The AI Edit: How Search—and Searchers—Are Changing

Reid Carr: Welcome back to the marketing remix. In this new series, the AI Edit, we're here to demystify AI and marketing, separating fact from fiction to help today's marketing leaders make smarter, more strategic decisions from real world applications to bold predictions. This series dives into the tools, trends, and challenges shaping the future of our industry. Whether you're a CMO navigating tech investments or a strategist exploring new efficiencies, this series is built for you.

AI is transforming everything, including the way people search and the way brands show up online. But with Google rolling out AI overviews and AI mode, and user behavior evolving fast, marketers may ask, are our SEO strategies still built for what's next?

In this episode of the AI Edit, we cut through the confusion around generative engine optimization, generative AI and search engines, and AI's broader impact on where consumers will seek answers into the future.

To help you understand what's changing and what you can do about it to help make sense of it all, I'm joined by two minds leading Red Door’s strategic response to this shift: our VP of Data and MarTech, Ron Hadler, and our SEO Manager, Angie Stuart.

Together we'll unpack the implications for visibility, performance and content strategy in an AI-powered search landscape. Ron, Angie, thanks for joining me here. Excited. This is a great topic. Yeah.

Why is search ground zero for AI disruption?

Reid Carr: Alright, well, so AI is transforming marketing across the board, but nowhere is the shift more immediate than in search. So I want to talk about what's happening and what marketers need to watch out for. So I'm going to start with you, Ron. Why is search ground zero for AI disruption right now?

Ron Hadler: Yeah, I mean in real time you could just kind of see that impact because of search. I mean, no other channel feeds the models better in the sense that every time there's a query or a click that's labeling data for these AI models. And so that's just feeding them. And basically labeling is a machine learning term that allows you to classify data. So if I do a query and I get a set of links and I click on something, boom, that's a label. So that allows Google to associate what's to humans the most important thing. So that allows them to also surface things and rate things within their model to surface in their own AI-generated answers. So there's just billions of those.

On a daily basis, search is still a very big market — about 41 cents of every digital dollar goes to search. So if you shave even a fraction off of that, you're really talking about billions in dollars.

The other thing is that the user experience is instant, right? You’re taking, I do a search, I look maybe at one link, and the next one it compresses that down to a single click. I click search and I've got my answer. So as humans, we want things instantaneous, we want things convenient, and so that just goes right along with that.

The other thing is that the infrastructure was already there for AI and search. It wasn't a rebuild, it was a retrofit, a pivot for Google and Bing and things like that. And the other thing is that the habit is kind of baked in, right? Search — I type something in, click a box, and I get an answer. So we are as humans just primed to do that.

How does device type (mobile vs. desktop) impact generative search?

Reid Carr: Yeah. I actually wonder on that part too — we’re primed to do it, but what about the device, right? I mean, we know that mobile search is different than desktop search because of your likely location. I'd be curious how much that you think that's adapting. If I do my generative search on a desktop, am I going to get a different result than I might on a mobile device?

Ron Hadler: I think both of them are taking location into account and almost 90% of the cases for at least Google search, you're doing it on a Chrome browser, you're logged into a Google account so it knows where you are.

What are clients most worried about with AI in search?

Reid Carr: Yeah, yeah. I'm curious about that. Angie, as an SEO manager, what are clients or strategists, what do they ask most about AI in search and where's the confusion? Where's the panic?

Angie Stuart: Yeah, definitely clients are wondering how all this change, what it means for them, what it means for their overall strategy within search. The biggest question I hear is will we still be invisible or visible in AI if they're answering all of the questions, what does visibility mean for us? How do we measure that? And I think that's a fair question because it is evolving quickly.

It used to be a clear clickable list of links, and now it's an AI-generated summary that might not even be included within that traditional organic result, or depending on the platform, a completely separate result as an AI chat like GPT.

What's really changing I would say is user behavior and how it's adapting to these new ways to search. People don't want to dig through results anymore. They want that direct answer, and AI is delivering that to them right at the top within that platform.

That means we're no longer just optimizing to show up within a sea of blue links, but we're optimizing to be cited within that AI to inform those summaries, to build that trust within the models that users are starting to shift and depend on. So yes, SEO is still incredibly relevant, but it's shifting from just that visibility game into visibility, credibility, brand building, a lot of those other areas that brands need to consider.

So content needs to be authoritative, well structured, and visibility now really means showing up within those AI-generated results, not just necessarily underneath them.

When brands and clients are asking, they're just having to think about this as a new kind of presence. So it’s less about positioning, though that still matters, but it’s also now shifting to influence and how you can show up in this space. It can feel disruptive, and there's definitely a lot of questions about it. People are asking about it, just trying to understand the space. But I think it's honestly a huge opportunity for brands and clients to adapt to that strategy with a holistic approach to marketing.

How can we measure success in generative search results?

Reid Carr: So now this kind of question for both of you to discuss I think is around measurability, right? I mean, I think since these results are so hyper-personalized, I think a long time ago you could say, oh, we're ranking number one on this thing, and then you could measure traffic. Are we getting traffic on a particular search term? And now we’re like — I mean, how do you judge whether or not we're successful in showing up in generative search results?

Ron Hadler: The KPIs are definitely changing and it's also getting harder to do this. There’s not built-in necessary platforms doing that, but really you need to understand your AI citation rate. So how often are you appearing in those citations at the top of a Google result?

How do you know? You have to essentially do those searches and scrape the data. So you have to build infrastructure. I assume eventually there will be platforms. We have ways to do that now here at Red Door, but you have to do that, and you don't have to do that just with Google or Gemini, but you have to do that with Bing, also Perplexity, making sure that you’re covering those three major search engines that are AI-generated in order to get that.

That’s one way. And then really it’s making sure that you show up as a link or even within the AI-generated, are you one of the top three links that it’s showing as a citation? I think that’s another important thing to track. So I mean, it’s shifting the way that we are measuring things, going from how many clicks are showing up.

And you can track also the references from when there are specific URLs that you check when somebody shows up on your webpage. So if somebody clicks on a link within that AI-generated content, you can see that they did come from that versus the 10 blue links in the regular search results.

How does optimizing for multiple AI search engines change the scoreboard?

Reid Carr: And one thing I've always thought was the coolest thing about your job is it's an ongoing battle. You're fighting a competitor or something and you kind of can know the score. Are we winning or are we losing? And now you bring this up in terms of the major search engines, I guess now is another way to say it, but I mean we used to really prioritize Google and now we're talking ChatGPT, Gemini, Claude, Perplexity. So how is that changing the scoreboard for you, Angie, in terms of what we're optimizing for, and do these different quote engines, they must all behave somewhat differently. Maybe talk a little bit about that. Like I said, the scoreboard has changed, and so has the field and the scoreboard.

Angie Stuart: Even the fields — how it's being measured, all of it. And when we talk about the data part, that's kind of the exciting part of it. A year ago these platforms were emerging, and you couldn't quite get your hands on the data behind it, but there's now a lot of new tools that are emerging where you can get data behind citations, market share, even starting to get query data, sentiment, things like that. Those are things that you need to consider.

And I would say in terms of the players, AI obviously is very hot these days, but Google still is the biggest search engine, has the biggest market share. Their market share is still growing, but ChatGPT and some of these other platforms are definitely on their heels. And that shift in user behavior I think is going to be the biggest thing in terms of considering the players.

Like I had mentioned, people want answers, they want answers right away, especially considering where they are within their phase or within their search journey. But that's going to, I think, based on platform, probably factor into how somebody is searching for something because there might be one type of person that's still going to Google, there might be another that's going to ChatGPT or Perplexity, using both within their search journey.

It’s just thinking about, from a brand's perspective, how are we considering those new platforms within the search journey of a person looking for you and the information that you have for them?

What about search on other platforms like TikTok or YouTube?

Reid Carr: Yeah, I failed to even mention the other ones being TikTok, Alexa, YouTube search. I mean, I think search has absolutely exploded, and I think a lot of folks are talking so much about generative AI — maybe ChatGPT is the one they think of — but I mean the challenge remains: people search in a lot of different ways. And it's going to be interesting to see how all of that either consolidates — you can use generative search to find things on other platforms — or do they start platform-first? The behaviors are fascinating to me. I think that's one of the things that we'll certainly get into the rest of the episode going through.

Ron, can you break down what AI overviews and AI mode are within Google and how it changes what shows up on Google search results pages? I think we really want to talk about the algorithm a little bit.

What are Google AI Overviews and AI Mode?

Ron Hadler: So AI overviews is Gemini-powered, Google's model synopsis on the top of the page. That’s being invoked roughly 13 to 16% of all your queries. I see it probably at a larger amount than that, but I've opted into that prior to when it was general availability for everyone.

It’s also cutting click-through rates by anywhere between 30 to 50%, but that’s where you really want to be. That’s where you want to get to because that’s the new target. You're not saying that you’re ignoring the top 10 blue links yet — you need to do both. But a synopsis, the other thing, AI mode is right now still a lab, but it’s more of a deeper reasoning in chat. So this would be doing a search within the Gemini model itself. And so it's doing a lot more sort of deeper sort of searching versus an instantaneous, like the AI overviews are instantaneous sort.

The 10 blue links are instantaneous. AI mode — and I'm starting to see this, and I only really started to see this in the last week or so — a lot more often is that I'll see at the very bottom of the AI overview a link to “search more” and then that'll basically pop up in a new window that allows Gemini to really give me more of an AI response that you would get if you just did your query directly in the model.

Is GEO (Generative Engine Optimization) a real shift or a rebrand?

Reid Carr: Yeah. Well, Angie, so generative engine optimization. I mean that's kind of the root of, I think, what we're talking about here in a lot of ways. Is this a real shift in how we think about SEO or just kind of new packaging? You alluded to authority and some of those things earlier in the discussion there, but what's the importance of structured data, authority, some of the other things that are traditional to SEO as it relates to now, GEO as we're calling it?

Angie Stuart: Yeah, definitely. GEO I would say is an evolution in search. So it’s not necessarily just a rebranding of SEO, but an evolution in response to these new platforms and as well as that shift in user behavior that we've been talking about. It's a real response in how those engines are processing and presenting that information and then how users are reacting to it.

So when you think back to traditional SEO tactics, like how you had mentioned — backlinking, keyword optimizations — those are still necessary, but I would say that they have now become a baseline. They're honestly table stakes at this point from a brand perspective; those have to be in place. So every serious brand should really be thinking about making sure that they’re doing those things to even become a player in the game.

And then what's changing about more of this sort of modern SEO is that it's no longer just helping crawlers read your page, but it’s about helping those large language models understand, summarize, and evaluate your content into AI-generated answers.

So if your SEO team has been keeping pace with the changes, they're already focusing on a lot of those things. Like you had mentioned — schema markups, structured data, clear architecture and headings — if we're getting super dirty into E-E-A-T signals, all of those elements have become very important to have.

But within this new AI-powered landscape, there’s kind of the next layer to think about. We're no longer optimizing just to show up within those 10 blue links for someone to choose from, but we're optimizing for a model that reads everything and gives people a synthesized answer, sometimes showing the links. It’s evolving where they’re showing and referencing the sources to that. But really, it’s the answer that’s being presented.

So that means your content, yes, has to be discoverable, but it also has to be customizable, have continuity between your website and all of your other channels where you’re communicating out to people. And it needs to be easy for AI to interpret that message and feel confident enough in your message that then you’re being presented as the brand.

I don’t think GEO is necessarily replacing SEO, it’s just stretching it, pushing it into a more holistic approach for people to think about how they’re generating and contributing value to their users that then can be delivered on behalf of AI.

Is the user experience actually better with GEO?

Reid Carr: So Angie, as a professional on both sides of this, being someone who helps deliver search rankings for clients, and then as a user or consumer yourself, is this the evolution that we're experiencing at the consumer level? Is this better? Are you getting better results as a consumer because of generative AI or GEO? I mean, do you feel like, oh man, this is so much better for the consumer?

Angie Stuart: I think it depends on what that person is searching for.

If they’re searching for products versus high-level information or something where they're a little bit farther down, I think that’s where, going back to what we were talking about before, these different platforms — if I'm looking for something on Amazon, TikTok, ChatGPT, and Google — those are probably three different journeys that I'm having within search, and I'm interacting with them in different ways.

And that, I would say, does provide a better user experience for someone because they're able to go to those different platforms that work for them in content that's delivered in the way that fits what they’re looking for. If I'm really video-oriented and TikTok’s my jam and I'm doom scrolling on the next restaurant that I'm going to hit up, then that can be a great place to go. But if I really want that lengthy, longer piece of information because I'm deep-diving onto a topic, well, that might be something where Google or ChatGPT presents the best answer.

How is user behavior shifting with generative search?

Reid Carr: Yeah, no, it’s interesting. I think we’re all still, as consumers, adapting to what's the best platform, the fit for the thing that I want to do. And I would say another question for you, Angie, is what's shifted in user behavior when it comes to search, and how does this affect how we approach SEO and content structure today for our clients?

Angie Stuart: Definitely. We always talk about how the user wants more information faster. They always say our attention spans are shrinking. Users expect instant, clear answers for what they’re looking for.

Going back to how search is evolving, people are always going to be asking questions, and it’s just where are they asking them and how are they getting the answer? People these days are not necessarily going to sift through a lot of multiple links. Instead they want to rely on an answer that is giving them a snapshot of what they’re looking for and getting what they need at that first glance rather than scrolling further.

And that shift is shifting the SEO playbook, but also just the search playbook and how we're thinking about these different platforms today. It’s not just about ranking high, but it’s about being selected as a part of that answer.

So to get there, your content needs to be well structured in a way that both users are easily understanding it. That means clear defined headers, concise, well-organized information, and having those direct answers, specifically direct question and answers, specifically in your content where someone can skim and scan and ultimately understand that.

But like I mentioned before, that's kind of table stakes. It's starting to go beyond just formatting. It’s about relevancy and intent of what somebody's asking. So if you ask yourself when you're thinking about a topic, you can think about — does it make sense for that brand to be talking about that topic? And then how can they back into showing that they’re the most credible answer and that they provide that brand expertise for somebody who’s searching for it?

Are they also talking about it on different platforms and really thinking about brand building and having that continuity of that conversation across the different platforms and in what they’re talking about? It ultimately also has to be something that aligns with their customer. You have to get somebody to stop scrolling to attract their attention and be genuinely helpful in the information that you’re providing for them and align with their audience’s expectations.

So the brand that wins in this space is going to be the one that shows up as that answer and takes that old sort of traditional SEO style, marrying it with content strategy, and then also blending in that brand building to show their expertise in their space.

What are extra tactics to help content stand out in generative panels?

Reid Carr: So the headline on this is make it chunkable. I like it. That sounds like a snack of some kind, maybe something I shouldn't have. So what should users expect from search with the way that tools powered by LLMs are changing? How should brands think about staying relevant within that? You're talking about chunkable content. I mean, I'm concerned about, for example, some of the results bringing back maybe old and outdated information if you're not keeping your content fresh and relevant and knowing what’s making updates there. What are some other things based on the way the LLMs are operating?

Ron Hadler: Yeah, I think that’s where Google does have an advantage and why they still get a lot of these lion shares, even though Perplexity is gaining quite a bit. I think they did say 780 million searches last month — it is kind of crazy — and Google does billions. But all of their work in understanding page rank in their algorithm and looking at E-E-A-T signals allows them to really get the right answers too, helping with that authority.

These models do have their own internal — and here’s the thing — most of these folks can’t explain the thinking inside of the models. It is a black box. So the things that we’re talking about still, bedrock SEO principles still apply to how you present your information. But a couple of the things that we’ve talked about as far as how you lay it on the page: making it chunkable, bylines so it’s very clear so the AI picks that up, and corroborating evidence that builds that trust — those help the model surface that and help you win the cite take.

Should prompt engineering change how we approach SEO?

Reid Carr: Yeah, so some of the stuff that I think about because of the trainings that we do here at Red Door, many of which you lead, Ron, is around the idea of prompt engineering. So we're training people on how to do good prompts, and I would think then, from an SEO standpoint or GEO standpoint, we have to think about the empathy. What is the state that they’re in? So we talk about search intent optimization for many years — who is this person? What do they want? What is the experience and how are they crafting the prompt such that on the other end we’re delivering results that line up with the prompt?

So are we piecing those things together? I think again, the trainings that we do on writing good prompts, probably not everyone's doing right now, but people will get better. I mean, people are going to think “act as a this,” “I want this in a table format,” — there’s five or six things that I think we've certainly been trained to do, at least at a generic level to set who am I, who are you, what is it I want, what are the results, those sorts of things, the output expectations. So do we consider that in the way that we actually — maybe this is for Angie — do we consider prompt engineering and reverse engineering that as it relates to GEO and how we're optimizing for clients?

Angie Stuart: That goes back to, I would say, the shift in user behavior and how somebody's interacting and who is that person, how are they interacting with the platform, and then how is it giving them the answers that they’re looking for?

When it comes to these platforms, there are going to be people that are digitally native picking it up immediately, or even kind of that first round of individuals who are going to be using it that are probably going to get a little bit more savvy in the prompts that they’re using.

And then as there’s more adoption with those platforms or even those platforms moving on to taking from something like ChatGPT to Google evolving into AI mode and things like that, where they have a broader user base and people are starting to use AI — I mean, I think we've seen evolutions of AI where you see the word “furthermore” and you know it’s AI, you see a lot of dashes, you know it’s AI. So people are starting to pick up on how it’s being used and interacting with the platform.

And then for us, it really breaks down to the customer journey. Who’s the customer? How are they using it? How familiar they might be with that? And then thinking about how those answers are going to be generated for them.

So there’s kind of the customer journey and how somebody’s interacting with the platform, but then there’s also, okay, however they might be asking those questions, however familiar they are with the platform, how can we strategically show visibility for our clients? On our team, we do these AI visibility reports, snapshots, where we can get an understanding of where a client is within the space right now.

Well, as the user base starts to evolve, are we applying the right tactics to be able to show it up within those as more users start to interact with these platforms, does that change the visibility? Same with other brands who are adopting it.

So it really breaks down to what that behavior is, how people are using it, but then using our tactics and making sure that we’re ultimately showing up within those platforms if somebody is asking for information that relates to our clients and their brand.

Ron Hadler: It’s a really fascinating question, kind of thinking about the prompt engineering section of that. We take training, we want to be experts at getting the best out of the model, and we know that the more information we give it, the better answer we’re going to get, right?

So I think the average query prior to LLMs was like four words, and now I think it is all the way up to seven words, but we understand that even beyond that is much, much more important. We as experts in this sort of work know to break up our prompts into different sections, and so we’re writing paragraphs and things like that.

I think we'll get there, and I think what’s going to change that is more voice interaction with these models. I spoke on Vibe Coding yesterday, and one of the things the guy quoted was that he’s not doing a lot of that coding via typing, he’s talking to the model. So he’s using an app on his Mac to just talk all the time. So I think as we move from typing and thumbing our questions and we move to more talking, those queries are naturally going to get longer, therefore improving our prompt engineering.

What about legal or ethical issues with AI reusing branded content?

Reid Carr: Well, that's interesting. That’s going to end up changing workspaces quite a bit, I would imagine — going around talking to their machines. Now we go back to the world of closed-off offices versus the open floor plans, I suppose, if we are all talking to our machines. For another episode, I think the elephant in the room on this a little bit is more on the legal side and the illegal or ethical minefield that we have — AI reused branded content, that sort of thing.

I mean, obviously we have clients who want to be discovered, they want to have their content scraped and presented to the consumers, but I'm sure there's also the concern about what should be blocked, where there’s misinformation. I mean, should brands draw a hard line blocking access entirely to things, seeding models with misinformation? I think there’s the still old-school version of this, Angie, of black hat versus white hat, and I'm sure there’s kind of some discussions and debate of what is that in this new world. But is there a smarter, more strategic way to manage how these engines use our content and how they’re represented by the engines and the different platforms that are out there for AI?

Ron Hadler: I definitely have a perspective, but I'd like to hear Angie first just from an SEO perspective before I jump in with my sort of legal and risk-oriented answer.

Angie Stuart: Yeah, I mean, it starts to break down on who is the client, what is your brand, what are you talking about, and how visible do you want that information? So sometimes if you have somebody who's in banking or finance or something like that where you don't necessarily want that visible, that's something to consider. These models are getting trained on all of the information that’s on your website and things like that. So just being a little more methodical about how reachable it is in terms of what industry you're in.

But if it is information that you do want people to see and get answers to, then that’s also something you need to consider in terms of how they’re showing up for those answers, making sure that they are crawlable, findable, customizable, like I was mentioning before, as far as those evolving technologies.

And then it comes to how maybe your internal team is working with the information before it’s ready to share out with the world and being sure that you’re being methodical about if people are within your organization using that information — just if it is being presented, that that was the clear final message that you want somebody to find and digest and understand.

Can you really block AI models from using your content?

Reid Carr: Yeah. Do they observe no-follow tags and some of the old stuff, I mean, how do you block these engines?

Ron Hadler: Yeah, I mean, my first answer is going to be to quote Star Trek: resistance is futile. The situation is that we have an honor system on the internet that these bots will respect the robots.txt file, which gives the directives of what you can and cannot crawl because of what they want.

That doesn’t mean that bad actors or deep-seek bots might not go through and basically pull in all your information. Because when it comes to getting ahead, this is — we’ve already seen something like Perplexity doing that many searches that’s taking away from Google’s share, and it means billions of dollars. So there’s a lot at stake there.

So playing by the rules is not necessarily in these folks' advantage to do that, but your bigger players are. I mean, you even go back, I think it was 2010, Zuckerberg came out and said “Privacy is dead.” And so I think Sam Altman had that same sort of moment, complaining the fact that people were cutting off or walling off stuff. It’s like, “give us all your information.” That’s important for him at OpenAI.

So I think there’s definitely some risk and reward there, but you have to have the models suck up your information. That’s how you get to the top of the search results, the top of the answers. If you don’t, you will be left behind.

So I think you have to do that. You have to bend your content so that it fits both things — traditional search and then AI-generated search. If you don’t, you’re not going to go anywhere.

People have some techniques where you can put things behind a paywall — that stops the robots. You can also — and I don’t know how well this is actually going through — there are folks seeding bad content for the robots. It’s not necessarily showing up if a human hits your webpage, but there’s seeding bad content when you drive hallucinations and mistrust in the models, then they figure “ah, they’re going to come back to my website.”

I don’t think that’s necessarily true either. These models — these models are now the gateways to the information, and I think life will be lived through these models going forward.