/nav-logo-desktop.webp "Marketing Architects"){kind=logo}
Episode 145
What Every Brand Should Know About AI Search with Josh Blyskal, Profound
When your grandma starts asking ChatGPT for recommendations, you know search has fundamentally changed. Josh Blyskal from Profound tracks billions of real AI search queries, and his data reveals a massive shift in how consumers discover and evaluate brands.
This week, Elena, Rob, and Jonathan sit down with Josh to discuss answer engine optimization (AEO) and what marketers need to know right now. Josh explains why traditional SEO tactics like domain authority matter less in AI search, how different engines cite content, and the surprising power of FAQs in product discovery. Plus, learn why SEOs have never had a better opportunity to become heroes in their marketing organizations.
Topics Covered
• [04:00] When AI search shifted from novelty to cultural necessity
• [06:00] How ranking signals differ between ChatGPT, Gemini, and Perplexity
• [09:00] Why domain authority matters less for AEO than traditional SEO
• [14:00] Tracking real user prompts across the marketing funnel
• [19:00] How instant checkout in ChatGPT changes brand visibility strategy
• [22:00] Why FAQs increased citations by 848% in top-performing domains
• [26:00] Why SEOs should lead the AEO charge at their companies
Resources:
2025 Profound Article
Profound Website
Profound LinkedIn
Josh Blyskal’s LinkedIn
Today's Hosts
Elena Jasper
Chief Marketing Officer
Rob DeMars
Chief Product Architect
Jonathan Elfreich
Head AI Architect
Josh Blyskal
AEO Strategy & Research at Profound
Transcript
Josh: When your grandma is like, I asked ChatGPT and this is what it showed me, you kind of have that moment of like, okay, this is gonna be cultural. We're gonna be using AI. Fundamentally, as people change their behavior towards wanting conversational experiences, it's easy to start to see that the money is gonna come from optimizing for those conversational experiences.
Elena: I'm Elena Jasper, run the marketing team here at Marketing Architects, and I'm joined by my co-host Rob Mars, the chief product architect of Misfits of Machines, as well as another Misfits of Machines team member Jonathan Reich, our head AI architect.
Rob: Hello.
Jonathan: Hello.
Elena: And we're joined by a special guest, Josh Blyskal. Josh leads AEO strategy and research at Profound, one of the first companies fully dedicated to helping brands win visibility in AI driven search. He's one of the pioneers building the answer engine optimization playbook. He advises Fortune 100 brands, analyzes billions of real AI search queries and maps how generative engines like ChatGPT, Perplexity and Google's AI overviews actually rank and cite content. Before joining Profound, he worked in AI and machine learning at HubSpot, where he built automation systems that touched over a million leads a month. Now at Profound, he's helping marketers navigate the biggest shift in search since Google itself, from keywords and backlinks to citations, prompts and generative visibility. Josh, very excited to learn from you today. Welcome to the show.
Josh: Thanks for having me.
Rob: Hey, Josh, I gotta open with this here, man. So you are this guru in AEO, right? You're mapping out the future of search. If my research elves have this correct, your first entrepreneurial venture was in streetwear with a brand called Sumi. How did you get into fashion and what caused this pivot to all things algorithms?
Josh: I always say that I'm glad I got out of fashion. I had gotten really, really transfixed on the idea of being able to pay for my own semester at the last semester of college. I was like, I want to try to do something entrepreneurial to understand how businesses work. So I started this fashion thing. We started making embroidered T-shirts, very high quality. What I basically learned is that it's very easy to sell dear friends, very hard to create an actual distribution network. So I made a few tens of thousands of dollars and was like, this is great. But then immediately tried to go online, tried to get into e-commerce, and it just sucked. It was awful. We were just getting beaten by everybody. To me it was fun. You do learn, you end up learning a lot about how business actually works. At the end of it, wrapping it up, I was like, I would've probably made more money if I just sold hotdogs on the street corner.
Rob: I love it.
Elena: I think that is an interesting background. It shows you have an entrepreneurial mind. You want to learn. I feel like that is necessary when you're diving into stuff like this that's always changing. So that's what we're gonna talk about today. We're back with our thoughts on some recent marketing news. We always try to root our opinions and data, research and what drives business results. And I'm gonna kick us off quick, as I always do with some research, and I chose some from Profound because it seemed fitting for this. It's an article titled, What is Answer Engine Optimization or AEO. Understanding AEO for the future of search. I thought it'd be helpful to cover this quick to get us started. This walks through a definition of AEO as the process of ensuring that a brand, product or service is accurately represented in AI generated responses on platforms like ChatGPT or Perplexity. This article argues that brands now must do more than optimize for keywords. We need to optimize for how AI systems surface and synthesize content. Strategies include structured data, clear answer-first content, and managing how a brand appears in the training or citation data of AI systems. And that's what we're gonna talk about today. So thanks again for joining us, Josh. I'm excited to dive into this topic and talk about Profound. We're actually new Profound customers ourselves. So I'm a bit familiar with the platform now myself, which I think will be helpful. But this topic of AEO has, I think maybe it's the biggest topic right now that's making marketers nervous. It definitely makes me confused, a bit nervous. So I think this will be helpful for many. I wanted to start at the beginning here. Can you walk us through when did you first realize that search was gonna be shifting away or starting to shift from traditional SEO to something more like, you know, zero clicks and people call it GEO or AEO?
Josh: The funny thing is once ChatGPT actually launched, it was clear that the way we're gonna get information and the way that we're going to interact with computer systems is gonna change. We are going to talk to these systems, they're gonna talk back. I think it took a while, like that's end of 2022, 2023. There's a lot of just adjusting, thinking, gestating, what does this mean? Bing's integration with ChatGPT came out. All that cool stuff is starting to come into fruition. Workers are using it. We're talking about gen AI. Images, videos, all that stuff is getting very popular. And then as 2024 kind of rolls around, it starts to become clearer that the little Bing integration that was a nice little add-on in ChatGPT, that might actually start to drive actual conversions and discussions. When your grandma is like, I asked ChatGPT and this is what it showed me, you kind of have that moment of like, okay, this is gonna be cultural. We're gonna be using AI. Fundamentally, as people change their behavior towards wanting conversational experiences, it's easy to start to see that the money is gonna come from optimizing for those conversational experiences. So if I had to put a date on it, just to keep it concise, I'd probably say early '24, it was very clear. By '22 and '23, it's like you could make some educated guesses, you might feel it in your gut, but there wasn't much to action on. So I don't really blame anyone for not moving faster there.
Elena: I remember as a marketer, first as a consumer being like, oh, this is so amazing. Like, I love the experience. And then as a marketer thinking, oh crap, this is gonna change things for my role. And there's a lot of confusion about how to work with AI search. We've spent all these years trying to perfect traditional SEO strategies. I think a lot of brands are still in that journey. So I wanted to jump just like straight into what marketers are all probably wondering right now, which is what are now the most important kind of like ranking best practices or visibility signals for brands in an AI search context. And how is that different from this traditional SEO world we've all been living in and optimizing for?
Josh: Do not throw your SEO out the window. Don't get rid of your SEO. SEO is not dead, not even by a long shot. It's 0% dead. If anything, SEO is even more relevant but the truth of the matter is this, it's that SEO is based on a few key fundamentals. One, you're optimizing for one system. It is typically for 98% of SEOs, it's gonna be the Google system. That in of itself has already changed with ChatGPT, with Gemini, with Claude, with Perplexity, with all the different Google models. I don't know what the future of that landscape is right now. If you want to win wherever people are searching in AI, you at least have to have two major approaches. One with ChatGPT, and one for the Google Suite. You're either gonna optimize for ChatGPT or the Google system, right? As you're going to be doing that, you're gonna be doing it in a slightly different way. So like these systems, they are also utility first content machines. I can't say this enough, that the keyword density, that the backlinks, that all these optimizations that you would typically do for SEO because, for lack of better terms, Google has locked those new providers out of its ecosystem. All this signal that would normally get piped right into Google's ranking algorithm isn't there for a ChatGPT for instance. They're relying on totally different signals, and for the most part, that signal is very data based, utilitarian based. I don't care what your domain is. Does it answer the question in the most concise way? And secretly does it align with the query fanout that I'm gonna be writing based off of the user's long tail query to directly find that resource in the first place? In my opinion, it is about, you've got the answer engine searching here. It's about making sure your content intersects in the right way. So what you need to do is you need to understand where the fanout is. You need to understand what information needs to be provided in the absolute most concise, most utilitarian, most data forward way. Place that information, contextualize why it exists. What I mean is like that explains why listicles exist and are so prominent in AI search. It's about building context and letting the AI model be lazy. Like the AI model doesn't want to figure out everything about your industry. Think about what OpenAI would want from an incentive perspective from the people who publish content that they want to feed into ChatGPT. If I'm OpenAI, my dream is that for every long tail question, the model just gets the perfect answer from someone who's gone out, taken the time, constructed the data, and has great factual sourced appropriate data that's not slop, written, not overly promotional. You're still doing EEAT, but you're feeding the model in a different way.
Elena: Just to make sure I understand correctly, having the, you want to make sure that your answers are accurate or like just what the model's looking for, but things like domain authority, you know, we used to obsess about getting certain backlinks on your site and having a lot of earned media. Does that matter less for something like ChatGPT?
Josh: ChatGPT is still using search APIs based on the traditional search engines. So if you're not indexed, if you have zero domain authority, if you're an actively spammy site, you're not gonna show up in those search APIs anyway. But if you're in this long tail query fanout and you have a piece of content for it, it's really not gonna matter whether you're ranked 10 in the search API results or rank one, so long as you have an exceptional amount of content and an exceptional ability to convey the raw value of that content to the agent or the model, who's gonna be choosing what to cite. So don't throw it away. Still gonna get you in the consideration set. But it is not the end all be all anymore. I wouldn't lose sleep over backlinks for pure AEO strategies. I like to say it in two ways. Earning a citation has two steps. Step one is getting your content chosen. So it's literally showing up in the SERP API results list of, imagine SERP API returns a list to ChatGPT of 10 URLs. If you're in that 10, awesome. Step two is making sure your content's accessible. So when ChatGPT decides to click on your content, make sure it's not rendered by JavaScript. Make sure it's not full of images and charts that can't be deciphered. Make things clear, make things contextual, especially once the AI has actually taken the effort to give you that click.
Elena: Speaking of what you can do, I know you just mentioned a couple of things there. What would you recommend as sort of a starter audit? Like what are a couple things I could start with to optimize for AEO?
Josh: This sounds so basic, but if you want to do AEO, the thing I would do is I would open ChatGPT. I would write 10 or 15 burning questions that, you know, your buyer or your ICP is gonna be asking about you, your industry, et cetera. Looking at how often you're actually showing up. Give yourself an actual mental spot check. See the top citation sources as well. Does it surprise you? Does your competitor contextualize it? If it's, is HubSpot an awesome CRM and Salesforce is getting cited every time, what does that mean for you? Why is that happening? Can you dig deeper? I recommend just following the results, following the data. It's more insightful than most people realize, especially if you just want to start.
Jonathan: I've got a follow up question on that. A lot of marketers when they're just first getting started in AEO, they treat AI searches though it's like a single uniform monolith. But I've read a ton of your articles and you've broken down the differences between different engines, which seem to have different citation behaviors. Could you break down just a little bit about how the different generative engines behave differently and treat content differently?
Josh: It is the Wild West in that regard. So ChatGPT is the one I think everybody fixates on the most. It's the most popular pure answer engine. So the way ChatGPT is working, you're gonna press enter, you're gonna get the query fanouts. ChatGPT is gonna create query fanouts. They're gonna hit their external search provider, get a list of links, decide which links to click on, view those snippets from those pieces of content, and then provide the answer. This is a very standardized kind of RAG format. Perplexity is slightly different. Perplexity is building its own index, and it relies more heavily on its own index than ChatGPT does on theirs. So ChatGPT's got an index, but I don't see it pulling from its index nearly as much or relying solely on that index. There's a little bit more that's coming out now about ChatGPT caching some certain pages, which could be very interesting for the future. Maybe that means they're starting to finally lean very hard into those specific things. I haven't seen signal that it's definitive yet. So Perplexity, for instance, they are pretty definitive. They've got an index. They cache the pages pretty aggressively, which means basically if you make an update to a page within some determined period of time when they're gonna go re-look at that page, the updates are not gonna be present in that page, but you're gonna have this basically static representation of the page. If I said, Hey, go to try profound.com/Josh's blog latest, it'll, for instance, say my latest post was one from a week ago, even though I have one from yesterday, because the last cached page was at that point. Gemini is very interesting. Gemini relies less on Google than I thought it would. The actual Google algorithm is really feeding directly into Google AI overviews and Google AI mode. Mostly Google AI mode's most different from the Google algorithm. As far as those in-search models, Google AI overviews, I basically think of AI overviews as a basic extension of search with a YouTube amplification. It loves YouTube. I mean, it loves YouTube immensely. Gemini, very interesting. Kind of a middle ground between ChatGPT and Perplexity, but with a pretty heavy reliance on Google, but also because they're providing such long tail responses and long-winded elements to the actual fanout, there's just a bit more randomness in there. Oh, one fun one. Claude uses Brave's browser. As of my last test at least, Claude is using Brave and it's pulling rank one, two, and three from Brave in that order. That's Brave's browser. If you want to win on Claude, just win on Brave. That's the situation for now. Very, very weird.
Elena: That's crazy. That seems unfair to switch things around that much. That's crazy how different it is.
Josh: Like, if you go on Brave's browser and you go Best CRMs, if that's the fanout, Claude uses the one, two, and three.
Elena: Yeah. It's a lot of business there potentially. Wow.
Jonathan: So since the models seem to behave differently and cite and pull different sources, one thing I know marketers are probably thinking about is in terms of measurement, what KPIs should I actually be tracking off all the different AEO efforts I need to be doing? Should I be looking at visibility score, citation counts, sentiments in the user's prompts, or excess share of voice that comes from using these types of systems?
Josh: I'll start with the most boring but also most interesting KPI that I've seen so far. And it only works if your business is set up for it. But I have been eternally surprised. I really do mean that. There is rarely a business that does this and the results don't surprise me. They will add on their checkout, how did you hear about us? Add ChatGPT as an example. Add a few of the models as an example, or just add AI search as a monolith as an example. How did you hear about us? Where did you find us? ChatGPT being a response gives you a ton of signal into how people are actually finding you. The issue here is the attribution problem could probably be its own multi-billion dollar startup because there are so many people. We did this survey. 80% of people right now, they go into ChatGPT. They talk about the problem that they're having. They don't necessarily even shop yet. They get recommendations from the problem like 47% of the time. That's actually gonna be a product recommendation, veiled as a solution to a problem. If they do take that recommendation, they're gonna go to Google, they're gonna copy it, Google it, and transact. And so Google has become this navigation and transaction layer, or it's at least moving that way. Like traditional search engines are where you close the deal, less commercializable, and then the actual decision making and consumer journey is actually now happening in these conversations. For me, tracking clicks is, we're moving towards zero click. Everyone agrees that even the traditional SEOs agree we're gonna be in a more zero click world, but if we use clicks or human referrals from these LLMs as our big KPI, I think we're putting ourselves on an increasingly sinking island. Basically, I would rather track visibility. I would rather understand how I appear for the questions that are critical to my business. I can track those prompts based on what people are actually most often asking. Like a huge misconception right now is, oh, you know, you have to optimize for many engines. ChatGPT may be using Bing, and also nobody has the prompts. Such a misconception that nobody has the prompts. You know, Profound has real user prompts, and we're beating ourselves over the head. We're like, how do we tell people that we've got prompts? But if you have those prompts, you can then track your visibility in a much more impactful way. You could track those citations as well.
Jonathan: There's a huge emphasis on tracking real user prompts across the marketing funnel from awareness to decisions. I wanted to know a good example from your perspective of how a brand could efficiently and effectively do that using your guys' tech.
Josh: We've worked with a ton of different brands. I like a good e-commerce brand. I'm not gonna say the name of the brand, but they do running shoes. And they came in and the question for them was more about different categories, like how do we weight our categories? AI, we know what's winning right now in traditional search in terms of volume. Like we know what people are searching about, we know where people have the most questions, but for AI search, is that landscape slightly different? How do we prioritize that landscape? And then how do people ask those questions? So we went in and we asked questions like, you know, durability, budget, price, spoiler, budget was for this specific company, really important. Women's running shoes were also overrepresented as well, which was really interesting. So lots of people were asking very technical questions about women's running shoes, but we wouldn't have known that if we were going off of traditional search volume. And so even at the high level, when we're setting up those high level questions, we can start to understand, okay, these are the topics that matter. Now, even within that, if you're working with us on an enterprise level, we go and we actually can provide really detailed representations of the most often asked literal types of queries. So people will ask, you know, everyone wants to know not just about technical questions, but they want to know about the actual construction of the shoe. What's the height? What's the actual slope? How are we gonna actually formulate it? Is it great for people who run on the inside versus the outside of their foot? Is it great for people with shin splints, IT band issues? What kind of problems do people talk about when they get into conversations about these brand's running shoes or these competitors' brand's running shoes? So we can provide a ton of different data and insights there to say, okay, it's gonna be very specific. When they want to find you or when they want to talk about running shoes, 30% of them are gonna be talking about IT band issues or shin splints at some level. And in that group, 40% of them are actually gonna talk about budget stuff as well. It's very interesting to build out clusters like that, 'cause then you can construct prompts very easily. There should be prompts about IT bands and budget. There should be prompts about IT bands without budget. There should be prompts about shin splints with budget, shin splints without, and then you have a great distribution. Beyond that as well though, the query fanouts are again, like despite having these long tail prompts, you come and spend three weeks in the platform, you'll see that, oh, well it turns out that the budget query where I said best shoes under a hundred and best shoes $50. The query fanout, both flattens to best budget running shoes. It actually, it doesn't matter, the LLM doesn't read it in as signal what the actual price between $50 and $100 is. I doubt that's a real situation, but we've seen things very similar to that before in actual production, in impacting real marketing strategies. A marketer will come in and be like, we created a page, best running shoes under $100. And it's not ranking for this particular prompt, which is what's the best running shoe on $100? Well, look at the fanout. It's like, oh, it actually should just be general budget. Wow. Who would've known? But that's where the model's pulling from. That's what it's looking for.
Rob: So Josh, a bit earlier you were talking about the inevitability of zero click and reducing the friction in the process, the launch of instant checkout within ChatGPT means users can go from conversation to purchase without leaving the chat. It sounds pretty amazing. From your vantage point, how are brands gonna really optimize for that, for visibility, relevance, and conversation and search conversation interfaces?
Josh: It is gonna be about tools, tools and invocations and product feeds. This is where the internet RAG looks messy and gross and totally uncouth. It's like, ew, we're gonna go to websites and we're gonna scrape data and pray to God that the JavaScript works and everything is all copacetic. The future of all this is going to be programmable merchants and programmable data. I'm gonna upload my product feed as a CSV and I'm gonna give that to OpenAI as a direct analog for what my storefront looks like, what the products do, all the fields. So it becomes much more about feeding in the right structured data at the right time for these models to start ingesting and querying on. Right now the data partnerships are so limited. I am emphatic about this with brands. You do not need to be a Fortune 500 to think like this. You need to think about this right now. Think about how as a plumbing business, you would actually categorize your services and upload them as a feed. I know it's crazy, but what are the five things that ChatGPT would need to know about your business? What are the ways that you would encode that into a spreadsheet? What are your five different services? Describe them, how do you perform them, things like that. Eventually, that's where we're gonna get. I don't know if you've seen this, but ChatGPT, they released an update where they're starting to actually cite different entities. That's definitely a part of this, where standardizing the brand names and standardizing how they're representing different brands is part of this drive to create an efficient and effective commerce platform. So it's definitely going to be about structured data, definitely going to be about product feeds. Definitely gonna be about creating tools and connectors like ChatGPT. If I asked what's the best leather belt right now for men on sale at Macy's, it better not go try to scrape Macy's. It better just go hit Macy's backend 'cause Macy's provides a connector to ChatGPT to query against Macy's backend and then come back to me and say, here's the five things I found. Or if I said, book me an Uber ride from Union Square to Bushwick. It's not gonna try to open an Uber app. It's not gonna try to scrape the front end of Uber. It's gonna just call Uber on the backend. But that's, every brand in the world now is gonna be like that.
Rob: Your team has been studying real time data around events in general, right? So are you seeing any sort of interesting shifts in consumer behavior as it relates to engine performance?
Josh: We see shifts in model behavior, but it really is only minor, but it has to do with the context in which the people are asking the questions. So like Black Friday deals, the actual propensity of seeing things that are discounted goes up massively. I actually would say it's probably more boring than you'd probably think in that it tracks just like tracking anything else with e-commerce. If you want to win on Black Friday, have steep discounts, contextualize that it is for Black Friday, so that when users ask the long tail query Black Friday deals, it will also show in your description or it'll also show in your product feeds. In general though, for winning in this kind of product environment, nothing has been better than FAQs, which is crazy. We took a sample of around 10,000 domains. 10,000 top performing, 10,000 bottom performing. FAQs were present 848% more often in the top performing set. So it's 848% more prevalence, not 848% more citations, but it's way more overrepresented in domains and pages that actually performed exceptionally well in shopping. So if something performed well in shopping, there's like a 10x chance that it had some FAQs in there.
Rob: All right. I'm gonna ask the idiot follow-up question here. So is it a no-brainer then that everyone should have FAQs on their page?
Josh: You'll mark this day as the day where FAQs start to decline here on out because of their adoption. Just like with listicles, those were extremely hot in AI search. Then everyone started doing them. It is like stock market dynamics. If we all go into FAQs, then it's gonna create a second layer optimization. There'll either be some hyper niche thing you can do with FAQs to make them win, or it'll be something besides FAQs once we're all armed to the teeth in that way. But the cool thing about AI search is that we're so early on this stuff that we're still only at the first layer of actually optimizing. We're not even at the layer where everyone has one or two optimizations under their belt. This is just good fundamentals.
Rob: So you're talking about early days, look ahead two to three years, where do you think things are going in terms of search and brand visibility when everybody goes to these AI answer engines?
Josh: Two to three years from now, the experience of searching is going to be conversation first. Show me the SERP in a little dropdown or a little click through. That's gonna be the same for Google as it is for ChatGPT as it is for everybody. Like we know there's a better UX out there. So now really the pace of innovation is dictated by Google, who has a $250 billion ads business based on clicks, who has to convert into this better UX and think about how they're gonna reconcile their long-term business plan. I think if the users will adopt it, shopping is gonna be the dream scenario for OpenAI, for Perplexity, for any of this stuff. Beyond that, if you're thinking tactically, I think this whole JavaScript thing, so impermanent, eventually we're gonna have JavaScript rendering. It's not gonna be a problem. But right now it's a huge blocker. Long term it's all gonna be programmable APIs and feeds and merchant feeds and feed aggregators. It is going to work kind of like getting your train cars on the rail lines so that they can go to the, you know, OpenAI station. You're just going to want to ready your stuff, package it up, get it ready to move, get it on those rails. Because right now we're basically trying to pluck them out of the basket. We're trying to find them in the wild. I don't think that's efficient for them. I don't think it's efficient for us. So we're gonna see a lot of change there. We'll look back on this moment and we'll think what a dirty and gross way of finding information this web open RAG thing was. It was so inefficient. It'll be great for top of funnel. Amazing. Top of funnel is gonna be insane for this stuff. It's never going away. Once you get to compare a few of these products, forget about it, it's done. It's not happening. Maybe there's gonna be a dimension where it checks Reddit and a few other sources, but really it's gonna start relying on product feeds and such for accurate information. I think there's always gonna be room for UGC by the way. The minute that it's only product feeds is dystopian to me.
Rob: That was a really good spicy take. We love a good spicy take on this podcast. So what is your most contrarian AI marketing opinion right now?
Josh: It depends who you're asking. I think my most contrarian AEO take is that SEO is still here to stay. It is absolutely a world where this town is big enough for the both of us. And that's absolutely true with AEO and SEO and I can see and think about all the GEO folks in the world saying, you know, no, or maybe, I don't know. Definitively, yes. It's not even a question. Not even a question to me. And a lot of people, they'll point out that some of the strategies are quite similar, but the application is totally different. This is about juicing every last mention, every last percentage of citations out of these answer engines. Broad strokes, we are still playing with the same channels. We are doing different exercises with the same muscle groups, the same muscle groups being URLs, pages, meta tags, tables. We are still using the same units. We're gonna use them in a very different way for AI search. For the SEO community, this is right now the most advantageous and incredible. This is basically the equivalent of total societal upheaval in the SEO world. SEOs have never had a better opportunity to become leaders and supercharged superheroes in their marketing orgs. And if you as an SEO are not taking advantage of it, you're leaving value on the table. That's my genuine perspective. CMOs are looking for guidance. E-commerce teams, are we gonna let the product teams and the engineering teams at these massive organizations try to create our product feeds and they're gonna do it from scratch? They're not gonna know, they're gonna have to come to all the same conclusions you will have as an SEO, or are we gonna rely on the people who have decades of experience with structured data and building out use case specific descriptions and keywords and phrases that align with semantic intent? Who knows that? Every SEO in the world should raise their hands there. So this is a moment in my opinion. As we're seeing clicks start to decline, this is a moment where SEOs can jump into e-commerce, start to prove some dollars from that, start to jump into actually dictating social, PR. A press release, a social post on Reddit. Those could dictate your future visibility as a brand for six months, a year, and so on. Everything that your company does or is on the web now trickles back and should have some impact and some knowledge through into that SEO team. The SEO team should be leading the charge there.
Elena: I like that perspective too, though. It's very empowering for SEO teams. Like you could become one of the most important people at your company right now.
Josh: You already are one of the most important people at your company. The question is, does everybody else know? That's the deal.
Jonathan: Before we wrap up, I had one final question for Josh. You had brought up a few times now the JavaScript issue that you feel like it's going to be irrelevant in the future, but it's relevant now. Since I'm a dev and this type of stuff hits close to my heart, but I know a lot of people listening to this podcast are primarily marketers, could you just explain a little bit what you mean by the JavaScript issue real quick?
Josh: Yes, you saved me. So when these answer engines access a page, they are accessing the static HTML elements of that page. They are not waiting for JavaScript to execute on that page. For instance, if I'm American Airlines, the dynamic pricing or in stock or out of stock, you'd be surprised how many things on certain websites are actually driven primarily by JavaScript. Things that would appear to be static in many respects, especially in e-commerce, they're actually driven by JavaScript. So, you know, if I were an LLM and I went to Levi's and the Levi's marketplace page was driven by JavaScript, I would just see a bunch of white blank squares, for instance. The importance here is that in this time, there is this need for static, it's called server side rendering, where the actual page and the responsibility to load the page is done just primarily through that HTML, primarily through the server. You don't need to wait, execute anything for that page to load in full. Most pages, most pieces of content aren't so dramatically impacted. But it's important to audit that. It's important to know. Atlas already has the ability because it's browsing through your browser to view these pages and it uses computer vision. But saving computer vision, it'll be a few months or years before we really do see prominent JavaScript rendering. So for now, big technical blocker.
Jonathan: Follow up on that. Do you think that that has any impact on the FAQs and the listicles being so popular in the ecosystem?
Josh: Absolutely. God, yeah. Yeah. Server side rendered.
Jonathan: Server side rendered.
Josh: Totally. And you know, I've seen people who have FAQs who are, it's actually JavaScript rendered, and so it doesn't help, like they'll be like, we just did this test. I'm like, folks, I disabled JavaScript and dev tools. We gotta go back to the drawing board here. We gotta figure something else out here. And I wish LLMs text got picked up more often 'cause that's what I would do. I would just be like, oh, let's make a markdown page. I so dream of a world where that is a situation where we can build a parallel architecture and that actually works. Right now, it doesn't. The markdown pages, the dot text page is really primarily, they don't get cited in the answer engines unless you can make a dot text page rank better than the normal page on SERP API and all these different search APIs for whatever black magic you can do, which you're doing that, you're really deep in some very gray hat stuff. So I wouldn't even recommend trying, but if you could, maybe then you'd have some inroads. But anyway, yeah, you get the point.
Elena: Josh, this has been really fun. I followed more of it than I thought I would, so thank you for the way you explain things. It's very clear and I appreciate that. I wanted to wrap up with something kind of fun. Do you have a piece of technology that you really love right now? And then is there one that you think is a little bit overhyped?
Josh: This is a spicy take. My favorite piece of technology is just my date wristwatch. I love it. I think that the most incredible products are the products that are durable, products that are omnipresent, universal products. I think you can't understate the importance of time in society and the nature of time. Like time is something that humans had to invent. I literally can't live without it, and I think the beauty and the craftsmanship of a wristwatch is something that I wish I saw more often in software. I'm always inspired when I look at that thinking, the actual features and the actual elements of this watch inside the case of the watch are totally hidden to me. But I know that if I went in and I unscrewed it, it would be, it's absolutely gorgeous. It's machined to perfection. So much right now, I think, in software, we're doing things very iteratively. Like the scary thing about software 20 years ago is software used to come on a disc. If you didn't get it right, the people would take the disc, they'd put it in their computers, and you're screwed. I mean, it's such a different world. And I think the fun thing is that we're building software, like we're building it for the disc, but we're building it at hyper speed. You know, if you were a software consumer 20 years ago, you'd get a pallet full of discs dropped off. That's how much we're shipping here. Maybe I could have chosen something more new. Maybe I could just said I love, you know, Claude Opus 4.5, but the truth of the matter is that the old things are often the best things. And there's still a lot of beauty out there that's not even AI related.
Elena: Yeah. No, I love that. What about, do you think anything's overhyped right now? Any new piece of tech you've tried or seen?
Josh: There was something recently actually. I think everybody's just a little bit crazed about agents. I think agents are cool, but I think an agent is a consequence, it's a concept. It's let's put a few LLMs together and give it some context. It is not a new way of doing business. It is a natural extension of having something that can simulate large language and speak back to people and reason through processes. I think we should not under index on how cool that is as a process, but I don't think we should treat it like, are we doing it agentically? Isn't it agentic product? Do we have agents doing this? Everything should be agentic and we should just be quiet about it and keep moving. I think that's my perspective right now is agents are totally priced in. It's, that is the A1 standard of doing this. It is not a feature. It is how it is done. Yeah, that's my perspective at least.
Elena: Love it. All right, Rob, Jonathan, did you have a quick one? Quick tech?
Rob: I've been in a cult since the mid nineties called Apple, and you talk about shipping software on a disc. I used to stand in line at the Apple store, like it was a rock concert waiting to get my disc of, you know, Snow Leopard or whatever animal they named it after. So as someone who's been such a diehard fan boy, I just recently, and I'm embarrassed to say this, I just recently switched from Safari to Chrome and my life has changed, and I know that sounds ridiculous, but I just refused to use anything but Safari and I get it now. I get it. I'm out of the cult. I'm like, okay, I've been in this bubble way the hell too long, this walled garden. I was not only in the walled garden, I was smoking the walled garden. I gotta get out. So I'm out on that. I love Chrome and I would say overhyped on that theme, any Apple announcement in the last five years has been overhyped. I'm just sorry. I've just, I'm throwing the darts. I'm out. Sorry. You know, Tim Cook is all about money. He's not about innovation, saying, I just, I missed Steve Jobs. I'm glad Tim Cook's leaving. There you go, Jonathan.
Jonathan: That is a spicy one, Rob. Okay, so mine's gonna be actually AI model related. On the, obsessed right now, there's a new image model that came out recently that's called Z Image. Very interesting. It's open source and you can make pretty much anything with it. It's crazy what it can do and it's fast. Like you can run it locally and get dozens of images, right? Like you can basically put together an entire book's worth of images in maybe 30 minutes with a good graphics card. It's insane. I was actually using it to illustrate one of the Chronicles of Narnia, but it was able to literally illustrate the entire book in 30 minutes, which is insane how fast and efficient this new model is. Overhyped, I'm gonna say this, Flux two Dev has not lived up to my expectations. It's huge. It's hard to run. It takes forever to get an image, and the images I get out sometimes aren't as good as Qwen Image Edit or even this new Z Image model. So those are my two things I've been absolutely obsessed with.
Elena: Josh, thank you so much for joining us today. This has been so fun and interesting. Where can people follow you and learn more about what you're doing at Profound?
Josh: I am on LinkedIn, but Profound's website is try profound.com. But check me out on LinkedIn. All my research goes on LinkedIn. We do have a research hub on Profound, but it's just literally what I did on LinkedIn, but then with more words. So really if you want to get insights into how AI search is moving, the latest and greatest in the field, check me out. Just search me, Josh Blyskal, just like the name.
Elena: Love it and I'll go ahead and plug Profound. It's a really cool tool and if your marketing team is looking to take action on AEO, that can also be a good thing to do is go talk to the team there.
Josh: Yes.
Elena: Thank you.
Rob: Thank you so much, Josh. That was so interesting.
Episode 145
What Every Brand Should Know About AI Search with Josh Blyskal, Profound
When your grandma starts asking ChatGPT for recommendations, you know search has fundamentally changed. Josh Blyskal from Profound tracks billions of real AI search queries, and his data reveals a massive shift in how consumers discover and evaluate brands.
This week, Elena, Rob, and Jonathan sit down with Josh to discuss answer engine optimization (AEO) and what marketers need to know right now. Josh explains why traditional SEO tactics like domain authority matter less in AI search, how different engines cite content, and the surprising power of FAQs in product discovery. Plus, learn why SEOs have never had a better opportunity to become heroes in their marketing organizations.
Topics Covered
• [04:00] When AI search shifted from novelty to cultural necessity
• [06:00] How ranking signals differ between ChatGPT, Gemini, and Perplexity
• [09:00] Why domain authority matters less for AEO than traditional SEO
• [14:00] Tracking real user prompts across the marketing funnel
• [19:00] How instant checkout in ChatGPT changes brand visibility strategy
• [22:00] Why FAQs increased citations by 848% in top-performing domains
• [26:00] Why SEOs should lead the AEO charge at their companies
Resources:
2025 Profound Article
Profound Website
Profound LinkedIn
Josh Blyskal’s LinkedIn
Today's Hosts
Elena Jasper
Chief Marketing Officer
Rob DeMars
Chief Product Architect
Jonathan Elfreich
Head AI Architect
Josh Blyskal
AEO Strategy & Research at Profound
Subscribe on
Enjoy this episode? Leave us a review.
Transcript
Josh: When your grandma is like, I asked ChatGPT and this is what it showed me, you kind of have that moment of like, okay, this is gonna be cultural. We're gonna be using AI. Fundamentally, as people change their behavior towards wanting conversational experiences, it's easy to start to see that the money is gonna come from optimizing for those conversational experiences.
Elena: I'm Elena Jasper, run the marketing team here at Marketing Architects, and I'm joined by my co-host Rob Mars, the chief product architect of Misfits of Machines, as well as another Misfits of Machines team member Jonathan Reich, our head AI architect.
Rob: Hello.
Jonathan: Hello.
Elena: And we're joined by a special guest, Josh Blyskal. Josh leads AEO strategy and research at Profound, one of the first companies fully dedicated to helping brands win visibility in AI driven search. He's one of the pioneers building the answer engine optimization playbook. He advises Fortune 100 brands, analyzes billions of real AI search queries and maps how generative engines like ChatGPT, Perplexity and Google's AI overviews actually rank and cite content. Before joining Profound, he worked in AI and machine learning at HubSpot, where he built automation systems that touched over a million leads a month. Now at Profound, he's helping marketers navigate the biggest shift in search since Google itself, from keywords and backlinks to citations, prompts and generative visibility. Josh, very excited to learn from you today. Welcome to the show.
Josh: Thanks for having me.
Rob: Hey, Josh, I gotta open with this here, man. So you are this guru in AEO, right? You're mapping out the future of search. If my research elves have this correct, your first entrepreneurial venture was in streetwear with a brand called Sumi. How did you get into fashion and what caused this pivot to all things algorithms?
Josh: I always say that I'm glad I got out of fashion. I had gotten really, really transfixed on the idea of being able to pay for my own semester at the last semester of college. I was like, I want to try to do something entrepreneurial to understand how businesses work. So I started this fashion thing. We started making embroidered T-shirts, very high quality. What I basically learned is that it's very easy to sell dear friends, very hard to create an actual distribution network. So I made a few tens of thousands of dollars and was like, this is great. But then immediately tried to go online, tried to get into e-commerce, and it just sucked. It was awful. We were just getting beaten by everybody. To me it was fun. You do learn, you end up learning a lot about how business actually works. At the end of it, wrapping it up, I was like, I would've probably made more money if I just sold hotdogs on the street corner.
Rob: I love it.
Elena: I think that is an interesting background. It shows you have an entrepreneurial mind. You want to learn. I feel like that is necessary when you're diving into stuff like this that's always changing. So that's what we're gonna talk about today. We're back with our thoughts on some recent marketing news. We always try to root our opinions and data, research and what drives business results. And I'm gonna kick us off quick, as I always do with some research, and I chose some from Profound because it seemed fitting for this. It's an article titled, What is Answer Engine Optimization or AEO. Understanding AEO for the future of search. I thought it'd be helpful to cover this quick to get us started. This walks through a definition of AEO as the process of ensuring that a brand, product or service is accurately represented in AI generated responses on platforms like ChatGPT or Perplexity. This article argues that brands now must do more than optimize for keywords. We need to optimize for how AI systems surface and synthesize content. Strategies include structured data, clear answer-first content, and managing how a brand appears in the training or citation data of AI systems. And that's what we're gonna talk about today. So thanks again for joining us, Josh. I'm excited to dive into this topic and talk about Profound. We're actually new Profound customers ourselves. So I'm a bit familiar with the platform now myself, which I think will be helpful. But this topic of AEO has, I think maybe it's the biggest topic right now that's making marketers nervous. It definitely makes me confused, a bit nervous. So I think this will be helpful for many. I wanted to start at the beginning here. Can you walk us through when did you first realize that search was gonna be shifting away or starting to shift from traditional SEO to something more like, you know, zero clicks and people call it GEO or AEO?
Josh: The funny thing is once ChatGPT actually launched, it was clear that the way we're gonna get information and the way that we're going to interact with computer systems is gonna change. We are going to talk to these systems, they're gonna talk back. I think it took a while, like that's end of 2022, 2023. There's a lot of just adjusting, thinking, gestating, what does this mean? Bing's integration with ChatGPT came out. All that cool stuff is starting to come into fruition. Workers are using it. We're talking about gen AI. Images, videos, all that stuff is getting very popular. And then as 2024 kind of rolls around, it starts to become clearer that the little Bing integration that was a nice little add-on in ChatGPT, that might actually start to drive actual conversions and discussions. When your grandma is like, I asked ChatGPT and this is what it showed me, you kind of have that moment of like, okay, this is gonna be cultural. We're gonna be using AI. Fundamentally, as people change their behavior towards wanting conversational experiences, it's easy to start to see that the money is gonna come from optimizing for those conversational experiences. So if I had to put a date on it, just to keep it concise, I'd probably say early '24, it was very clear. By '22 and '23, it's like you could make some educated guesses, you might feel it in your gut, but there wasn't much to action on. So I don't really blame anyone for not moving faster there.
Elena: I remember as a marketer, first as a consumer being like, oh, this is so amazing. Like, I love the experience. And then as a marketer thinking, oh crap, this is gonna change things for my role. And there's a lot of confusion about how to work with AI search. We've spent all these years trying to perfect traditional SEO strategies. I think a lot of brands are still in that journey. So I wanted to jump just like straight into what marketers are all probably wondering right now, which is what are now the most important kind of like ranking best practices or visibility signals for brands in an AI search context. And how is that different from this traditional SEO world we've all been living in and optimizing for?
Josh: Do not throw your SEO out the window. Don't get rid of your SEO. SEO is not dead, not even by a long shot. It's 0% dead. If anything, SEO is even more relevant but the truth of the matter is this, it's that SEO is based on a few key fundamentals. One, you're optimizing for one system. It is typically for 98% of SEOs, it's gonna be the Google system. That in of itself has already changed with ChatGPT, with Gemini, with Claude, with Perplexity, with all the different Google models. I don't know what the future of that landscape is right now. If you want to win wherever people are searching in AI, you at least have to have two major approaches. One with ChatGPT, and one for the Google Suite. You're either gonna optimize for ChatGPT or the Google system, right? As you're going to be doing that, you're gonna be doing it in a slightly different way. So like these systems, they are also utility first content machines. I can't say this enough, that the keyword density, that the backlinks, that all these optimizations that you would typically do for SEO because, for lack of better terms, Google has locked those new providers out of its ecosystem. All this signal that would normally get piped right into Google's ranking algorithm isn't there for a ChatGPT for instance. They're relying on totally different signals, and for the most part, that signal is very data based, utilitarian based. I don't care what your domain is. Does it answer the question in the most concise way? And secretly does it align with the query fanout that I'm gonna be writing based off of the user's long tail query to directly find that resource in the first place? In my opinion, it is about, you've got the answer engine searching here. It's about making sure your content intersects in the right way. So what you need to do is you need to understand where the fanout is. You need to understand what information needs to be provided in the absolute most concise, most utilitarian, most data forward way. Place that information, contextualize why it exists. What I mean is like that explains why listicles exist and are so prominent in AI search. It's about building context and letting the AI model be lazy. Like the AI model doesn't want to figure out everything about your industry. Think about what OpenAI would want from an incentive perspective from the people who publish content that they want to feed into ChatGPT. If I'm OpenAI, my dream is that for every long tail question, the model just gets the perfect answer from someone who's gone out, taken the time, constructed the data, and has great factual sourced appropriate data that's not slop, written, not overly promotional. You're still doing EEAT, but you're feeding the model in a different way.
Elena: Just to make sure I understand correctly, having the, you want to make sure that your answers are accurate or like just what the model's looking for, but things like domain authority, you know, we used to obsess about getting certain backlinks on your site and having a lot of earned media. Does that matter less for something like ChatGPT?
Josh: ChatGPT is still using search APIs based on the traditional search engines. So if you're not indexed, if you have zero domain authority, if you're an actively spammy site, you're not gonna show up in those search APIs anyway. But if you're in this long tail query fanout and you have a piece of content for it, it's really not gonna matter whether you're ranked 10 in the search API results or rank one, so long as you have an exceptional amount of content and an exceptional ability to convey the raw value of that content to the agent or the model, who's gonna be choosing what to cite. So don't throw it away. Still gonna get you in the consideration set. But it is not the end all be all anymore. I wouldn't lose sleep over backlinks for pure AEO strategies. I like to say it in two ways. Earning a citation has two steps. Step one is getting your content chosen. So it's literally showing up in the SERP API results list of, imagine SERP API returns a list to ChatGPT of 10 URLs. If you're in that 10, awesome. Step two is making sure your content's accessible. So when ChatGPT decides to click on your content, make sure it's not rendered by JavaScript. Make sure it's not full of images and charts that can't be deciphered. Make things clear, make things contextual, especially once the AI has actually taken the effort to give you that click.
Elena: Speaking of what you can do, I know you just mentioned a couple of things there. What would you recommend as sort of a starter audit? Like what are a couple things I could start with to optimize for AEO?
Josh: This sounds so basic, but if you want to do AEO, the thing I would do is I would open ChatGPT. I would write 10 or 15 burning questions that, you know, your buyer or your ICP is gonna be asking about you, your industry, et cetera. Looking at how often you're actually showing up. Give yourself an actual mental spot check. See the top citation sources as well. Does it surprise you? Does your competitor contextualize it? If it's, is HubSpot an awesome CRM and Salesforce is getting cited every time, what does that mean for you? Why is that happening? Can you dig deeper? I recommend just following the results, following the data. It's more insightful than most people realize, especially if you just want to start.
Jonathan: I've got a follow up question on that. A lot of marketers when they're just first getting started in AEO, they treat AI searches though it's like a single uniform monolith. But I've read a ton of your articles and you've broken down the differences between different engines, which seem to have different citation behaviors. Could you break down just a little bit about how the different generative engines behave differently and treat content differently?
Josh: It is the Wild West in that regard. So ChatGPT is the one I think everybody fixates on the most. It's the most popular pure answer engine. So the way ChatGPT is working, you're gonna press enter, you're gonna get the query fanouts. ChatGPT is gonna create query fanouts. They're gonna hit their external search provider, get a list of links, decide which links to click on, view those snippets from those pieces of content, and then provide the answer. This is a very standardized kind of RAG format. Perplexity is slightly different. Perplexity is building its own index, and it relies more heavily on its own index than ChatGPT does on theirs. So ChatGPT's got an index, but I don't see it pulling from its index nearly as much or relying solely on that index. There's a little bit more that's coming out now about ChatGPT caching some certain pages, which could be very interesting for the future. Maybe that means they're starting to finally lean very hard into those specific things. I haven't seen signal that it's definitive yet. So Perplexity, for instance, they are pretty definitive. They've got an index. They cache the pages pretty aggressively, which means basically if you make an update to a page within some determined period of time when they're gonna go re-look at that page, the updates are not gonna be present in that page, but you're gonna have this basically static representation of the page. If I said, Hey, go to try profound.com/Josh's blog latest, it'll, for instance, say my latest post was one from a week ago, even though I have one from yesterday, because the last cached page was at that point. Gemini is very interesting. Gemini relies less on Google than I thought it would. The actual Google algorithm is really feeding directly into Google AI overviews and Google AI mode. Mostly Google AI mode's most different from the Google algorithm. As far as those in-search models, Google AI overviews, I basically think of AI overviews as a basic extension of search with a YouTube amplification. It loves YouTube. I mean, it loves YouTube immensely. Gemini, very interesting. Kind of a middle ground between ChatGPT and Perplexity, but with a pretty heavy reliance on Google, but also because they're providing such long tail responses and long-winded elements to the actual fanout, there's just a bit more randomness in there. Oh, one fun one. Claude uses Brave's browser. As of my last test at least, Claude is using Brave and it's pulling rank one, two, and three from Brave in that order. That's Brave's browser. If you want to win on Claude, just win on Brave. That's the situation for now. Very, very weird.
Elena: That's crazy. That seems unfair to switch things around that much. That's crazy how different it is.
Josh: Like, if you go on Brave's browser and you go Best CRMs, if that's the fanout, Claude uses the one, two, and three.
Elena: Yeah. It's a lot of business there potentially. Wow.
Jonathan: So since the models seem to behave differently and cite and pull different sources, one thing I know marketers are probably thinking about is in terms of measurement, what KPIs should I actually be tracking off all the different AEO efforts I need to be doing? Should I be looking at visibility score, citation counts, sentiments in the user's prompts, or excess share of voice that comes from using these types of systems?
Josh: I'll start with the most boring but also most interesting KPI that I've seen so far. And it only works if your business is set up for it. But I have been eternally surprised. I really do mean that. There is rarely a business that does this and the results don't surprise me. They will add on their checkout, how did you hear about us? Add ChatGPT as an example. Add a few of the models as an example, or just add AI search as a monolith as an example. How did you hear about us? Where did you find us? ChatGPT being a response gives you a ton of signal into how people are actually finding you. The issue here is the attribution problem could probably be its own multi-billion dollar startup because there are so many people. We did this survey. 80% of people right now, they go into ChatGPT. They talk about the problem that they're having. They don't necessarily even shop yet. They get recommendations from the problem like 47% of the time. That's actually gonna be a product recommendation, veiled as a solution to a problem. If they do take that recommendation, they're gonna go to Google, they're gonna copy it, Google it, and transact. And so Google has become this navigation and transaction layer, or it's at least moving that way. Like traditional search engines are where you close the deal, less commercializable, and then the actual decision making and consumer journey is actually now happening in these conversations. For me, tracking clicks is, we're moving towards zero click. Everyone agrees that even the traditional SEOs agree we're gonna be in a more zero click world, but if we use clicks or human referrals from these LLMs as our big KPI, I think we're putting ourselves on an increasingly sinking island. Basically, I would rather track visibility. I would rather understand how I appear for the questions that are critical to my business. I can track those prompts based on what people are actually most often asking. Like a huge misconception right now is, oh, you know, you have to optimize for many engines. ChatGPT may be using Bing, and also nobody has the prompts. Such a misconception that nobody has the prompts. You know, Profound has real user prompts, and we're beating ourselves over the head. We're like, how do we tell people that we've got prompts? But if you have those prompts, you can then track your visibility in a much more impactful way. You could track those citations as well.
Jonathan: There's a huge emphasis on tracking real user prompts across the marketing funnel from awareness to decisions. I wanted to know a good example from your perspective of how a brand could efficiently and effectively do that using your guys' tech.
Josh: We've worked with a ton of different brands. I like a good e-commerce brand. I'm not gonna say the name of the brand, but they do running shoes. And they came in and the question for them was more about different categories, like how do we weight our categories? AI, we know what's winning right now in traditional search in terms of volume. Like we know what people are searching about, we know where people have the most questions, but for AI search, is that landscape slightly different? How do we prioritize that landscape? And then how do people ask those questions? So we went in and we asked questions like, you know, durability, budget, price, spoiler, budget was for this specific company, really important. Women's running shoes were also overrepresented as well, which was really interesting. So lots of people were asking very technical questions about women's running shoes, but we wouldn't have known that if we were going off of traditional search volume. And so even at the high level, when we're setting up those high level questions, we can start to understand, okay, these are the topics that matter. Now, even within that, if you're working with us on an enterprise level, we go and we actually can provide really detailed representations of the most often asked literal types of queries. So people will ask, you know, everyone wants to know not just about technical questions, but they want to know about the actual construction of the shoe. What's the height? What's the actual slope? How are we gonna actually formulate it? Is it great for people who run on the inside versus the outside of their foot? Is it great for people with shin splints, IT band issues? What kind of problems do people talk about when they get into conversations about these brand's running shoes or these competitors' brand's running shoes? So we can provide a ton of different data and insights there to say, okay, it's gonna be very specific. When they want to find you or when they want to talk about running shoes, 30% of them are gonna be talking about IT band issues or shin splints at some level. And in that group, 40% of them are actually gonna talk about budget stuff as well. It's very interesting to build out clusters like that, 'cause then you can construct prompts very easily. There should be prompts about IT bands and budget. There should be prompts about IT bands without budget. There should be prompts about shin splints with budget, shin splints without, and then you have a great distribution. Beyond that as well though, the query fanouts are again, like despite having these long tail prompts, you come and spend three weeks in the platform, you'll see that, oh, well it turns out that the budget query where I said best shoes under a hundred and best shoes $50. The query fanout, both flattens to best budget running shoes. It actually, it doesn't matter, the LLM doesn't read it in as signal what the actual price between $50 and $100 is. I doubt that's a real situation, but we've seen things very similar to that before in actual production, in impacting real marketing strategies. A marketer will come in and be like, we created a page, best running shoes under $100. And it's not ranking for this particular prompt, which is what's the best running shoe on $100? Well, look at the fanout. It's like, oh, it actually should just be general budget. Wow. Who would've known? But that's where the model's pulling from. That's what it's looking for.
Rob: So Josh, a bit earlier you were talking about the inevitability of zero click and reducing the friction in the process, the launch of instant checkout within ChatGPT means users can go from conversation to purchase without leaving the chat. It sounds pretty amazing. From your vantage point, how are brands gonna really optimize for that, for visibility, relevance, and conversation and search conversation interfaces?
Josh: It is gonna be about tools, tools and invocations and product feeds. This is where the internet RAG looks messy and gross and totally uncouth. It's like, ew, we're gonna go to websites and we're gonna scrape data and pray to God that the JavaScript works and everything is all copacetic. The future of all this is going to be programmable merchants and programmable data. I'm gonna upload my product feed as a CSV and I'm gonna give that to OpenAI as a direct analog for what my storefront looks like, what the products do, all the fields. So it becomes much more about feeding in the right structured data at the right time for these models to start ingesting and querying on. Right now the data partnerships are so limited. I am emphatic about this with brands. You do not need to be a Fortune 500 to think like this. You need to think about this right now. Think about how as a plumbing business, you would actually categorize your services and upload them as a feed. I know it's crazy, but what are the five things that ChatGPT would need to know about your business? What are the ways that you would encode that into a spreadsheet? What are your five different services? Describe them, how do you perform them, things like that. Eventually, that's where we're gonna get. I don't know if you've seen this, but ChatGPT, they released an update where they're starting to actually cite different entities. That's definitely a part of this, where standardizing the brand names and standardizing how they're representing different brands is part of this drive to create an efficient and effective commerce platform. So it's definitely going to be about structured data, definitely going to be about product feeds. Definitely gonna be about creating tools and connectors like ChatGPT. If I asked what's the best leather belt right now for men on sale at Macy's, it better not go try to scrape Macy's. It better just go hit Macy's backend 'cause Macy's provides a connector to ChatGPT to query against Macy's backend and then come back to me and say, here's the five things I found. Or if I said, book me an Uber ride from Union Square to Bushwick. It's not gonna try to open an Uber app. It's not gonna try to scrape the front end of Uber. It's gonna just call Uber on the backend. But that's, every brand in the world now is gonna be like that.
Rob: Your team has been studying real time data around events in general, right? So are you seeing any sort of interesting shifts in consumer behavior as it relates to engine performance?
Josh: We see shifts in model behavior, but it really is only minor, but it has to do with the context in which the people are asking the questions. So like Black Friday deals, the actual propensity of seeing things that are discounted goes up massively. I actually would say it's probably more boring than you'd probably think in that it tracks just like tracking anything else with e-commerce. If you want to win on Black Friday, have steep discounts, contextualize that it is for Black Friday, so that when users ask the long tail query Black Friday deals, it will also show in your description or it'll also show in your product feeds. In general though, for winning in this kind of product environment, nothing has been better than FAQs, which is crazy. We took a sample of around 10,000 domains. 10,000 top performing, 10,000 bottom performing. FAQs were present 848% more often in the top performing set. So it's 848% more prevalence, not 848% more citations, but it's way more overrepresented in domains and pages that actually performed exceptionally well in shopping. So if something performed well in shopping, there's like a 10x chance that it had some FAQs in there.
Rob: All right. I'm gonna ask the idiot follow-up question here. So is it a no-brainer then that everyone should have FAQs on their page?
Josh: You'll mark this day as the day where FAQs start to decline here on out because of their adoption. Just like with listicles, those were extremely hot in AI search. Then everyone started doing them. It is like stock market dynamics. If we all go into FAQs, then it's gonna create a second layer optimization. There'll either be some hyper niche thing you can do with FAQs to make them win, or it'll be something besides FAQs once we're all armed to the teeth in that way. But the cool thing about AI search is that we're so early on this stuff that we're still only at the first layer of actually optimizing. We're not even at the layer where everyone has one or two optimizations under their belt. This is just good fundamentals.
Rob: So you're talking about early days, look ahead two to three years, where do you think things are going in terms of search and brand visibility when everybody goes to these AI answer engines?
Josh: Two to three years from now, the experience of searching is going to be conversation first. Show me the SERP in a little dropdown or a little click through. That's gonna be the same for Google as it is for ChatGPT as it is for everybody. Like we know there's a better UX out there. So now really the pace of innovation is dictated by Google, who has a $250 billion ads business based on clicks, who has to convert into this better UX and think about how they're gonna reconcile their long-term business plan. I think if the users will adopt it, shopping is gonna be the dream scenario for OpenAI, for Perplexity, for any of this stuff. Beyond that, if you're thinking tactically, I think this whole JavaScript thing, so impermanent, eventually we're gonna have JavaScript rendering. It's not gonna be a problem. But right now it's a huge blocker. Long term it's all gonna be programmable APIs and feeds and merchant feeds and feed aggregators. It is going to work kind of like getting your train cars on the rail lines so that they can go to the, you know, OpenAI station. You're just going to want to ready your stuff, package it up, get it ready to move, get it on those rails. Because right now we're basically trying to pluck them out of the basket. We're trying to find them in the wild. I don't think that's efficient for them. I don't think it's efficient for us. So we're gonna see a lot of change there. We'll look back on this moment and we'll think what a dirty and gross way of finding information this web open RAG thing was. It was so inefficient. It'll be great for top of funnel. Amazing. Top of funnel is gonna be insane for this stuff. It's never going away. Once you get to compare a few of these products, forget about it, it's done. It's not happening. Maybe there's gonna be a dimension where it checks Reddit and a few other sources, but really it's gonna start relying on product feeds and such for accurate information. I think there's always gonna be room for UGC by the way. The minute that it's only product feeds is dystopian to me.
Rob: That was a really good spicy take. We love a good spicy take on this podcast. So what is your most contrarian AI marketing opinion right now?
Josh: It depends who you're asking. I think my most contrarian AEO take is that SEO is still here to stay. It is absolutely a world where this town is big enough for the both of us. And that's absolutely true with AEO and SEO and I can see and think about all the GEO folks in the world saying, you know, no, or maybe, I don't know. Definitively, yes. It's not even a question. Not even a question to me. And a lot of people, they'll point out that some of the strategies are quite similar, but the application is totally different. This is about juicing every last mention, every last percentage of citations out of these answer engines. Broad strokes, we are still playing with the same channels. We are doing different exercises with the same muscle groups, the same muscle groups being URLs, pages, meta tags, tables. We are still using the same units. We're gonna use them in a very different way for AI search. For the SEO community, this is right now the most advantageous and incredible. This is basically the equivalent of total societal upheaval in the SEO world. SEOs have never had a better opportunity to become leaders and supercharged superheroes in their marketing orgs. And if you as an SEO are not taking advantage of it, you're leaving value on the table. That's my genuine perspective. CMOs are looking for guidance. E-commerce teams, are we gonna let the product teams and the engineering teams at these massive organizations try to create our product feeds and they're gonna do it from scratch? They're not gonna know, they're gonna have to come to all the same conclusions you will have as an SEO, or are we gonna rely on the people who have decades of experience with structured data and building out use case specific descriptions and keywords and phrases that align with semantic intent? Who knows that? Every SEO in the world should raise their hands there. So this is a moment in my opinion. As we're seeing clicks start to decline, this is a moment where SEOs can jump into e-commerce, start to prove some dollars from that, start to jump into actually dictating social, PR. A press release, a social post on Reddit. Those could dictate your future visibility as a brand for six months, a year, and so on. Everything that your company does or is on the web now trickles back and should have some impact and some knowledge through into that SEO team. The SEO team should be leading the charge there.
Elena: I like that perspective too, though. It's very empowering for SEO teams. Like you could become one of the most important people at your company right now.
Josh: You already are one of the most important people at your company. The question is, does everybody else know? That's the deal.
Jonathan: Before we wrap up, I had one final question for Josh. You had brought up a few times now the JavaScript issue that you feel like it's going to be irrelevant in the future, but it's relevant now. Since I'm a dev and this type of stuff hits close to my heart, but I know a lot of people listening to this podcast are primarily marketers, could you just explain a little bit what you mean by the JavaScript issue real quick?
Josh: Yes, you saved me. So when these answer engines access a page, they are accessing the static HTML elements of that page. They are not waiting for JavaScript to execute on that page. For instance, if I'm American Airlines, the dynamic pricing or in stock or out of stock, you'd be surprised how many things on certain websites are actually driven primarily by JavaScript. Things that would appear to be static in many respects, especially in e-commerce, they're actually driven by JavaScript. So, you know, if I were an LLM and I went to Levi's and the Levi's marketplace page was driven by JavaScript, I would just see a bunch of white blank squares, for instance. The importance here is that in this time, there is this need for static, it's called server side rendering, where the actual page and the responsibility to load the page is done just primarily through that HTML, primarily through the server. You don't need to wait, execute anything for that page to load in full. Most pages, most pieces of content aren't so dramatically impacted. But it's important to audit that. It's important to know. Atlas already has the ability because it's browsing through your browser to view these pages and it uses computer vision. But saving computer vision, it'll be a few months or years before we really do see prominent JavaScript rendering. So for now, big technical blocker.
Jonathan: Follow up on that. Do you think that that has any impact on the FAQs and the listicles being so popular in the ecosystem?
Josh: Absolutely. God, yeah. Yeah. Server side rendered.
Jonathan: Server side rendered.
Josh: Totally. And you know, I've seen people who have FAQs who are, it's actually JavaScript rendered, and so it doesn't help, like they'll be like, we just did this test. I'm like, folks, I disabled JavaScript and dev tools. We gotta go back to the drawing board here. We gotta figure something else out here. And I wish LLMs text got picked up more often 'cause that's what I would do. I would just be like, oh, let's make a markdown page. I so dream of a world where that is a situation where we can build a parallel architecture and that actually works. Right now, it doesn't. The markdown pages, the dot text page is really primarily, they don't get cited in the answer engines unless you can make a dot text page rank better than the normal page on SERP API and all these different search APIs for whatever black magic you can do, which you're doing that, you're really deep in some very gray hat stuff. So I wouldn't even recommend trying, but if you could, maybe then you'd have some inroads. But anyway, yeah, you get the point.
Elena: Josh, this has been really fun. I followed more of it than I thought I would, so thank you for the way you explain things. It's very clear and I appreciate that. I wanted to wrap up with something kind of fun. Do you have a piece of technology that you really love right now? And then is there one that you think is a little bit overhyped?
Josh: This is a spicy take. My favorite piece of technology is just my date wristwatch. I love it. I think that the most incredible products are the products that are durable, products that are omnipresent, universal products. I think you can't understate the importance of time in society and the nature of time. Like time is something that humans had to invent. I literally can't live without it, and I think the beauty and the craftsmanship of a wristwatch is something that I wish I saw more often in software. I'm always inspired when I look at that thinking, the actual features and the actual elements of this watch inside the case of the watch are totally hidden to me. But I know that if I went in and I unscrewed it, it would be, it's absolutely gorgeous. It's machined to perfection. So much right now, I think, in software, we're doing things very iteratively. Like the scary thing about software 20 years ago is software used to come on a disc. If you didn't get it right, the people would take the disc, they'd put it in their computers, and you're screwed. I mean, it's such a different world. And I think the fun thing is that we're building software, like we're building it for the disc, but we're building it at hyper speed. You know, if you were a software consumer 20 years ago, you'd get a pallet full of discs dropped off. That's how much we're shipping here. Maybe I could have chosen something more new. Maybe I could just said I love, you know, Claude Opus 4.5, but the truth of the matter is that the old things are often the best things. And there's still a lot of beauty out there that's not even AI related.
Elena: Yeah. No, I love that. What about, do you think anything's overhyped right now? Any new piece of tech you've tried or seen?
Josh: There was something recently actually. I think everybody's just a little bit crazed about agents. I think agents are cool, but I think an agent is a consequence, it's a concept. It's let's put a few LLMs together and give it some context. It is not a new way of doing business. It is a natural extension of having something that can simulate large language and speak back to people and reason through processes. I think we should not under index on how cool that is as a process, but I don't think we should treat it like, are we doing it agentically? Isn't it agentic product? Do we have agents doing this? Everything should be agentic and we should just be quiet about it and keep moving. I think that's my perspective right now is agents are totally priced in. It's, that is the A1 standard of doing this. It is not a feature. It is how it is done. Yeah, that's my perspective at least.
Elena: Love it. All right, Rob, Jonathan, did you have a quick one? Quick tech?
Rob: I've been in a cult since the mid nineties called Apple, and you talk about shipping software on a disc. I used to stand in line at the Apple store, like it was a rock concert waiting to get my disc of, you know, Snow Leopard or whatever animal they named it after. So as someone who's been such a diehard fan boy, I just recently, and I'm embarrassed to say this, I just recently switched from Safari to Chrome and my life has changed, and I know that sounds ridiculous, but I just refused to use anything but Safari and I get it now. I get it. I'm out of the cult. I'm like, okay, I've been in this bubble way the hell too long, this walled garden. I was not only in the walled garden, I was smoking the walled garden. I gotta get out. So I'm out on that. I love Chrome and I would say overhyped on that theme, any Apple announcement in the last five years has been overhyped. I'm just sorry. I've just, I'm throwing the darts. I'm out. Sorry. You know, Tim Cook is all about money. He's not about innovation, saying, I just, I missed Steve Jobs. I'm glad Tim Cook's leaving. There you go, Jonathan.
Jonathan: That is a spicy one, Rob. Okay, so mine's gonna be actually AI model related. On the, obsessed right now, there's a new image model that came out recently that's called Z Image. Very interesting. It's open source and you can make pretty much anything with it. It's crazy what it can do and it's fast. Like you can run it locally and get dozens of images, right? Like you can basically put together an entire book's worth of images in maybe 30 minutes with a good graphics card. It's insane. I was actually using it to illustrate one of the Chronicles of Narnia, but it was able to literally illustrate the entire book in 30 minutes, which is insane how fast and efficient this new model is. Overhyped, I'm gonna say this, Flux two Dev has not lived up to my expectations. It's huge. It's hard to run. It takes forever to get an image, and the images I get out sometimes aren't as good as Qwen Image Edit or even this new Z Image model. So those are my two things I've been absolutely obsessed with.
Elena: Josh, thank you so much for joining us today. This has been so fun and interesting. Where can people follow you and learn more about what you're doing at Profound?
Josh: I am on LinkedIn, but Profound's website is try profound.com. But check me out on LinkedIn. All my research goes on LinkedIn. We do have a research hub on Profound, but it's just literally what I did on LinkedIn, but then with more words. So really if you want to get insights into how AI search is moving, the latest and greatest in the field, check me out. Just search me, Josh Blyskal, just like the name.
Elena: Love it and I'll go ahead and plug Profound. It's a really cool tool and if your marketing team is looking to take action on AEO, that can also be a good thing to do is go talk to the team there.
Josh: Yes.
Elena: Thank you.
Rob: Thank you so much, Josh. That was so interesting.