Episode 25: How Google’s Regulatory Battles and AI Are Redefining the Future of Search

David Mihm, partner at Near Media and renowned local SEO expert, shares his perspective on how Google’s antitrust and regulatory challenges, coupled with the rise of AI, are reshaping search marketing. With insights from his research at Near Media, David offers a unique view into the diversification of search, the evolution of user behavior in healthcare, and what an AI-powered future means for digital marketers.

Listen & Subscribe:

The Digital Markets Act and Its Impact on Google

Aaron Burnett: Tell me a little bit about the Digital Markets Act (DMA), its construct, its implications for search overall, and I’m interested in implications for Google in particular.

David Mihm: Yeah, totally. So the DMA, I think came out in 2022. There were a series of sort of hearings or workshops and then September, 2023, I believe that the law crystallized and said, “Okay, now these are what are called gatekeepers.” The law applies to, at the time, six, now seven, very large multinational companies. It’s all the big tech platforms, Google, Apple, Amazon, Meta, Microsoft, ByteDance. And then a seventh one is actually the first European gatekeeper, Booking.com, was designated a gatekeeper just last year.

This Digital Markets Act came into effect in 2023. Companies had until March of 2024 to demonstrate their compliance with the law. It covers everything from app stores to iOS and Android platforms to browser choice to default search engine choices, specifically as it relates to our business search. The law only applies to companies with tens of billions of dollars in revenue, over 10,000 business users in Europe and over, I don’t know, 50 million European users, like consumer users. It’s like a very small set of companies. As it relates to search, Google is subject to the Digital Markets Act.

The provision that’s most interesting to me is Article 6:5, which basically says they cannot preference their own services and products if a third-party offers a similar service or product with respect to crawling, indexing, ranking, right? And they have to present these results in a fair and non-discriminatory manner. That’s very interesting to me as a local search practitioner because Google Business Profiles containing basically all first-party data, right? Photography that’s uploaded by Google users, business categories that they’re soliciting directly from the business, reviews that they’re soliciting from Google users.

This is all first-party data. The primacy of the local pack in basically every local search result is very interesting because there are plenty of businesses that are excluded from the local pack who offer very similar business profile services. Think of Yelp or Yellow Pages companies, or TripAdvisor. Any large directory offers very much a competing product, and they are not eligible to appear in Google local results. Article 6:5, I think, is going to have serious teeth as it relates to local search once the European Commission gets around to investigating that specific vertical.

Currently there are three active investigations, shopping, hotels, and flights where Google clearly is showing its own products. Google Shopping, Google Flights, Google Hotels, very prominently when there are plenty of other competing players in Europe. There’s been a longstanding lawsuit, I think it was like a seven or eight yearlong lawsuit specific to shopping that Google actually finally lost last year at the European Supreme Court, and they ended up having to pay a multi-billion dollar fine and that there’s. The outcome of that case I think has maybe given regulators a little bit stronger leg to stand on in terms of the legal sort of foundation for applying penalties to Google for this exact kind of behavior, this self-dealing, self-preferencing behavior.

Google’s Self-Preferencing in Search Results

Aaron: The model that you use at Near Media is a very interesting one. It’s a combination of your consulting expertise along with your partners and primary research that you conduct. You’ve done a good bit of primary research around whether in fact Google is satisfying the stipulations, the Digital Markets Act. Yeah, for sure. And what’s your conclusion?

David: Yeah, so to cut to the chase, our conclusion is that they’re not complying with the Digital Markets Act, which is a shocker because self-referencing. Yeah. There, so there’s all kinds of studies out there that you can get just with any basic rank tracking software to identify one of the new elements. So Google introduced three new SERP elements, a new chip at the top of every search result page that has local intent, which is called places, sites, or comparison sites. Now that users can click on, similar to news or shopping or images or maps. There’s a new tab for places, sites for local queries. We found almost no engagement with that tab. Something like, I don’t even know, 20 out of 700 and some clicks in our most recent study across Germany, France and Ireland. That’s the first introduction.

The second is a carousel of URLs from places, sites. These are directories like I said, Yelp, TripAdvisor, those kinds of sites. A carousel of URLs from those sites. The third is an aggregator carousel, where any one of those individual sites, so if yelp.com is ranking fourth or fifth in a traditional search result, they can then have a carousel within the yelp.com organic listing of individual businesses on yelp.com. That requires a certain level of JSON-LD markup, but for any competent company with a significant development team that’s not that difficult to implement. We don’t see a lot of that in the United States interestingly enough, but I wonder how much of that is due to lack of awareness of this specific markup or if it’s something that Google is actually just not showing. The reason I’m curious is because we found a) those things don’t show up very often, even in Europe and b) very few users click on them. The thing that we focused on most in our user research is this places sites carousel, which includes multiple aggregators in a prominent place on the search results. This is where, getting back to the rank tracking question, any rank tracking software is going to tell you like, oh this element exists in the first position and the local PAC exists in the second position, sometimes they’re flipped. But generally speaking, Google I think, could argue that they are presenting this places sites module prominently.

In our data, we’ve found that it’s not quite as prominent as Google Business Profiles are, but like basically 50/50 in terms of which one occupies the top spot. The issue is of course, that the map interface that accompanies Google Business Profile is so compelling. There are these five gold stars alongside every business and you have little review snippets and you have photos of these businesses. None of those things are necessarily true of those places, sites modules. Nobody is clicking on them. This to me gets at the fair and transparent and non-discriminatory parts of the provision 6:5 given that this is not, that Google is not presenting results from these competitors in the same way that they’re presenting results from their own product and service.

I think that is a clear example of self-preference. And Google’s one of the smartest companies in the world, they’ve probably done their own user research. Sure. And they’re saying, oh, let’s put this out there because it’s going to look like we’re compliant. But the reality is, we know we’re still going to capture the same clicks that we were always getting for Google Business Profiles. That’s basically exactly what we found. That typically if a user is on a SERP, where both the places, sites, carousel and Google Business Profiles are present, they will click the Google Business Profiles interface about 40-45% of the time, they’ll click the places sites result less than 2% of the time. It’s 20x the engagement that Google is keeping for itself on Google Business Profiles, even if it appears directly below this place’s, sites carousel.

David: The one exception we found in that behavior actually is in hotels. The reason is because there are so many ads in hotels, many of which have their own very compelling interface that includes a photo and reviews and pricing information that users don’t scroll far enough down the page to see either the places, sites carousel or the Google Business Profiles hotel finder. The only way that Google’s dominance of the local SERP with its own Google Business Profiles product, the only place we don’t see that is when Google’s collecting all the ad revenue from these hotel carousel ads.

How User Research Reveals SERP Engagement Patterns

Aaron: Tell me a little bit about how you go about conducting the user research that you do it.

David: Yeah, totally. We crowdsource a panel of real human beings. These are not bots. This is not done through ChatGPT. We source a panel of real human beings who meet our client’s sort of demographic desires. We recruit them to essentially a user testing platform. We present them with a prompt. We say for example, you’ve become dissatisfied with your primary care provider and you’re looking for a new one. You go to Google to help you find a new provider. We’ll drop them on the google.com homepage at both mobile and desktop.

We pay them for their time, and it’s very clear like they’re sharing their screen and their audio. We’re watching them do their searches and narrate their experience. And hearing what they’re reacting to, and then we’re aggregating, we’re annotating their entire video and aggregating all of the clicks they make and the attributes of the results that they’re clicking on. What was the star rating? What position was it in? What was the photo? Was there a tagline showing up? Was there a review snippet? All of those things that relate to Google Business Profiles and organic search, kind of the same thing.

For example, product review schema on a legal website query, which aren’t supposed to show up in an organic result, and by golly, those things still work really well. Then I think even more so, we aggregate all that data and we present it back to our clients and we say, here’s essentially a unique click-through rate curve for your specific vertical you’re interested in, across both Google Business Profiles and organic search and paid as well. One of the really interesting findings we’ve done for clients or come across for clients is when LSAs are present, just how dramatic local service ads. How dramatically they impact click through rate and choice of provider or choice of business. Google must be making a fortune on those, to be perfectly honest. It’s been really interesting. We aggregate all of the sort of quantitative data together. Yeah. Give clients a view of what that looks like.

Some of the most interesting findings that we’ve come up with have been around messaging. What messaging resonates most with clients.? Is there a particular type of photo, for example, that they really want to see? The examples we found in OBGYN was one of the sites that, that people clicked on had a medical stirrups on the service page for a checkup. Women absolutely were like, “Oh my God, this is this is horrible. How can you possibly have that on this page?” Things like that, that are just like, depending on this sort of acuity of how people react. Some things are really actionable for messaging and photography and taglines and use of reviews and all these kinds of things depending on the vertical. And I think that’s the really interesting thing that we’ve found is. Yeah, there’s a certain set of things that is pretty consistent from medical to legal, to restaurants, to hotels, to home services.

There’s really different behavior depending on the vertical, just for one example. Lots and lots of clicks on the map that shows up above Google Business Profiles for home services. This was a kitchen remodeling prompt. You’re remodeling your kitchen. You need help from a professional, you do a search on Google. People really wanted, at least in the EU, they really wanted to go to a showroom and see what the kitchens looked like. Huge percentage of people clicked on the map, and so ranking in that context becomes totally irrelevant. It’s all about. Are you somewhere in the person’s local area? And if you are, what’s the first impression that your business profile is putting out there?

So what are your stars, essentially, that people are using stars, heuristically as a disqualifier. What are the sort of one or two photos that show up first on your Google Business Profile? A lot of these, a lot of the businesses anyway, clearly had not claimed their profile. These were user photos, they were poorly lit, they were grainy, that sort of thing. People would disqualify businesses from the map which in a category where I would’ve felt like, eh, this guy’s coming to my house, or This company’s coming to my house. I don’t necessarily care where they’re located. I do want to see their before and after projects. That would’ve been my sort of default mindset going in. But now it turns out ranking less is important. Presentation is more important. So that’s just one example of the sort of really interesting sort of behavioral findings that we’ve come up with in just one category in just one market.

Aaron: I would imagine as well, I mean our experience with user testing in a different context, but similar approaches that hearing the voice of a customer saying repeatedly, “This was horrible or I got stuck here, or I like this much better,” is so much more compelling than aggregated data.

David: Yeah I couldn’t agree more. We certainly have played clips for our panel for clients. In this case though, I think what’s I, and I don’t know how you guys have done user testing historically, but I do think that there’s value to we use panels of several hundred folks. Which I don’t think traditional user testing typically, like you wouldn’t probably have 300 people usually looking at your own website. We’re essentially showing folks, “Here’s why you might be missing a click.” Google Analytics is showing you what clicks you got, and Google Search Console might be showing you some of the keywords, that people were using when they found you. But what about the clicks you’re missing? Why aren’t you getting the click even if you’re in the first position? There is some sort of quantitative value, I think to having a large enough panel to deliver that with some. Statistical probability in terms of why you might be losing a collection.

Aaron: I absolutely agree. I didn’t mean to discount the value of the quantitative result. More that our experience is that the quantitative data particularly for people at a C-level who don’t necessarily understand right. Intricacies of search or user experience, it’s much easier for them to breeze by a stat or a number than it is the sound of a customer who is frustrated. Yeah. In particular, we have at times created highlight reels of, here are exactly 16 customers encountering exactly the same problem. Yeah. Yeah. And that. That’s incontrovertible. Yeah. In a way that the data was not Right, for some reason. Yeah.

David: Yeah. Not everyone is as data driven as you and I are probably.

Aaron: That’s fascinating. I can see why that would be really valuable.

David: Yeah. It’s been super fun to work. This is very time intensive. We have dabbled with finding AI video reviewers essentially that can, but we the intricacies of collecting all the data points that we need to collect. And the value of us actually hearing. Customers narrate these experiences. Like it just takes a lot of time to, to do this. I think the output has been really valuable for our client.

Aaron: You mentioned a couple of things. One is directories and aggregators. Historically, there were aggregators, horizontal platforms that were very important in terms of local search, and that through your research you have found that is generally not the case any longer. That this is very much a verticalized space. There are directories that matter for a particular vertical. They may or may not have weight outside of that vertical, but that’s the way to think about and focus on optimization for a locality or a vertical.

David: Yeah, so when I first got started in local search, this notion of a citation, which has a new meaning in the world of AI, but at the time essentially a mention of a business in within structured format where Google could identify, oh, this is the business name, this is the address, this is the phone number. Maybe there’s a mention of the website, but not necessary, right? The more of those structured mentions that you had across the web the higher you ranked in the original Google Business Profile results, Google Places results. This was before the Venice algorithm update of 2012 and where Google artificially inflated the rankings of local businesses ahead of very strong national websites. The organic part of the local search results was all of these directories. It was Yelp City Search, Super Pages, Yellow Pages. I mean that you could basically say that the same 10 to 12 domains ranked for basically every local keyword across the entire United States. And that is no longer the case.

Again, the Venice updates are, we’re now seeing plenty of localized organic results for the last 12 years now flowing into to local search. Also, I think we’ve seen the growth of vertical platform. Yelp essentially now the only horizontal directory left standing with any kind of significant presence in organic search. There are certainly very strong vertical players like Houzz in home remodeling and Trip Advisor in travel and that sort of thing. And we do see users engaging with directories on a regular basis in, but they’re all these vertical directories. Yelp has had very low engagement in all of our user research.

But people in the healthcare vertical, they’ll go to Zocdoc, they’ll go to Healthgrades. You find the horizontal directories have really faded away, in terms of their organic search rankings and consequently the benefit of being mentioned on one of these sites. These sites are no longer as prominent as they used to be, so the value of getting mentioned on them has also gone down. Meanwhile, the value of getting mentioned prominently on a strong vertical directory, I would say gone up. If you are one of the top ranked businesses on Healthgrades and MediFind and Vitals and all these other sites. Your Google Business Profile for that doctor probably also has a pretty significant boost. So the same concept applies, but the sites that Google values from a organic ranking standpoint have definitely changed dramatically in the last 12 to 15 years.

Healthcare Systems and the “Find a Doctor” Challenge

Aaron: You’ve made another good point related to that research as well, which is a lot of healthcare systems have a Find a Doctor function that is so poor that they’re just handing business to the aggregators. They really are. And I think you made a very shrewd point, which is, “Look, you should be looking at Zocdoc or one of these platforms that is tuned and tested or find a doc functionality and yours should mimic that because you already know that works.”

David: Exactly, yeah. Zocdoc specifically, I don’t know anyone at Zocdoc. I’ve never worked for Zocdoc. So this is not a, necessarily an endorsement, but just based on our user research. Sure. The people loved using that website to find a provider. They loved the sort of very fast, faceted search at the top of every sort of search result that they landed on. They loved the reviews. Generally, Zocdoc has, I would say, better photography of providers than most other medical sites. In particular, they do a good job of highlighting the insurance plans that each doctor takes. Now, how accurate that is, sure. I can’t vouch for, but certainly consumers are engaging with that.

I think there’s a lot to be learned from Zocdoc’s user experience for sure. If you are building or revamping your own find a doctor interface. I think the biggest thing. The biggest error that I see health systems making with Find a Doctor is either that it’s only a search interface, so there’s not a corresponding sort of clickable directory to providers by specialty and location, which is essential for SEO. You can’t just say, “Oh, we’ve got a search feature and we’ve got an XML site map, we’re done, our SEO’s good.” That’s not how it works. Then the second thing is hosting your Find a Doctor solution on a subdomain. A subdomain is still treated as a quasi-independent website, so you’re not any of the value of any of the equity, the site authority that your main domain is built up doesn’t necessarily transfer to doctor profiles hosted on “doctors.healthcareorg.”

Aaron: Yeah. I can tell you in real life we’ve rebuilt Find a Doc functionality for some large health systems taking very strong cues from Zocdoc, and it works very well.

David: I’m sure it does. Like I said, the consumer engagement with that site was noticeably stronger than any other aggregator in any individual Find a Doc system that we found.

Google’s Antitrust Issues

Aaron: So sticking with the regulatory environment, let’s shift to the world of Google in the United States. There have been a few very interesting things that have happened. One is the antitrust finding, finding that Google is monopolistic, and the second sort of corollary is the testimony that was given to the Department of Justice, combined with a leak of some Google documentation. Let’s kind of start there with the Google documentation and DOJ testimony. What did we learn out of that corpus of information that was a surprise and or a point of validation around something that Google had, I’ll use the word “lied,” about for a long time.

David: I think that’s an appropriate term, particularly given the testimony. The thing that was validating to pretty much every long time SEO is that Google absolutely uses click through data right from Google search results as a meaningful lever for organic rankings. And that’s been a core part of their algorithm for a very long time. That was very clear in both the API league as well as the DOJ testimony and discovery documentation. I don’t think that the behavior surprised me, but the degree to which Chrome browser data is leveraged for search rankings seemed to be a surprise based on the number of references in the API documentation. So we don’t know, there’s no like weak variables assigned within the API leak, but certainly the number of references to Chrome user data was interesting. And that also came out in the DOJ testimony as well.

So I would say those were the two, if you were a Google Kool-Aid drinker. You may have been totally shocked by the first finding, the first release in terms of the click-through rate, meaning making meaningful change in rankings. But I think that those of us who long looked at Google with a little skeptical eye were not surprised, but definitely validated.

Aaron: The implications then of the remedy for the antitrust finding are also pretty profound. One of the things that’s being explored is divestiture of Chrome, which we now understand to be the core signal for Google.

David: At the very least I would think Google would or should be forced to license its Chrome data to competitors like DuckDuckGo or Ecosia or even Bing. I think that is clearly such a strong competitive advantage and monopolistic advantage for Google. That that would be to do anything short of that, to ignore the value of Chrome, I think would be a mistake. Absolutely divestiture Chrome, I think is on the table.

There’s certainly very interesting ramifications if Google is forced to divest Google Maps. I don’t think that’s part of the DOJ case necessarily, but in terms of antitrust remedies, generally mapping platforms are very expensive to maintain. They have been very expensive to build. So, there is a question whether a standalone maps application would even be financially viable. I’m not a legal expert, I’m not a business analyst at that scale. I can’t speak to whether that’s viable or not, but I do think that’s one thing. I think that Chrome, it strikes me like given that Mozilla has survived for so long as a right, very strong browser alternative as a, essentially a nonprofit. I don’t think there’s the same sort of financial considerations with divesting Chrome as something like Google Maps would be.

Potential Regulatory Outcomes for Google

Aaron: Considering both of these environments, Europe and the DMA and the US in this antitrust finding, I guess two questions. What do you think will happen with Google and what should happen with Google?

David: I think very little will happen here at the United States. It’s somewhat helpful that the decision is in the judicial system at the moment in terms of what to do about remedies as opposed to legislative, where I think absolutely nothing is going to happen for the next four years at least. I don’t think on the regulatory side of things, I think I’d be very surprised if the new FTC chair pursues antitrust policies with the same gusto and vigor that Lena Kahn did right in the previous administration. I don’t see much happening here at the sort of federal level, aside from the current case that’s already in, I think, judge Meta’s lab in terms of deciding what to do so right.

In Europe I am less clear about what’s going to happen. I think Reuters broke the story that the EU is like strongly considering implementing damages in this initial investigation, in this initial DMA self-preferencing investigation. So that story came out on a Friday and on a Saturday morning the White House put out a statement about, “You better not regulate our tech companies, Europe. We’re going to put tariffs on you.” Jim Jordan put out a response who I guess is now in the Commerce Department saying something like, “You need to explain yourselves why you’re regulating American tech companies.” I can’t say what’s going to happen in Europe. If the investigation had happened a year ago, I think I’d be more sure that there would be fines and penalties imposed on Google. Less clear that’s going to happen, just given the changed political environment. What should happen, I absolutely think Google’s a monopoly. I think that they should be. Essentially we’ve put out a policy statement Near Media about the DMA and we think that Google should be subject to publicly verifiable, clickstream sharing, essentially not for any given query. I don’t think it would impact, our jobs as SEOs. But just, “Hey, this is how many queries we’re getting for these.” Kind a Google Trends dashboard essentially, and “Here’s how many we’re sending to Google Business Profiles and to individual hotel suppliers and to aggregators from various interfaces or modules on the SERP.” That kind of public transparency, I think, would be a good starting point to help people understand just how people are engaging with Google search results and to verify that if they are complying with this law, either that trend is going downwards at a minimum or has disappeared entirely. That would be the starting point for what should happen with regulating Google is a publicly verifiable data set about what modules people are clicking on.

AI’s Current and Future Impact on Search Behavior

Aaron: Yeah, that makes a ton of sense to me as well. Let’s talk about AI and the implications of AI in search. What, in your research or in your practice, are you seeing as the implications, the incursion of AI into the search domain?

David: Yeah, for sure. So obviously, I’m monitoring, as you guys are as well. You guys have a great analytics team. We work together on one client, but we’re certainly not seeing a significant dent in Google’s market share currently. I think, all of the Generative AI products combined are still under 1% for most of my clients anyway. Google’s still 85-90%, if not higher for all organic search traffic. And those numbers are staying. I’ve been surprised at how steady they have stayed. For organic search, even with the onslaught of AI Overviews, particularly in the healthcare space the click, the organic search performance hasn’t dipped as much as I might have predicted seeing those things show up.

Having said that, one of the key sort of pieces of advice I’ve been giving my clients is just that we can’t necessarily equate click-throughs from ChatGPT or Perplexity or anything else as a substitute for how many people are using those products because you’re getting just full funnel experiences in a single ChatGPT session or a single Gemini session for the percentage user that are using something else where you can ask ChatGPT a very high funnel question, “Tell me about knee replacement surgery.” A very broad query, but one that has a lot of search volume at least by traditional keyword measurement tools. ChatGPT will give you the full, “Here’s everything involved in knee surgery. Here’s the recovery time, typical recovery time, here are the side effects, here’s what you need to ask your doctor,” all those kinds of things. That’s all content. That would’ve been probably 5, 6, 7, 8 search queries. Historically, that’s now all in a single session, ChatGPT, with no clicks out, typically, to a website. The thing that I think is actually hopeful about AI in search as it relates to businesses with brick-and-mortar service delivery models or local service delivery models, is that I’ve noticed ChatGPT will prompt me with an explicit follow up at the end of almost every initial sort of higher funnel chat, “Hey, do you want recommendations for providers or companies or whatever who do X in your area?” They will almost invariably ask that at the bottom of one of these higher funnel searches, and those are quick, like knee replacement surgery.

Any mom-and-pop orthopedic practice is going to have a very hard time ranking for that term against Mayo Clinic and Johns Hopkins and the Cleveland Clinic and Penn Medicine. It’s a very strong set of national competitors. That’s not a term that Google would typically localize where you’d see a Google Business Profiles result. I think there’s a chance that local companies actually get a bigger chance at the pie, Generative AI is going to hurt these big national global content producers, because I do think it’s going to be answering a lot more of the questions that people are typing in. But I think when it comes to actually delivering a medical service or going to a coffee shop or what have you, I think that those, I think the GenAI products might actually deliver more real-world traffic than the current search paradigm does.

Aaron: Yeah. I think that’s probably right. What’s curious to me is the difference, as you suggested, when we look at data, we don’t see much incursion into conventional search. But when you look at attitudinal research or consumer preference research, there’s a very strong preference for using AI over conventional search. I wonder like you, if the nature of the queries are still there, the nature of the query has changed, and it’s now a query that’s much further down the funnel than it was before.

David: I think that might be true. Users may be using GenAI to do their initial research. Then they still would probably be conducting follow up brand queries, for example, in Google or Bing or whatever their search engine is. They’d probably still be interacting with a website based on that brand query or a Google Business Profile based on that brand query. To your point, I think that maybe we’re going to just see a shift in search behavior. Again, for a typical sort of regionally focused medical organization or a typical local service business, I’m not sure that I see ChatGPT or AI Overviews as the same existential threat that a site what was it? Chegg just sued Google for, stealing traffic or something, so I’m not sure I would see it as the same existential threat for my business.

Adapting SEO Strategy for the AI Era

Aaron: What adjustments, if any, are you making to search strategy, local strategy to ensure that you’re hedging your bets in both directions?

David: I’m not sure this is necessarily an adjustment, but a sort of shifting in strategy and really trying to help clients understand, “Look, we, you have to be explicit about this service is available at this location and these providers practice at this location and offer this service,” and having a very strong structured data set behind the scenes that is powering not just your Find a Doctor tool. The content page on your orthopedics silo that’s talking about knee replacement surgery, you need to be very explicit about these are the places where you can get your knee replaced and these are the providers that are delivering that service. Then that will help you show up hopefully at the bottom of the sort of ChatGPT response where they’re prompting users, “Do you want help with this problem locally?”

That kind of activity I think has been important for organic search and I think to some extent for GBP results for quite a while. But I do think that level of data fidelity is, has been a struggle for healthcare organizations, I think, to get their hands around. At least with the ones that I work with. I think it’s now becoming even more important than it was before. I think one interesting thing that I think maybe will change moving forward, it hasn’t yet again given because Google is still the 800-pound gorilla, but site architecture has always been something I’ve been super focused on. Making sure that your internal link graph is aligning to show your most popular pages. The pages where you want to sell stuff as actually having the most internal links, large language models may be less dependent on links as a sort of dominant ranking signal. And it might just be more about making sure that you’re mentioning things like you are the number one, US News Hospital in Seattle or whatever. It may be less about link graphs and more about mentioned graphs and being very consistent with your messaging about what makes you special, why patients should choose you, that sort of stuff.

Aaron: Yeah, and then this gets into the notion of building a brand and building a brand with the right descriptors in as many places as possible not on your site.

David: Exactly, and understanding what those sites are. To some extent, if a language model is trained on data that’s two years old, you might not have that much of a chance to influence it. But increasingly, paying attention to the media sources that are cited by ChatGPT and AI Overviews and adjusting your PR strategy to target the sources that the sort of RAG models are pulling into the initial LLM results.

Aaron: I think that’s right. And then I think, coming back to one of the things you mentioned early on, which is the use of schema, make all the machines jobs easier, as easy as possible.

David: That’s exactly right. Even if your content management system doesn’t fully support all of the various schema elements that you, we as SEOs would like it to.

Even looking at schema to understand how the topics and the concepts in your space can be structured right, I think would be a really useful, interesting starting point for your content teams and your website product teams to understand how all these topics and content should be fitting together. Is there a specific protocol treatment for a given medical condition, for example, that Schema has already identified as a series of things it expects to see structured data on, or is available to be structured in schema, I think is a really interesting way to approach your content.

Aaron: There’s obviously a significant diversity of opinion on the actions that people should and shouldn’t take in the face of the rise of AI on one end of the continuum. There’s a faction that would say seed, the SEO game, build your brand as much as possible. You’re not going to rank for informational queries that will be subsumed by AI or other aggregators, that sort of thing. And then others say, particularly in verticalized niches, it’s still worth doing that. You just have to be very careful about your focus and the structure of the content. Where do you fall on that in particular?

David: I’m definitely in the second camp. I’ve never worked with a large enough, Meredith brand publisher or something like that, where they could instantly rank for a sort of high-level keyword with their content. I’ve never worked with a site that could just do that. I’ve always advised clients on targeting the terms that are actually achievable to rank for. Is Google showing some evidence of localizing terms in this related to this topic? Or is the national competition just missing this on when they talk about a particular condition or whatever. And it’s like trying to identify those gaps. So I don’t think the strategy’s necessarily changed. But I also don’t think that seeding the SEO game is like necessarily a good strategy. I think you should still be trying to be one of the sites that is pulled in as authoritative for the AI overview, for example, or making sure that you are mentioned as a source in Chad, GPT as a domain expert in this particular topic.

Even if that doesn’t lead to a click, I think again, to your point about brand, like you still want your brand to be out there as being an expert in this particular thing. I think you should still be very much paying attention to search visibility. In fact, I would say maybe visibility is even more important than it has ever been, but I don’t know that traffic is really the right way to measure that anymore. And that’s something that I don’t have a good answer right now. What is the right way to measure visibility? I think AI Overviews, may be a little more straightforward, but from my own anecdotal research, you have a very different set of results in ChatGPT if you’re logged into the product for basically every query. If I do a search from Portland and you do the same search here in Seattle, logged into ChatGPT, you get a very different set of results. And it’s not the same set of results that the API returns. So all of these studies that we’re seeing now about, here’s our brand visibility metric based on the ChatGPT API that has no bearing in what actual people are seeing if they’re logged into ChatGPT, which I think is going to be a real challenge for measuring visibility.

Aaron: It’s the same quandary that’s existed in local forever.

David: Yeah, exactly. I think there’s always been more sort of variability in Google Business Profile results than organic results since as long as I can remember. So you’re right. It’s just turning that dial up on, “Okay, we need to be taking a broader cross section of terms and also looking at a broader cross section of locations when we’re doing these searches to understand how we’re appearing.”

Aaron: Yeah. We have been actively exploring monitoring applications and haven’t found anything that I’m willing to pull the trigger on.

David: No, absolutely. I think it’s still I think at this point, it’s good to be aware that this is a problem, but I don’t think that the solution has really come about yet. I think, honestly, I think it’s going to have to be something to do with just like crowdsourced rankings through Mechanical Turk or something like that. And even then, I think it’s going to be very difficult to say, “Oh is that a representative set of people? What have they put into their ChatGPT sort of query stream historically? What does the application know about them? And does that match your own audience?” So that may I’m not sure. I’m not sure what the solution is, but what I can say is, at least as it relates to localized queries, I don’t think that just taking the API’s word for it is a real solution. I don’t think that’s a productive metric to follow.

Aaron: Yeah, I agree. It’s a very interesting time to be in search and focused on the search industry, as you suggested. We were chatting previously, this is the first instance in years and years that Google’s had meaningful competition, and they now have a lot of competition that they’ve had to react to. I think I heard you on a prior interview I was listening to make a pretty controversial suggestion that maybe Google should consider putting search on autopilot and focusing very heavily on Gemini.

David: I think that they’re really at the innovator’s dilemma crisis point right now. Two years ago, they were correct to perceive ChatGPT as an existential threat. And I think a lot of what we saw very early on with the search generative experience was very much performative for Wall Street. All of the early days of hallucination and everything else like that was Google in a full-on panic. I do think that their R&D efforts in Generative AI have paid off. I think that Gemini is a viable product at this point. It’s probably something north of minimum viable products. The other thing is obviously we’ve seen a lot of at least tech press decrying the degrading of Google search results over the last three, four, five years. And I think part of the problem, for Google the company is that they. It seems like they are federated to such a way into all these little individual product fiefdoms.

Where the shopping team needs this for shopping search results, and the local team needs this for local search results, and the medical team needs this for medical search results the way they think is like boxes in a search result. For ChatGPT to come along and to totally say “No, we don’t have any boxes, we just have answers,” I think that was like such a shift in search paradigm and user experience. I’m not sure that Google, in its current mode, is fully equipped to deal with a radically different proposition like that, just because of the way the company has been structured for the last 20 years.

I think it’s really a time of the innovator’s dilemma. Are we going to bet that this thing is the next thing? Much in the same way that Facebook, I think was like, “No, we can see Facebook’s kind of dying.” Which I think you could make an argument that Google search is in the Facebook of 2015 kind of phase right now. I haven’t logged onto Facebook in four or five years probably. I think that’s become secondary to Meta’s success and that Instagram and WhatsApp are really where they’ve placed their bets. I think Sundar Pichai in the quarterly earnings calls has said “Yeah, we’re aiming to get 500 million Gemini users.” That is where we see the future.

I would at least be experimenting with and we’ve seen some rumblings about Gemini mode or AI mode where you just get a Gemini experience for a certain query. Right now, I think that Google’s thinking about that as “Oh, if I click this button, then I can pop open a comparative window.” I’m talking about like by default, like show if you think that a certain query is better solved. A pure Gemini result. Do that. I think you have to start displacing your own sort of historical golden goose a little bit if you want to maintain users and stop the bleeding away and trying GPT and try and frankly, younger users have been using TikTok for four, five, six years.

Prabhakar Raghavan said as early as, I think, 2020 that the majority of Gen Z folks started their searches on TikTok instead of Google. I think this time is like very, there’s so many different options. Now, the one thing Google has going for it is habit, right? Still people are still typing in google.com and doing a search, or they’ve got it installed as their default search engine on their Chrome browser, which sure may go away as well. The competition truly is one click away at this point, in a way that it’s never been before. I think if Google wants to retain its existing installed base, I think it may have to be a little bit more aggressive about displacing its golden goose.

Aaron: Yeah, I think that’s right. It’s certainly my perception that there’s a significant tax to using Google because they’ve been a monopoly, and they’ve been able to impose that tax at will, and I’m unwilling increasingly to suffer that tax. My own habits have changed markedly. I usually start with Perplexity, and then I’ll go to Google for something specific, some sort of recovery search to find a particular business, but it’s often based upon information that I’ve gleaned from Perplexity.

David: I do the same thing with ChatGPT. I’m less aggressive about GenAI being my default. I still find myself using Google quite a bit, but yeah, I use ChatGPT for complex queries that my brain knows Google is going to do a poor job of answering. If it’s a long tail query about some Looker Studio formula I’m trying to do, or frankly, even for an upcoming, I’m speaking at SMX Munich coming up next month and trying to figure out good restaurants to go out with colleagues in Munich. Historically that is a search I totally would’ve started on Google, but I actually know that the experience has been very frustrating because of all of the boxes that Google puts on the page, essentially.

I prefer the cleaner sort of starting point experience that GenAI offers. And that’s what I’m saying. I think even to hold me as a user, I think Google needs to be more, I would be more I would be less averse to switching over less, less eager to switch over if Google would just present me with a radically different experience.

Aaron: Yeah, I think that’s right.

David: I’m more willing to try new things today than I’ve been in 15 years probably.

Aaron: Yeah, I am as well. To try new things and then to try that same new thing 15 times over the next 15 weeks because the rate of changes.

David: Exactly. Yeah. And people rightly still complain about ChatGPT data quality in local search. That is another massive competitive advantage that Google has is Google Business Profiles are generally the most accurate, complete set of data about any given local business. Most current, most photography, et cetera. And ChatGPT, even if they just suck in the Bing index, like that’s not as good.

People have rightly said “Oh, the data’s not as good for local searches when I do these queries on ChatGPT,” and I’m just like they will find a way to solve that. They will. They have how many billions of dollars in venture funding that Yelp would love one of those billions of dollars to be their like exclusive local data supplier. They can do a similar deal with TripAdvisor and Houzz and everything else. If and when local search becomes a real vertical for ChatGPT. I think that problem is solved with a few billion dollars, and I see the experience as being much more important right now. And yes, I’m still fact checking every business that Yelp pulls up in Google, but there will be a time where that won’t be necessary. I think pay more attention to the experience and, as we were talking about in schema, the kinds of facets that ChatGPT highlights for these searches, and the kinds of prompts that it follows up with asking users. I would be paying very close attention to what that looks like right now, even if the data quality isn’t there. because I do think over time we’ll see whether it’s ChatGPT, or Perplexity, or Claude or who knows who’s going to win the game right now. But I do think what we start to see over time are bleeding away of Google users.

Aaron: Yeah. I think my experience is if I’m looking for a synthesis or distillation of information, then I’ll start with some form of AI. If I’m looking for interaction and challenge and exploration, I’ll start with AI. If I need a particular fact or data or address, then I’ll go to Google.

David: It depends, I think. So I’m a sort of an NBA salary cap nerd, and so I asked ChatGPT a bunch of questions about how much money does this player make, and like, how long does this player’s contract go on and whatever. And I would say, it didn’t do a good job last year before the introduction of search GPT. But in this most recent trade deadline in February, it did a really good job. The introduction of this real time pull from search GPT, I think has really helped with ChatGPT’s fact accuracy. I actually search depending on the discipline, the domain. I still would probably go to ChatGPT if I just needed a sort of a Siri-like fact that I might have looked up through voice search.

Aaron: I’d have to take my own advice and go back and try it on Perplexity.

David: Yeah, maybe Perplexity just needs a better set of data points maybe.

The Future of Digital Marketing in an AI-World

Aaron: Yeah. Let’s talk about the implications of AI for digital marketing. How do you use AI in your own practice, and what do you foresee as the future of ‘what does a digital marketer do’ in two years, five years?

David: ChatGPT is empowering all of us to be more productive and tasks that we might have historically spent hours in a Google Sheet or Excel or whatever we can now get that analysis done in seconds or minutes as opposed to hours or weeks. So there’s a productivity factor I’m excited about as, essentially, a solo consultant.

GenAI is going to allow me to be more productive and to mimic agency level services because I can just do more faster. So I think that’s exciting. We’re going to need to obviously keep up to date with what’s happening in all these GenAI products, but just from a professional skill standpoint I think we’re going to, we’re going to need to understand like where and what we can delegate and what GenAI is good when we do delegate it, where it’s particularly skilled, and where we still are going to be actually hold value with a particularly strong opinion or a particularly curated set of training data or whatever that is. I think being able to make those calculations on the fly, I think will be an important skill. I have not built that skill up myself at all, but I can see, based on my very early fumblings with GenAI in the last sort of six to nine months, I think that kind of thing will be important.

I feel like I have leveled up my Looker Studio game four levels in the last six weeks based on having ChatGPT as my tutor. Just saying, “Look here’s how I would do this in Google Sheets. Show me, tell me, what the formula is that I need to add as like a calculated field in in Looker Studio.” It actually works. I mean it. I feel like I have learned six months worth of Looker Studio, watching videos from experts in Looker Studio and BigQuery and all of this. I feel like I’ve learned that much in the last two or three weeks with ChatGPT as my tutor. So I think that’s that kind of like technical specificity is really compelling. I don’t know Python, happy to admit that freely, but like now I’m thinking about maybe I should just learn like intro to Python, like what are the sort of concepts of how this language sort of works. Then having ChatGPT do all of the syntax when I actually need to make a query or something like that.

I think that there’s that level, like that’s where I see ChatGPT being particularly useful in my day-to-day work today. Where I encourage clients to use it is to have your content teams and your product teams putting ChatGPT in the persona of your customer. If you say you are a caregiver of a patient with a 104 fever and a history of ear infections, if they’re looking for care, what are the questions they’re going to be asking? What are they going to be looking for on a website? It does a really good job of highlighting really big content themes that a page that is trying to address that particular person should be talking about. I don’t do that myself because I’m not the one writing the content or like putting the content briefs together, but that is definitely, I think, a real strong utility for GenAI is not writing the content whole cloth, which I still think humans do a slightly better job of, at least expert humans do a slightly better job of. But just like the outline of content and understanding the themes and the messages that need to be highlighted.

Aaron: It does a good job of developing personas and filling in the gaps in personas when you’ve got an initial framework. What am I missing? As you suggested, a good job of content themes and, as you allude to, in healthcare and MedTech, the need for expertise and the implications of having content that is not fully accurate are profound. But also, this is an industry where true expertise actually is overweighted even algorithmically.

David: Oh, for sure. Yeah. I mean, as we were talking about earlier, it’s the same cabal of like four sites that dominate every prominent query and in healthcare, and they have parlayed their brand, even if into strong rankings for everything, even if that individual piece of content isn’t particularly strong. So I don’t know if that’s you’re getting at by expertise being overplayed or whatever, but it certainly feels that way to me.

Aaron: I was actually alluding to something different, which is, we’ve had experience particularly with large health systems large sites in the healthcare space of shifting content strategy from a generic content strategy. We have a content team and they produce content, and it’s generally on target to actually, your content should be written by someone who is a medical doctor. We should make quite a lot of that credential by implementing schema, creating an ecosystem that very much drives home the expertise and the qualification of this author, and that can yield profound performance improvement.

David: For sure, and it is supposed to have yielded profound performance improvements since the EEAT release four years ago, whenever that was. I guess I’m less bullish on that being the reason that a particular piece of content ranks than I am. I still I still think it boils down to brand and domain authority in the con like domain being like the topical authority and brand as being by far the two sort of dominant signals that Google’s using.

My colleague in Germany, he’s been involved in some of these DMA lawsuits. Actually his company is suing Google and in Germany. He’s put out a couple of like really awesome LinkedIn posts about how to better optimize content for LLMs. Two of his takeaways are, yes. Having more like true experts taking a stand and having those stands be quoted and a prominent part of the page. And also being very explicit about statistics and facts with numbers. Is more, those two things will make it more likely that your content gets pulled into and an LLM, so the example I love that he gave us, don’t just say that Berlin is a large city in Germany, say it’s a city of, 250 square miles with a population of 8.2 million residents, something like that. Those sort of numeric things are more likely to get pulled into an LLM. In healthcare, don’t just say that this treatment has been shown to be more effective than an X treatment. Say it’s been shown to be 54% more effective than Y treatment. Something like that.

Aaron: Yeah. On a related note, I’ve heard that described as non-consensus data. If you publish non-consensus data, which is the intersection of data and taking a stand, you are more likely to be included inside it, which makes sense to me.

David: It makes a lot of sense. And especially since we’re seeing, at least I see in my own, I don’t know if Perplexity does the same thing but ChatGPT likes to present pros and cons of almost every topic. If you’re on either the pro side or the con side, you have a better chance of taking a strong position on one of those. You have a better chance of showing up somewhere in that answer.

Aaron: You describe your use of AI as, starting with this fumbling process. One of the things that I think is interesting and uncomfortable about using AI is that’s how I pretty much everybody starts, and that’s how you continue for some time. Each time a new model is released or you try to do something new using AI, there is this uncomfortably long period of time where you’re super amateur before you achieve some unlock.

David: Exactly. Which is definitely different from historically, definitely a trained monkey can use Google and get value, right? It could have used Google12 years ago and gotten value out of it. It’s definitely a very different sort of paradigm where you do have to work with AI and understand, try to figure out what it’s good at, what it’s not. And that definitely in my experience, just takes time. There’s no great guide for that. I found maybe I should write it on like how to use ChatGPT for Looker Studio. That information didn’t really exist, nor would I necessarily have searched for it. I don’t know. When I go to ChatGPT, I need help. I’m not looking for a guide on how to do something. I’m expecting ChatGPT to be my guide. And so I think that there’s just some sort of trial and error that’s endemic to GenAI tools, that is a new thing.

Think of it as a very low paid intern starting out and where you have to be just very explicit about exactly how you want things done. Then that intern has worked at your agency for a little longer and it understands a little bit more about what you’re trying to do and all that. Then over time, it actually will get better at responding to what you’re looking for. I personally think I’ve found that, I don’t know if it was like explicit in the model that like it just learned more about my Looker studio questions, but I found its answers getting better and more explicit with me about like “No, I really am looking for the exact whatever SQL query to add here. I’m not looking for like the logic behind why to do this.” I just want you to give me the box that I can copy.

Aaron: If we’re talking about work that can be clearly documented, the process can be documented and it’s repeatable. That’s work that AI can do and is doing in most instances. I think the conundrum, the puzzle that I think about is that’s also work that a junior employee does, an entry level analyst, a junior strategist, an intern, and you only get deep experts, which is I think what will remain valuable and prized in the market, through lots of reps as a junior level employee. So I do wonder what it’s going to look like in five years when we didn’t need to hire a whole bunch of junior level folks, and they didn’t advance to a degree of expertise that now makes them differentiated invaluable in the market. Do you have thoughts about that?

David: I guess, yes, I think so. I think you’re right that it may be harder to get on-the-job training in a traditional sense, or junior folks. But I actually think ChatGPT is going to reward beginners in any fields, any discipline, who have significant initiative and desire to learn. The notion of sort of repetition. I think that there will be new sort of training paradigms where you’re given a real-world tasks, and you as the agency owner give a junior level employee who’s just starting a task and say, “Go figure out how to do this with ChatGPT.” Start them with ChatGPT as the tutor, and then level them up once they’ve reached a certain frustration point or whatever. But I think that there will still be a place for like entry level folks. We’re still going to need client interactions and sort of translation of client goals and priorities. That is not something that AI’s going to do. I still think that there’s going to be a place for junior level folks. It’s just that you may have them figure something out first so that they understand why you’ve set up your Claude project in a certain way. I think it’s going to be a little bit of a different model than what’s existed historically where people are learning over the shoulder of someone more senior.

Aaron: Interesting. So almost an accelerated apprenticeship.

David: Yeah, I would, yeah. More like an apprenticeship, I think. I’ve been a largely a solo practitioner for most of my career. I’ve always enjoyed just experimenting with things. I’m the guy who’s just, “Give me what you’re trying to do, and let me go into a Google Sheet and figure out, and I can get you the analysis.” That’s what I love doing for clients. And I think that ChatGPT is actually going to accelerate the people who are starting from that position, which I think can be anybody, right? I think that’s the amazing thing. You’re absolutely right, deep expertise is going to be even more valuable than it is it has been historically, but I also think that a junior level person can level up faster with GenAI than they would’ve been able to previously.

Aaron: Yeah. I think that’s true, but I think that the people who rise quickly will maybe be a different type of junior level employee who might not have risen in the past. In the past, specific expertise, deep expertise in a particular domain or discipline was rewarded. I think that in the future, more flexible thinking, the ability to connect dots across multiple disciplines, the ability to think strategically, comfort with the discomfort of not knowing and excitement at not knowing, excitement at exploration, I think will be characteristics that are rewarded.

David: I guess I’m surprised that they haven’t been rewarded so far. Again, this is like me not being in the agency world, but to me, those are the people that I can teach how to manage a Google Business Profile in two hours. Are you the type of person that wants to know why your client is fourth and they should be third? I want people who are excited and like competitive and that sort of thing, to understand why. I can train them even without ChatGPT, I can train them on process fast enough.

Aaron: I think those people have been rewarded, but they’re not prevalent. They’re in rare supply, and they always have been, I think, in any business context, whether an agency or other marketing context. And now they will be even more highly prized. Yeah. Yeah. Yeah.

David: I don’t know. I’m just thinking back to we had a small team at when I was working at ThriveHive, sort of 2018-19, that was a set of marketing guides for small businesses. We had an awesome, it was a really great PR product, where we essentially gave small businesses like two hours of consulting and two hours of work every month for $130 a month or something like that. These were by necessity, so like very junior, like just out of college. I don’t even know, most of them didn’t even have marketing degrees, but they were just like, I want to be in digital marketing. These are the kind of people, they were all just like super eager to help these businesses and they wanted to understand how Google worked and like they were just go-getters and maybe we just lucked into finding a half dozen of these people. But I don’t know that ChatGPT is going to affect those folks’ job prospects moving forward. And I actually think that they probably can move up faster and scale their knowledge faster than they ever have been able to.

Aaron: It’s a very interesting time to be in digital marketing.

David: It really is. It’s exciting. I think the last time I was this excited was when Google first introduced what was, at the time, the 10-pack, the sort of radical change to the search results, right? I was just like, “Wow, this is really big, and Google’s probably put a lot of resources into this, let’s focus on local search.” I definitely feel this is arguably many times bigger than that. Even if you’re just talking about the impact on local businesses, I think that this new era that we’re in is just a total change, and it’s going to be really exciting to try to stay on top of and help businesses maintain and grow visibility.

Aaron: I agree. I agree. I really enjoyed talking with you. Thank you.

David: Yeah, likewise. This has been great. Thanks for the invite.

Aaron: Yeah, absolutely. Thank you for taking the time.