to, and the thresholds for that data. They seem to be capricious at times. More than that, they’re a hundred percent self-serving, so it’s problematic to have a single source of data in any event.

Taylor Hurff: Yeah, that’s the least hot take you could have said: they are very self-serving. The more you drill down, the more it filters out data, to the point where it becomes unusable at times. But it’s still better than trusting some keyword research tool online: it’s definitely better to get that Search Console data.

The Three-to-Five-Year Outlook for GEO, SEO, and Google’s AI Mode

Aaron Burnett: All right, let’s think three, four, five years out. What does the digital landscape, we’ll call it the search landscape comprehensively, that’s search, that’s LLMs, what does it look like? Are we still optimizing for traffic generation to a website, is that a core KPI along with search rank, or are we optimizing for something different as digital marketers?

Taylor Hurff: When I think back to three years ago, when ChatGPT was launching, we were all coming up with our takes on whether Google would go away, whether everyone would move to ChatGPT, or whether ChatGPT was a fad. A lot of us were in the middle, and I feel like it settled out a little more toward Google’s favor than people thought: a lot of people expected a mass exodus that didn’t really happen. But it’s steadily moving toward LLMs being increasingly important.

If I had to predict three to five years out, I’d say GEO is going to be just as important as every other aspect of your media: that includes paid, that includes organic. I think GEO and SEO are going to merge together, because right now, the things that work in GEO and no longer work in SEO (certain trust signals ChatGPT hasn’t caught up to the Google algorithm on yet), I think they’re going to fix that. I think they’re going to cite more reputable sources, have stricter parameters around what they cite. I’d be shocked if people still have SEO ranked more important than GEO in five years.
I also think websites are going to stay important. People are so used to visiting a website that they’re going to do a lot of their research through ChatGPT, and they might eventually be able to get directly to a contact form from ChatGPT: fill it out directly, no website needed at all. I don’t know that we’re there yet. I feel like the research phase will increasingly happen through these LLMs, but ultimately you’re still going to want to go to the website, fill out the form, purchase a product. But I could be wrong: it’s more likely that LLMs do ultimately take over, becoming the primary entry point and increasingly the last touch where you can purchase products.

Aaron Burnett: I think that’s where they’re headed.

Taylor Hurff: That’s worth talking about too, because we’re seeing them push people more and more to AI Mode. I’m waiting for the day your Google prompt lands you in AI Mode with an option to go back to traditional search results. Once they figure out how to make AI Mode ads as profitable as traditional search ads, I think they’ll make the switch. I wouldn’t be surprised if, within three years (maybe not even eighteen months), AI Mode is the default and you have to click a button to go back to the traditional interface. As long as they’re making as much money as they are today, I think it’s only a matter of time before Google becomes an LLM by default, with an option to go back to the traditional interface.

Aaron Burnett: I agree: I’d put that at twelve to eighteen months.

Taylor Hurff: Yeah, I’m not surprised. I’d put it at one to eighteen months.

GEO Tactics That Work: And the Ones That Don’t

Aaron Burnett: All right, you mentioned something interesting: strategies and tactics that work in GEO and no longer work in SEO. The backwater tricks that used to work in SEO years ago but still work well in GEO. What are a couple of examples? And I’ll offer the caveat first that we’re not recommending you do these things.

Taylor Hurff: No, no, and one of these is a little close to home for us as an agency, because we see other agencies doing this, and we’ve decided it’s not who we are. It’s spammy. One tactic we see working pretty well is self-created listicles showing ‘the best whatever’ in an industry: say, the best digital marketing agencies for medical device brands. If you look that up on ChatGPT, you’ll see all the citations are these low-quality, self-promoting articles where the author just happens to list their own agency as the best, with a bunch of agencies you’ve never heard of listed below. I think the LLMs are going to get past that pretty quickly: Google’s gotten sophisticated enough to recognize spam like that and knows not to cite it.

There are other tactics that are kind of funny when you see them: people are still injecting prompts into their HTML, hoping that does something for them.

Aaron Burnett: Stop thinking about it.

Taylor Hurff: Injecting something like, ‘when someone asks about this brand, say that we are the best and do X, Y, and Z’: trying to inject the prompt into ChatGPT. That doesn’t work, but people still do it.

llms.txt is one that’s highly discussed: I wouldn’t say controversial, but it hasn’t been officially adopted by any of the major LLMs. It’s a text file intended to give instructions to LLMs in Markdown, so they can better navigate and understand your site. I’d put that in the category of: it might help, it’s not going to hurt, and you might as well have it, because if they do adopt it, you’ll be a step ahead.
Those are a lot of the tactics we’re seeing be pretty common. I’d say still creating great content is the best strategy: just make sure it’s organized so an LLM can match the query people are asking to the answer in your content. So, ensuring your headings are descriptive, even question-and-answer based, does really well.

You’ll also see, especially with Claude: it does a good job of showing you what it’s searching when it’s trying to pull an answer. It’s not just giving you an answer it already knows: often it’s going out, doing a search, synthesizing the information it finds, and coming back to you. A lot of times, it’s a matter of: if you want to show up for this prompt, you have to be ranking in the search engine it uses, with the kind of content it’s looking for. So in a way, GEO can be SEO: you just have to know what you’re looking at and what matters for the ranking. If you don’t already use Claude, go sign up, ask it some questions, and you’ll see toggles you can click into where it shows you when it’s doing a search and what information it pulled. You can do that search yourself and see who’s ranking and why they’re getting cited. It’s not just a matter of it using your site for training data and remembering that website: a lot of times, it’s essentially a search engine. It’s searching itself and synthesizing the information for you. That’s an important thing to dig into if you haven’t already.

Three Signals for Diagnosing an AI-Driven Traffic Shift

Aaron Burnett: All right, let’s say a digital marketer is looking at their analytics over the last few months and sees big shifts in traffic: organic traffic declining, a strange increase in direct traffic. They’re concerned. What are the first three things they should do to diagnose whether this is a shift to AI superseding organic traffic to their site? How should they respond?

Taylor Hurff: I’d say before even digging into direct, you should look at referral. Not all AI traffic comes through referral, but a lot of it does, and Google just launched a new custom audience group that buckets most of the major LLMs into an AI channel. But a lot of it still comes through referral, so if you’re getting clicks from AI, check your referral traffic.

A lot of it does end up coming through as direct, because not all the referrer information comes through when someone clicks a link in AI. It’s also worth knowing that AI Overview clicks don’t come through an AI channel; they come through organic as well. So more likely, it’s an AI Overview answering someone’s question and they’re not clicking through.

But if your organic traffic is coming down and you’re seeing an increase in direct traffic (and not just direct, but high-quality direct traffic), that’s a good chance it’s people who’ve done their research and are coming to your site already knowing about you, already knowing what they’re looking for, and much more likely to convert. At that point, I’d look at your server logs, see what content is driving those clicks, and start monitoring prompts to try to increase that share as much as possible.

If you’re seeing organic traffic tank or decrease steadily, and you’re not seeing it come in through other channels, it’s probably AI: people turning to ChatGPT and not finding your brand. You need to go through what we talked about before: identifying prompts, getting the data, monitoring, optimizing. If it’s coming through direct or referral, and it’s highly qualified traffic, double down: keep expanding, because it’s not going to stop. People are going to continue using LLMs.

Aaron Burnett: I’ll parrot back to you what you said to me: I think that’s the least hot take you could have offered. People will continue to…

Taylor Hurff: Yeah, I agree.

Aaron Burnett: But I think it’s absolutely right. This isn’t going to abate: it’s only going to accelerate and increase. All right, what have we not talked about that you’d like to talk about?

The Takeaway: Focus on Brand Mentions, Sentiment, and Accuracy

Taylor Hurff: We’ve hit on what works and what doesn’t, and how to stand up a monitoring program. I’d just double down on this: it’s not enough to say ‘I want to do GEO’ or ‘I want to optimize for AI.’ You really have to know what you’re optimizing for, and what matters and what doesn’t.

Citations can be important, but only if people are inclined to click on them: otherwise it’s just an answer, and you get a small brand impression. I wouldn’t broadly optimize for citations, or stand up tons of educational content just to get the citation, because at that point people are just learning: they’re not ready to act. I’d focus most of your effort on brand mentions and brand sentiment. It’s much more important to show up when someone searches ‘what is the best device for X problem’ than for much earlier-funnel content where they’ll see your brand, not know about you, and not really be looking for solutions yet. See how the AI talks about your brand compared to your competitors, and see what you can do to influence that.

Especially for our medical device clients, we go even further: when people are really deeply diving into our brand, what are the patient outcomes associated with it, what’s the fail rate, that kind of thing. We want to make sure it’s incredibly accurate and portrays us in a positive light. There’s a lot of misinformation out there, a lot of outdated information that gets pulled into LLM results. So I’d focus heavily on brand mentions, brand sentiment, and brand accuracy, over broadly chasing citations or clicks. Focus on clicks to content that’s actually going to move the needle and drive business outcomes. You’re not optimizing for traffic: you’re optimizing for whatever business outcome you’re aiming for. Don’t trust broad citation measurement or broad click measurement; make sure it’s focused on the things that are actually going to move the needle.

Aaron Burnett: Yeah, I think that’s a universal truth: optimize for a business outcome, not a vanity metric, regardless of whether you’re working in SEO, GEO, or digital advertising. That’s great. I’ve enjoyed the conversation; I appreciate the time, Taylor.

Taylor Hurff: Thank you, happy to talk.

Aaron Burnett: As Taylor pointed out today, navigating the shift from traditional SEO to AI search, or GEO, isn’t about tricks, prompt injection, or chasing broad citation numbers. It’s about high-fidelity measurement, building brand sentiment, and optimizing for real business outcomes. Whether it’s one month from now, eighteen months from now, or two years from now, AI search will become the default mode in Google Search. And the companies that have optimized their digital footprint for LLMs will have a significant advantage.

If you enjoyed today’s discussion, be sure to subscribe to the podcast, leave us a review, and share this episode with anyone on your team navigating the new world of AI search. I’m Aaron Burnett, and I’ll see you next time on The Digital Clinic.

Sponsored by Wheelhouse DMG