
Ahead of the Game · 2026-06-19 · 50 min
Key moments - from our scoring
Substance score
61 / 100
Five dimensions, 20 points each
Crystal Carter, with nearly 20 years in digital marketing and current role at Wix, explores the fundamental shift from websites as destinations to websites as resources in an AI-first world. While Google still dominates with 82-83 billion monthly visits, AI search tools (ChatGPT, Claude, Gemini, Perplexity, Grok) collectively saw 27 billion visits in Q1 2025, representing explosive growth. However, the behavior differs - AI searches handle complex, multi-step queries that would previously require dozens of Google searches. Carter introduces three critical adaptation strategies: knowing your customers deeply through actual sales calls and CRM data rather than generic keyword research; making websites "do-thingsy" by adding actionable elements (forms, purchase buttons, calls-to-action) on every page; and preparing for agentic commerce through protocols like the Universal Commerce Protocol (UCP) and Agentic Commerce Protocol (ACP). She details Wix's implementation of these systems, including NL Web protocol and one-click agentic shopping configuration. The episode addresses the transition curve marketers face - similar to the shift from CDs to MP3s to iPods - and emphasizes that optimizing for humans first remains the foundational strategy, whether for algorithms or agents.
Google still gets 82-83 billion monthly visits with flat growth, while combined AI search tools (ChatGPT, Claude, Gemini, Perplexity) grew from 15.6 billion visits in Q1 2024 to 27 billion in Q1 2025. The growth is explosive for AI search, but behavior differs - AI searches handle complex multi-step queries that would historically require many Google searches.
Make your website "do-thingsy" by adding actionable elements (forms, purchase buttons, contact calls) on every page, implement commerce protocols like UCP or ACP through your tech stack, and configure your site for agentic shopping if possible through platforms like Wix.
Traditional SEO foundations remain critical, but add a relevance layer on top by understanding how your content appears in ChatGPT, Gemini, Claude, and Perplexity. Build content from actual customer questions in CRM, sales calls, forums, and comment sections rather than generic keyword tools.
Adoption is early, but the trend mirrors Amazon's one-click purchases and Alexa shopping - as integration deepens across Google Home, Android Auto, and Gemini, and trust builds, frictionless agentic commerce will likely follow.
Humans browse destinations; agents do stuff. Agents require declarative information (about things) and imperative opportunities (to do things) - your website needs both to be useful to agentic systems, with clear paths to action on every page.
Our reviewer’s read on each dimension, with quotes from the episode.
The episode contains several genuinely useful non-obvious ideas - preempting hallucinated URLs, the training-data vs. RAG content distinction, semantic triples for disambiguation - but these are diluted by significant filler, repetitive affirmations, and repackaged basic SEO advice given a light AI veneer. The ratio of actionable insight to conversation padding is moderate at best.
LLMs will hallucinate URLs and if you look at the hallucinated URLs for LLMs, they very often make logical sense...we didn't have one that was domains, we had one that was domain. So we made up a new one that was domains and redirected it to the other one
Gone are the days where you can just write a, you know, what is a platypus article and just expect the traffic to roll in
There are a handful of genuinely fresh angles - preempting LLM URL hallucinations by building redirects, the framing of PR syndication as training data for LLMs, and the declarative vs. imperative website structure - but most of the episode recycles the classic three-pillar SEO framework and standard 'optimize for humans' advice simply relabelled for an AI audience.
make your website do thingsy. Agents do stuff. Your website needs to be do thingsy
when you think about your pr...make sure that the phrasing benefits you, the phrasing will benefit the kinds of, the kinds of ways you want to be spoken about in LLMs
Crystal Carter is a genuine practitioner operating at meaningful scale - Head of AI Search at Wix with 300M+ users, citing first-party data from their AI Search Lab, and referencing concrete implementation work (first CMS signatory for ACP, one-click NL Web, Stripe/PayPal agentic shopping integration). She is not a pure thought-leader; she has done the work, though her role is communications rather than deep engineering or P&L ownership.
we are a CMS. That is the first, we're the first CMS that is an official corporate signatory for the ACP
I've updated content and I've seen it reflected in ChatGPT and you know, within 30 minutes
The episode provides real data points (82 - 83 billion Google monthly visits, 27 billion AI search visits in Q1 2026 vs 15.6 billion the prior comparable period from SimilarWeb), named protocols (ACP, UCP, NL Web, WebMCP, LLMs.txt), and concrete anecdotes like the hallucinated URL redirect fix and the acronym disambiguation failure in Gemini - enough specificity to be credibly actionable, though one period reference appears erroneous and some claims remain at assertion level without sourcing.
Google gets around 82,83 billion visits every month...Q1 2026 there were, there were 27 billion um, visits to AI search. In Q1, 2021 there were only 15 uh, point 6 billion uh, visits to, to AI search
we didn't have one that was domains, we had one that was domain. So we made up a new one that was domains and redirected it to the other one
Will Francis adds genuine value by pushing back on the AI search hype narrative and asking for clarification when Crystal uses jargon ('do thingsy'), and he occasionally contributes his own observations rather than just teeing up answers. However, he rarely challenges Wix-centric claims, lets several assertions go unprobed, and the conversation frequently drifts into mutual affirmation rather than productive disagreement.
but a lot of that uh, behavior is not what you would be doing on Google. Right? I mean if you want to know, um, plumber near me, you put that into Google. I don't think many people are asking that
But that's a pretty, I mean that is quite niche. I would never do that. Anyone that does that.
Computed from the transcript - who did the talking, and the words that came up most.
Search is changing faster than most marketers realize. In this episode, Will Francis speaks with Crystal Carter , Head of AI Search and SEO Communications at Wix Studio, a role focused on helping brands understand how AI-powered search is reshaping online visibility and discovery. They discuss the rise of AI search, the growing role of AI agents, and what these changes mean for brands trying to stay visible online. As AI-powered search transforms how people discover and engage with information, Crystal explores the emerging trends marketers should be watching closely and why some long-held assumptions about search may need to change. Crystal’s top 3 tips for marketers: Measure your AI visibility: track how much traffic is already coming from AI tools and monitor how it changes over time Make your website more actionable: ensure every page gives users and AI agents a clear way to take action Check how AI tools describe your brand: test prompts across multiple LLMs and look for consistency in how your business is represented The Ahead of the Game podcast is
Transcribed and scored by The B2B Podcast Index.
Speaker A: Welcome to Ahead of the Game, a podcast brought to you by the Digital Marketing Institute. I'm your host, Will Francis, and today I'll be talking to Crystal Carter, all about AI search, AI SEO, AI visibility. Crystal is the head of AI search and SEO communications at Wix and an SEO in digital marketing professional with almost 20 years of experience. Her, uh, clients have included people like Disney, McDonald's, Tomi. She is an active AI and SEO communicator. She hosts webinars, keynotes. Her expertise has been featured by Google, Microsoft, Mars, Semrush, and, um, more. Kristel, welcome to the podcast.
Speaker B: Thank you so much for having me. It's a really real pleasure.
Speaker A: Yeah, no, it's great. You are someone that we really need to talk to, clearly, because AI SEO, AI search, AI discovery, algorithmic discovery, all these things, you know, how people stumble across us and how we increase our surface area for that I think is crucial to any business today. Right. So let's just start by explaining what to you, what those things that I've just ratt settled off are and what you're working on there at wix. Around those things.
Speaker B: Yeah. So I'm the head of AI Search and SEO Communications. And so my role is essentially I communicate the value of AI search and SEO to Wix, uh, users. And we've got over 300 million users worldwide and I communicate the value of WIX to folks who are working in the AI search and SEO space. And with regards to what SEO and AI search are, essentially it's about building your visibility, building your um, your footprint online and making sure that people can find you when they need you. So SEO stands, uh, for search Engine optimization. And that is where you're sort of optimizing your website so that robots, um, like googlebot for instance, can crawl your website and index your website and so that humans want to go to your website. Right? So you're writing content so that humans want to see it and you're, you're making your technical configuration so that bots can see it. Um, with AI search, what happens is you're essentially adding in another layer onto top of that, where you're thinking also about how you're visible in things like ChatGPT and Gemini Claude Perplexity Deep seek all of these different tools, um, and how you're, and you're essentially adding another relevance layer on top of, on top of your SEO layer. So all those tools still use SEO, but then they'll add on another layer on top of that. And you said things, um, that we're looking at, we're also looking at now is the agentic element which is another layer further still on top of that because digital marketers don't have to wear enough hats. Um, and essentially it's where you're sort of looking at making your website more ready for agentic users who essentially use the website, the web in a slightly different way from humans, but potentially are making decisions on behalf of humans.
Speaker A: Interesting. And I mean you know, obviously you've been in the game a long time, over 20 years and you've seen a lot of change. Um, just about this current hype about AI search, is that a bit overplayed? I mean largely is Google still by far the biggest is player?
Speaker B: So we have an article on the AI Search Lab which is an AI search, uh, communications hub that I uh, envisaged and launched um, in 2025 with my team. And essentially we have a thing on there called um, AI search versus Google. We've been tracking this. Okay, so um, Google gets around 82,83 billion visits every month. And that's been the case for a while. But what we tend to see is that if you look at it over a 12 to 12 month average, the growth to Google doesn't seem to be that much different. It's pretty even. And actually if you pull it back over the last 12 months from Q1, um, so Q1 2025 back to Q1 2024 and then the 12 months before there was actually ever so slightly according to our data anyway, an ever so slight um, decrease in traffic could be margin of error, but essentially it's not growing significantly according to the data that we have from similar web. What we see on the other hand though is that if you look at um, if you look at searches that are made on Claude, Gemini, chatgpt, Grok, Copilot, Perplexity, all of those combined, the amount of growth that they're seeing is immense. So Q1 2026 there were, there were 27 billion um, visits to AI search. In Q1, 2021 there were only 15 uh, point 6 billion uh, visits to, to AI search. So that's like percent increase which is, you know, that's significant.
Speaker A: Yeah, I mean, but a lot of that uh, behavior is not what you would be doing on Google. Right? I mean if you want to know, um, plumber near me, you put that into Google. I don't think many people are asking that. ChatGPT and isn't most of the time people spend with AI different kind of behavior?
Speaker B: So that's definitely That's a question that I get a lot about this particularly when I started looking at that, that research. And what I would say is that the, arguably the AI searches are worth more because if I go to an AI tool and I say this is my cv, help me write it, then that's, that's a search that you wouldn't have made for you know, example of a marketing executive CV or how to format a CV or like all of those search. So you would, you would traditionally like you know, if we spin it back five years ago, you would have had to do, you would have had to do a bunch of searches to get that uh, to get that content, to get that information.
Speaker A: So it's replacing searches but not like for like it's, it's replacing a lot of searching with a different behavior behavior because we've got these new tools so there's different tools.
Speaker B: And also the other thing is, is that like it's constantly evolving and, and it's, and the tracking this data is, is also really tricky because Google is becoming more and more AI first. So there's AI mode, there's AI overviews and Google IO, um, they've launched more uh, AI features coming into, coming into search. Additionally when you go into um, when you go into some of these tools, many of these tools they have lots of features within them. So for instance we have a Wix MCP to use Wix within, within Claude, within Perplexity, within ChatGPT. Historically speaking if I wanted to go to my WIX website I may very well Google you know, Wix.com to get to it. So that would, that would reduce the number of, of clicks that we're getting from, from Google. Additionally um, you know one of the top things that people Google is Google um, and, and so then there's lots of different tools. Like I can't remember how many times I've had to Google um, you know sheets for Google or, or you know, Google Docs or that, that sort of thing. And so those searches are going to go down because people are just able to access them within, within Claude or within whatever tool that they're using. So um, it is, it is tricky but the way I like to think about it is this. I had a look at, at the transition between, between different types of um, music, um, formats. In high school I had a Discman and I remember a friend of mine showing me like a first gen MP3 player and it was just this little bo and it looked like a circuit board and he was like oh, you can Put you can play like 400 songs. And I was like, where are they? Where are the songs? I don't understand. And he was like, they're all in these files. I'm like, where am I gonna get all these files, dude? Like, I don't know what you're talking about. And, um, and it didn't make any sense to me, right? But then the ipod dropped and the ipod was like two buttons and they had the whole infrastructure that was ready to go and everybody was using it and it was really, really easy. And so essentially, like when we hit that inflection point for AI searches, you're going to have fewer and fewer people. Like even, even accessing, you know, classic Google, you're going to have fewer, fewer and fewer people using that because it's easier to use AI because you literally just vocalize to your phone or your car or your refrigerator, um, or your watch, what you need. And they would just give you the information. And, um, they'll give it to you in a way that you understand. So if it's too complex, you can say, I don't know what that means. Explain it to me. And it will. And so I think that, that for marketers who are trying to navigate this, this transition phase, because essentially Google's in beta mode right now. It's like it's, you know, they're testing everything all the time. Yeah. And if you're trying to navigate that, I think it's worth, um, you know, putting, putting a few plays in, in terms of the AI, the AI search, monitoring it you can monitor in a couple of different ways. You can either buy one of those tools or we have a tool built into wix, for instance. That's the AI visibility overview. But you can also monitor it with a Google sheet. Like you can, like you can do that too. Um, so just keep an eye on which, which, uh, queries are shifting, where the traffic's coming from. And you can also see your traffic in GA4 as well.
Speaker A: And there's actually a free, um, Google, uh, looker Studio dashboard someone created and has gone around it breaks down your AI traffic taken from your GA4 and just presents it in a nice neat way. But um, that's really interesting because of course yesterday was Google I O, they, they launched this thing called Spark, which is going to roll out their version of OpenClaw, you know, these kind of always on 24. Seven agentic AI systems that work away on your goals in a persistent way. And so does that throw up more opportunities for our, uh, business and our website to be found, do you think? And um, if so, how do we adapt to that?
Speaker B: Yeah, so I think that what people need to realize is that we are shifting from websites being a destination to being a resource. Okay. So for an agent, one of the things that people, you know, the people hear the word agentic and the agentic web and AI agents and it all sounds super futuristic and it sounds super techy and all that sort of stuff. But at the core the TLDR about agents is that agents do stuff, right? So this thing that you're talking about, you know, Gemini Spark, it's on all the time, it's trying to help you achieve your goals. So let's say one of your goals is to increase your protein and grow more lean muscle mass as many people who are my age are pursuing at the moment. Then it will be constantly looking at new information about, about different things that you can, that you can do. So then it would say to me, hey Crystal, there's this new thing that you should be doing. You should be doing peptides now, not
Speaker A: just collagen, now comes to you and says, hey, uh, you didn't ask me to do this, but I've found this thing because I know this is one of your goals and it's interesting, isn't it? It's proactive, right?
Speaker B: So with that, if you are a brand, then there's a couple of things you need to do. One of, one of them is you need to know your customer inside and out, right? Because any of these agents are going to know them. Essential for these agents, like they are working on behalf of the human orchestrator. If I have an agent, it's essentially like my Alfred, like I'm Batman, it's my Alfred and they're doing my bidding, right? So if you as a, as a business don't know your customers, don't know who they are, don't know the kinds of ways that they behave, the kinds of, you know, the kinds of things that they do, the kinds of things that they're worried about. Then you're not going to be prepared for this moment, right? Because, because the bespoke is going to be the benchmark, right? People are going to be expect to see, to see, you know, bespoke custom, really, really highly tailored experiences both on websites and you know, and in their, their interactions with, with each other. So that's super important. Know your customers, use your customer data instead of, Instead of pulling FAQs from some, some SEO tool somewhere, pull your FAQs from your actual customers. From your sales calls, from your um, you know, from your user voice recordings, from the questions that people email you in your CRM. Um, I use the Slack thing. We have a monthly webinar that we host for um, on the WIX SEO hub and we get questions from there. I use those questions and we write blogs on them. We have a forum and I know that our forum is full of people going what about this? What do we do here? How do we do this? We answer those questions in our blogs to make sure that people have the answer so that when they ask me I can just give them the link and they go oh great, thank you.
Speaker A: Well it sounds obvious but uh, I do the same thing in my content with comments or comments on my competitors content as well.
Speaker B: Absolutely.
Speaker A: Because that preempts. It's not just you're being good and nice and customer centric and all that. Of course that's lovely. But it's also a very accurate predictor of what people will probably be going searching. Right, yes. Um, uh, and in a lot of ways gives a lot more detail and nuance than sort of traditional keyword research. It gives you more specific pain points, um, and things that just lead you to more detailed relatable content. Um, so yeah, no, that's, that's interesting idea about, you know, just to, I want to just really highlight that thing you said about see your website not as a destination but as a resource.
Speaker B: Yeah.
Speaker A: And I, I can, that, that's, that's a very good way of seeing it, you know, because none of us do things enough. Uh, what I mean by that is I play tennis, right, Mike or my chatgpt knows I play tennis and I can imagine a world in which that starts to go, hey, you haven't asked me about grip tape for three months. And you really should be changing every three months factors. I only change every six months. So you can have these proactive AIs saying hey, you should be reordering that supply of this or that and then it's going to go well, I can find that for you. And it's going to go to the easiest source. It's going to go to the first website that comes up that's got a good market price on that and you want that to be your website, but it also might need more information and might go to your content, uh, for that. So yes. Do you have anything, do you have any guidance on optimizing for that? Or is it just, is it the same SEO tactics, um, essentially that we've talked about for many Years.
Speaker B: Yeah, so this is great. So that's a great example. So, so this brings me onto the second thing that websites need to do. Which is, which is to make your website do thingsy. Agents do stuff. Your website needs to be do thingsy, right? It needs to be a place where people can do something. A place where agents can do something.
Speaker A: What do you mean by that?
Speaker B: So what I mean by that is that basically like let's say your tennis, your tennis example, you've got a blog that talks about, talks about grip tape, okay? But uh, maybe there's no button, there's no form to, to, to buy it. There isn't an opportunity to buy the grip tape on the blog. It just talks about grip tape. I've seen this a million times. So make sure it's there. Maybe there isn't a place where they can a button where they can click and call you. Um, um. So in, in agentic stuff they talk about imperative and declarative. Declarative is where you tell someone about something. Imperative is where you give them the opportunity to do stuff. There's something called Web webmcp and within webmcp you can declare which parts are the parts that the website can do things on. Uh, uh, so in the WIX and in um, and in something called NL Web, for instance, you can search on a, on a page that doesn't even have a search bar, for instance. So for instance, if you have a website that doesn't have a search bar, you're also less do thingsy if they, if you don't have um, you know, if you've just got a brochure site that's not, that's not useful necessarily. That's they'll end. The other thing is, is that you were talking about, you know, where they can order it, the easiest place to order it in terms of again making it more action actionable, making it a more interactive. If your website is configured for agentic shopping, then they will use you. This is another buzzword is called interoperability. You'll hear that a lot. And that just basically means that it works with, with different systems together. And essentially if your website is configured for like the UCP which is a universal commerce protocol which Google is using, um, or is configured for the acp which is the gentic commerce protocol, then you're going to be more likely to be able to, to be used by agents because agents can use those things. So in the Agentix space you'll hear all this talk about this protocol and that protocol and that protocol what's happening with those protocols is essentially we're, people are rewriting the web for agents because agents, they need to, they want to be able to do stuff but they can't just go out in the web like just willy nilly, they can't just go and do whatever. They need guardrails and that's what these protocols do. So these protocols have handshakes, they have ways to verify, you know, whether an agent is a, is a real agent and is a good agent. They have um, they have ways where you can protect private information and things like that. So essentially one thing everybody can do is to add more actionable points on your website right now.
Speaker A: Do I do that? Is there a one and done plugin for that?
Speaker B: You know, so the forms and stuff like that you can do, anybody can do that. You don't need to, you don't need to be a super techie whiz person to do that. Um, all of these protocols and things, a lot of these are happening at scale. So for instance like at wix, we are implementing this for people. So there's something called NL Web, which is a protocol from um, from Microsoft. We've built a one click implementation for that. You literally click it and we've built it in the backend and then you're configured for that. We are the, we are a CMS. That is the first, we're the first CMS that is an official corporate signatory for the ACP. We've configured something that works with Stripe, works with PayFab, PayPal for agentic shopping. We're working alongside Google for, for other protocols as well and things like that. So have a look at who your, who your tech stack is and where they are, where they are making moves to um, make you applicable for that. So like if you're using Etsy for instance, Etsy is set up for a lot of these things. Booking.com is set up for some of these things. Um, uh, ebay is doing a lot for these as well. So how the look, I mean I
Speaker A: remember a time when, and it's not that long ago when people wouldn't buy stuff on their mobile phone, right? They would be maybe on the train looking at products but they were like, no, I'll have to buy this when I go home because no idiot buys anything on a mobile website. Um, and I think uh, we're in that stage now. I mean, I don't know, um, I don't know many people that are buying that are letting their agents buy stuff. Their AI agents buy stuff on their behalf, you know, with free access to their credit card. But do you, do you think that's coming?
Speaker B: I think, you know, if you think about Amazon, the Amazon one click ah, like purchase like you like on Amazon, people just have, they just have the thing and people buy stuff with Alexa, um, and have subscriptions with Alexa and things and they just say Alexa, buy me, buy, you know, buy me a new phone charger. And she'll go okay. And then the phone charger shows up and people don't even go to a website. They just, they just get the phone charger.
Speaker A: Um, but that's a pretty, I mean that is quite niche. I would never do that. Anyone that does that.
Speaker B: But I think what I think is that, I think when it becomes more integrated and this is the thing that Google is really pushing for. They've talked a lot about um, you know, um, artificial general intelligence, which is what I think is their big vision for Gemini, which is you know, a single entity that's, that's looking across all of the different systems and essentially with, with these tools like Google already has my Google Wallet details. They already know all of this stuff. So and, and, and they're also configured for, for Google Home, they're con. They are. Gemini is the, is the, is the assistant for, for Google Assistant for Gemini Home for Android Auto now for Siri. So I think that once people get more and more used to it and, and because it becomes more and more convenient and are going to be more likely to use it.
Speaker A: Yes, I think we are, I think we're getting there. I just think we're in that stage where it's just nowhere near the mainstream. It seems a bit scary and a bit like giving an agent your credit card details. Are you insane? Um, but like we've seen with a lot of, you know, the evolution of trust in the digital world, I think um, we will, as it proves itself, we'll, we'll quite possibly get there. Hello. A quick reminder from me that if you're enjoying our podcast series, why not become a member of the DMI so that you can enjoy loads more content from webinars and case studies to toolkits and more real life insights from the world of digital marketing head to digitalmarketinginstitute.com ahead of the game. Sign up for free. Now back to the podcast. So practical, practical things for now are uh, how did you put it so that you can.
Speaker B: That's, that, that's, that's my, my official term.
Speaker A: Yeah. So that, so that every page on your Website people or agents, because remember, it's going to be better for humans anyway. If, if I can take action on any page on your website as a human, that's a pretty good thing.
Speaker B: Yes.
Speaker A: Because I'm closer to conversion.
Speaker B: Yes, absolutely, absolutely. And you get more data on like what's going on and what's interesting to people and what matters to people and all of the, all of that sort of stuff. It's um, it's, it's a win, win. You can't lose if you, if you do that.
Speaker A: Well, I'm always, I don't know about you, but I always think that anytime people ask me how do I optimize for the algorithm, how do I hack the TikTok algorithm, the Google algorithm, I just say, look, it's all about humans. If you optimize for humans and humans are seen to engage, dwell, do, et cetera, then the algorithm will be your best friend. Um, so I'm always a big fan of that, you know, optimize for humans first. Um, but how, how, what do you think about that in, you know, in terms of what we've talked about here?
Speaker B: So I think it's interesting that you talk, you talk about TikTok and you talked about, you know, videos and responding to comments on videos is. Because I think this whole thing is going to become a lot more algorithmic. So essentially like at the moment we have like page one of Google, like oh, are you on ranking on page one? And it's like, and, and, and it's become more and more and more, but it's been for a while where it's like which page one? Right, which page one are you talking about? Because my page one is gonna be very different from yours, especially going forward. So like when I go on to, when I go onto YouTube, for instance, my algorithm is completely different from my husband's. Right? Mine is all like puppies and Beyonce and like that's, that's what my algorithm is and my husband's is like news and sports and all of that sort of thing. And, and what we're going to see with um, with the, the um, Google's rolling out something called personal intelligence that's, that's bringing through like your YouTube data to give you, to give you recommendations on, on things and bringing through like your email and all of that different stuff. So when you're engaging with comments, engaging with people and then driving the algorithm and engaging with humans essentially like this, this is where it's headed where, where um, brands on the webs, on the, on the web are going to need to think more like content creators who have been, who have been working with this like very human led, very engagement led, um, ah, space in social. Right. So people who've been doing like social media management, like, they know how, they know that, like, I need to be very strong on my niche. I need to be very engaged with my audience, um, I need to interact with my audience, they need to want to interact with me. I need to have regular touch points and things like that. You can't just publish one thing and hope for the best. You have to engage and things like that. And so I think that it is more involved, uh, and I think it is gonna change how we interact. But as you said, it does come back to people and it will come back to those interactions. And if the, are interacting with you and the people are engaging with you.
Speaker A: Yeah.
Speaker B: Um, then, then the agents will have more confidence that you are the person that they should recommend. Because, because based on the previous, previous interactions that they've had with you, they liked you. So they'll say, we'll send you back to Will because Will did that thing, you know, he, he was great the last time. So you should, you should go back to Will because Will is there and he can help you and that sort of thing. Whereas, you know, if they're just having to go out of nowhere, then, then they, you know, they're going to be less confident, they're less likely to recommend.
Speaker A: It makes me think that, you know how there are three pillars of SEO. Technical SEO or content or on page SEO as people call it, and then off page or reputational SEO. Right. And it seems to me that we're talking about this and I'm thinking, well, those things are all important still, right? We have to have a technical SEO. We have to have a website that works, that's crawlable, that's usable by, you know, machines and humans, content SEO. We need to have a website that's got, got useful, helpful content on, and also really clearly defined services and products and also reputational, uh, resio. We need to be seen to be an engaging entity on the Internet and important because people engage with us and also mention us and link to us and all that kind of stuff. So you're the expert. I mean, you're far more of an SEO expert than me. Would you agree that that's a good definition of SEO? Sort of the last time.
Speaker B: That's a classic framework.
Speaker A: Yeah. Right. So in, uh, each of those three buckets, um, of work, essentially they're three work Streams aren't they? Essentially what uh, do we need to do to make ourselves to set ourselves up for this future? I mean let's start with technical. What are the extra things over and above the obvious of having a good website that loads quickly and is crawlable, et cetera? What are some things there people can do technically?
Speaker B: So for the, for the agentic element there, there are lots of very specific things. So again I mentioned like protocol making yourself more relevant for protocols. Thinking about structured data for instance is, is really relevant if you have NL web enabled. And also we've seen lots of data to suggest that um, structured data can be really useful for RAG based queries. So RAG being retrieval, augmented generation which is essentially when um, an LLM will go and do a search and they will tend to do this for newer information. So um, if I, if I was to say what happened at Google yesterday, they would have to carry out a fresh search on that information. Most of the technical things uh, that you do in SEO to be crawlable and indexable are standard, are the same for LLMs for agents. There are additional things that you can do in terms of making the content really, really easy to digest for agents. So for instance people are experimenting with markdown. People are using um, you know, JSON, uh like well known JSON files to um, expose like REST APIs and different endpoints for agents. And there's also something called LLMs T which has been very controversial within the SEO space. However developers and Google have recently added um, LLMs Txt verification to their agentic browsing um audit within Chrome dev tools. And the reason why is because essentially you can, you can use it as a file to give agents to ground them in information. So this means that it costs less for the agents to do their job. So those are technical things.
Speaker A: So that's data structure a lot of what just to just to really break it down to smart and less technical listeners. Those things you've just talked about there are all about structuring data to make it more machine readable. So JSON for instance is a way of structuring um, information in essentially organized chunks so that it's quicker for a machine to go through and find the bits it needs. Or like you say you talked about, oh LLM Txt, which is essentially just a text file you have in your root folder of your website, which is a quick overview of what, what a machine would need to know about this website without it having to crawl through the entire website to make its mind up what this site's about. So things like that is fair?
Speaker B: Uh, yeah.
Speaker A: Um, right. Cool. Like it. So then in terms of content SEO, we've always been told, like I say, to have you know, content that is easy to read, easy to access and then to continuously update with, you know, like blogs and articles, uh, to target largely long tails, more niche specific keywords that we can see respond to what people are searching for. Um, obviously. So oversimplification. There's a lot more to say there.
Speaker B: Yeah, of course.
Speaker A: What do you think, what else do you think we should be doing in terms of content SEO to make life um, easy for these agents?
Speaker B: Um, so, uh, um, for agents in terms of content, I don't think that that's too much like too much of a big deal other than making sure that your CTAs are really clear for all of the buttons and actionable elements across your website. Um, and making sure that you' got good logical navigation. Um, additionally on the content point, I'm not sure if this counts as content or if it counts as technical. But accessibility. So again, that Chrome, that Chrome developers um, audit has um, sorry, Lighthouse audit for, for ah, for agentic tools also has an accessibility tree in it because agents use area tags to navigate a website. So in the same like your, your H1s, your H2s, your alt text, all of that sort of stuff, your end and mouse free screen reader navigation, that's the same thing that agents use to navigate a website, particularly for like agentic browsing and things like that. So making sure you have good accessibility on your website is really important. Making sure that you have clear, uh, clear logical navigation is really important. Thinking about your URLs is actually really important. So LLMs will hallucinate URLs and if you look at the hallucinated URLs for, for LLMs, they very often make logical sense. Right? So we had one where it was saying like um, you know, oh, this is where domains. And we didn't have one that was domains, we had one that was domain. So we made up a new one that was domains and redirected it to the other one.
Speaker A: That's such a good point, you know, because I see that all the time, all the time, um, hallucinated made up URLs uh, that are linked to in LLM content. But you're absolutely right, they kind of imagine the most predictable URL, don't they? Like what it would logically probably be in its simplest way. So that's such a good idea actually to sort of imagine preempt what they might be. Suppose what I was asking was like, should we still be creating article content and going after those long tail keywords, is that still important?
Speaker B: Yes. Okay, so here's the, here's the other thing. Okay, so in terms of the type of content you should make, essentially, um, if we were thinking about, there's two different types of uh, LLM responses. If you're thinking about responses, there are, there's um, training data responses and there are, there are RAG responses. A training data response is essentially the LLM coming off the dome. So it's whether, using their, their, um, their training data, the information they've been trained on. So in the PT and ChatGPT stands training. Right. So that's. So they, they read all of the Internet and then you ask them a question and they will give you an answer. If I was to ask an LLM, why is Dolly Parton a great musician? They don't need to look anything up, right? Because all of the trading data that they have will cover lots of reasons why Dolly Parton's great. If I was to say, you know, um, how did Dolly Parton, how has Dolly Parton influenced Beyonce's Cowboy Carter album? Album? They would have to look that up because that's something that happened fairly recently. Right? Um, so that, so, so the, the training data information that, the content you need to make about that is content that, that reflects who you are to your core. In SEO, we've called this content evergreen content previously. And so it's that kind of content. And in that kind of content you need to, you need to utilize the kinds of phrases that there's something called semantic triples where you would say like WIX is a website builder, Crystal Carter is an SEO expert, Will Francis is a podcaster, that sort of thing. And you need to make them really, really clear and you need to make sure that they disambiguate you, um, with, with you know, both the entity and who and the relations. So for instance, if I was to say Steve McQueen is a man known for his films, that's a semantic triple. That's cool. But that is two people, right? There's Steve McQueen from the 60s who drove fast cars and all that sort of stuff. Then there's Steve McQueen who's working now. So the way that you disambiguate, you say Steve McQueen is a, uh, is an Academy Award winning man known for his films, right? Because. Or Academy Award winning director known for his films. That's only one person. Okay? So you think about this with your brand I've had people who are saying, oh, I don't not showing in ChatGPT when you ask for like best marketing agency in Paris and then I look on their website and they say the same thing as every other website. Make sure your website says who you are, why you're important, important and that it's super consistent on your website, on your Facebook, on your LinkedIn.
Speaker A: We skip it because it's the basics, it's basic. But the sort of thing, well, I shouldn't have to say, I mean really so and so is a shop that sells, you know, you think, um, does this mean there's more of a role for like an about page now?
Speaker B: Yes, the about page is super important for training data responses. It's really important to have that. So the training data responses you need to have your brand. You're building up your relationships, you're building your relevance, right for who you are. So you need to make sure that you are in the right space and so that for your evergreen content, you're doing that for the rag responses. This is where you're more likely to get a click from an LLM and this, and to do this we need to be thinking more like reporters. We need to think, be thinking more like um, you know, more like content creators on, on platforms like TikTok or platforms like, like Instagram. So on um, TikTok, if you spend any time on TikTok, any news breaks, then there's loads of influencers who are immediately on it and who are like, I saw this, this happened. Find out about the information. I've got all the piping hot gossip, all that sort of stuff, okay? And similarly with, with websites where you will get clicks is where you have content that is both fresh and spicy. So fresh information is like uh, Google IO happened yesterday. Here's my, my hot take on what's going, you know, or here's, here's all of the information that they covered. People are going to get that and they're going to come to a website for that because they, they can't necessarily get that from, from an LLM. And then the spicy thing is your opinion, okay? LLMs don't have opinion opinions on don't, don't have opinions. And if I wanted to hear the hot take on what happened on Google I O and what it means for all of us, I wouldn't go to an LLM for that. I would go to a human for that because the human would be able to process it in, in an interesting, in an interesting way. So when you're thinking about the content for your website. It. If you're, particularly if you're trying to get traffic from LLMs and even traffic from, from humans, think about stuff that is like new and is pushing it forward and is, and you know, articulates your point of view. Okay. Um, and the other thing that you can do in terms of newness is something that is adding something new and exclusive to the conversation. I've mentioned some of the research that we've done on the AI search lab. That was research that we did. Nobody else did that. So if somebody's looking for that information they will. The what? The LLM will send people to us because we are the source of that information. No matter which um, business you're in, you will have exclusive information that, that you can share that they will have to cite you for because that you're the only person that has that information and it's, and it's interesting information that they would have to refer to you for. So the content is, the content does have value but it has serves a different, different purpose. Gone are the days where you can just write a, you know, what is a platypus article and just expect the traffic to roll in that, that traffic. You know that article's been written 17, 17 million times. Um, unless you're adding something that's genuinely new, that's fantastic, fresh and or spicy then you're unlikely to get a click um, from, from Google or from, from LLMs.
Speaker A: Well there's quite a lot in there that's so interesting to hear your take on it. When you, when people talk about rag, Rag can mean a lot of things but it's really where the LLM goes and um, gets information rather than draws on its existing training data, you know, its existing knowledge. Um, and you're right that's it's more likely to do that with topical things that it know, it knows are likely to be newer than its knowledge cutoff date, um, the date after which it doesn't know anything. So um, that is interesting and do you know what? I don't know if this is why but I've seen uh a data that and I feel like I see this anecdotally but it might be confirmation bias that Chat GPT does seem to really like fresh. It seems to like freshness when it cites websites in the way that you know Google will happily show a 20 year old website. Um, ChatGPT and Claude seem, do seem to be quite swayed by fresh.
Speaker B: Yeah, no, um, statistics uh, and you know research uh backs that up. So there's been lots of people who've been running research on, on this stuff and, and yeah it does seem to be some, something that's a fact. And, and also ChatGPT can pick things up really quickly. So ChatGPT has lots of content partnerships with different newspapers and, and also they are bing feeds into ChatGPT as well. So anecdotally I've, I've seen websites that use IndexNow, which is a tool that instantly indexes new content. Um, from, from websites that are enabled with IndexNow. Like all Wix websites have the ability to be connected to index. Now with that are premium websites. Um, for instance I've seen where I've updated content and I've seen it reflected in ChatGPT and you know, within 30 minutes I've seen where um, I've seen updates to, to um, other other pages and I've seen it reflected um, almost instantaneously in, in chatgpt. So um, it's, it can be, it can be really quickly. Um and it's, and it's something that's worth thinking about. But it's all. The other thing is, is that like loads of people are updating their content content a lot really quickly. The window between, between being able to rank and you know, your, your competitor making an update is getting closer, is closer and closer. So I think when I first started this, you know, you, you would expect to make an update to content and then you probably have a couple of weeks, couple of months maybe um, of ranking before, before your competitor caught up. But it, you know, it can be down to like you know, a few hours, few minutes these days.
Speaker A: Yeah, no, that's interesting. And the other thing you said was because there's been this whole thing around. You know, I get asked all the time in my workshops if I write articles of see do they rank. And often, you know, I'll quote that how Google changed the wording and their helpful content guidelines from we reward content written by humans for humans. And now it just says we reward content written for humans. Um, and it appears that Google maybe doesn't care whether it was written by AI, but what it cares about is it's not just the same regurgitated thing that like you said, has been already written about in the same way 90 times. Um, and so it's about information gain. It's about bringing new new original information. Whether that's uh, internal data, internal opinion, expertise, uh, more original research towards the edges of the Internet, books, academic papers, all that kind of stuff. Um and that seem, does seem to help. So that sounds like you think that's also important for appearing in LLM results.
Speaker B: Yeah, so I think that when people say, uh, oh, content written by AI. I think that people think that like, you literally just go to ChatGPT and just say, write me an article. And it goes. And then you just post it. And I think there's lots of different grays in between. So. So I've worked in content long enough to know that subject matter experts are busy and they don't want to talk to the content people. Okay.
Speaker A: Um, massive problem, very common problem. I hear that it's incredibly annoying.
Speaker B: Um, um, and like, you have to respect it because they're off doing the things that they're, they're subject matter experts on. You have to respect it. And you're like, I need a blog. And they're like, I don't care. So what, what AI allows you to do is that you can have a 20 minute phone call with this person that's really. And you can, and you can have a 20 minute phone call. You can use AI to transcribe the phone call. Um, you can use another AI to help you knock that phone call into an article. Um, you can carry out some research and then you can give it back to them. And then they only need another 20 minutes to make sure that you're not talking complete craziness. And then you have an article from M, that subject matter expert that you would not have otherwise. Okay? And that AI allows you to do that. That. And that is super, super valuable to everyone. Right? So I think that, you know, use all the tools. And the other thing is, is that like, use them well. Okay, so, um, people, somebody asked me what's a good tool for writing AI blogs? It's like, well, there's a million tools for writing AI blogs. What's really important is that you learn how to train whatever tool you're writing with to write them in a valuable way. So give them good grounding knowledge. So every time you brief an AI on, on, you know, the content, you know, give them the webinar that talked about it, give them the podcast from your team that talked about it. Give them the call, um, that you had with that expert. Give them the support documentation that you have on it. If you, if it's internal, use that, use that internal briefing. Give all of that to it to help it write the article. Don't just let, don't just, don't just send it out to the Internet, um, to, to come up with Stuff it's going to make up, give it, give it good grounded knowledge. Um, a tool like NotebookLM is incredibly useful for this. It's free, it's from Google and essentially you can add in lots and lots of resour, then you can use it as a knowledge base and it will only reference what's in the knowledge base that you give it. Um, um, and it can be super, super useful for marketers. You can use it, you can, you can use um, it with your clients. So you get a client, a new client and, and you know, let's say they're, I don't know, they're, they do circus tents and you know nothing about circus tents. You can take all of the information that they have, their PDFs, um, um, podcasts, websites, YouTube videos, everything. You can stick it all into a knowledge base and then you can ask it questions. I've done it myself with things internally. There was a couple of, a couple of projects where I'm the person who knows the most about the project and people kept asking me stuff and I was like, I don't have time, I don't have time. So I, so I did a couple, I did a bunch of videos and I talked m to myself for you know, a few hours and I stuck it in this notebook lm and now when people ask me, I will literally take their question, paste it into NotebookLM and it will reply as if it was me and I go, yes, that is exactly what I would say. And I give that to the person and it's super useful.
Speaker A: Well that grounding is, yeah, that, that is a big part of it, isn't it? Because the kind of, the flip side of all of this is getting AI to write stuff for you. You kind of, you may lose your grip on reality potentially. You know, it might start creating stuff that's not particularly true. But then also, is there not a problem of the uh, LLMs being manipulable? Are the LLMs manipulable? Could I, could I say that I'm the most award winning marketing and AI educator on the planet and if I say it enough times around the Internet, will ChatGPT just start to parrot that back at people?
Speaker B: I've seen that happen, but to be fair, I've seen that happen with Google. Like I've seen, I've seen people manipulate Google results. And you know, I remember back in the day there was this woman in new in Los Angeles and she put up a bunch of billboards all over Los Angeles. This is like in the 90s or something saying how great she was and, and you know, just trying to make herself like a famous person or something, you know. So I think that this isn't necessarily an exclusive to AI issue, but it's, you know, it's just a thing that happens.
Speaker A: Yeah, yeah, no, um, it's really interesting. And so that third pillar of SEO, reputational SEO, which we've. For years, for 20, 25 years, we've been trying to get more backlinks. We've, there's agencies who will just get you backlinks, you know, um, what do you think about that? What's, what does that job look like now? Is it important? What should we do?
Speaker B: So I think when I first started doing SEO, the name of the game was to get lots of backlinks. Didn't really matter where, where it was just lots and lots of backlinks. I remember our agency, we, we got a backlink on the, the website for the plant watering company that we were work, that we were working with. And we were like, well done us. We got a, we were a marketing agency. We've got a link on a plant watering website. And um, and you know, the name of the game now is relevant, okay, because if you get lots of links on a bunch of websites that are nothing to do with your niche, right? So let's say you are a plant watering business, but then you get a bunch of links on a, on a shoe website, that's not, that's not gonna make sense. That's gonna dilute the information that the LLM has. Has for you. Because they understand that website to be the other website to be about shoes. They understand your website to be about some, something else. And they haven't connected those two. Right? So essentially what you need to think about is, is building up relevance. Okay? So Google has the knowledge graph. They. And Google has, you know, and they're, they're trained on entities. And Google and all of the LLMs, for instance, are trained on Wikipedia. Wikipedia is basically just entity after entity after entity. And essentially if you look at that, what they do is they'll say like, you know, this is, um, you know, this is Beyonce. This is her, her husband. Husband. These are her children. This is da da, da, da da. Those are all the relationships that they have, all the correlations that they have, um, about her. There's a similar profile for you around, around your business. And it's really important that you use your backlinks to verify that, to help build confidence in the LLM, um, about who you are and what you're for. So for instance, if you were a business, it's worth making sure that you're listed not just on a general directory. Right. So not just on, on uh, like any directory, but on, like, let's say if you were a vegan restaurant, you'd want to be list happy cow. Because that's a website that's, that's niche and it's all about vegan food. So I think with all of the backlinks, it's not just about quantity anymore and it's not just about da even um, you know, domain authority, but it's specifically about. About which ones build up the relevance and, and help to build that brand story about you that, that you are this, you serve this audience, you provide this service and that's, that's sort of thing.
Speaker A: So is, but is PR still a good way to get backlinks because they have such high domain authority, but we want to just skew more towards relevant PR rather than any pr.
Speaker B: Yeah. And I think also with pr, so with pr, what you, what you also get is you very often get a lot of content syndication. So for instance, um, I, I've had coverage or, um, gotten coverage for clients, um, in my local newspaper. Um, but the local newspaper is a syndicate of the Daily Mail, for instance. Um, and so, and that they've got a huge syndication, they've got their own Daily Mail, and then they've got lots and lots of local newspapers, for instance. Um, and so just because we were in that one local one, we were also spread out to lots of them. So when you think about your pr, it's really important to think about the freezing that's in your PR because it will be parroted across the web again and again and again and again, literally verbatim, particularly for syndicated content. Um, make sure that it's ex. That the phrasing benefits you, the phrasing will benefit the kinds of, the kinds of ways you want to be spoken about in LLMs.
Speaker A: Okay, and is that about consistency? Is that about always, you know, really kind of hammering home the thing that you are. Ah, the thing that you do, the thing you're good at. And having that phrase always around the name of your business, it's.
Speaker B: It is about that, but it's also about the way the LLMs work. When you watch an LLM and it literally go. And it's putting out word, word, word, word, word, word. That is the LLM predicting the next word every single time. So every single time it's predicting the next word. If it sees Mary had a little. It knows that the next word is lamb, right? If it says like, WIX is a website, it knows the next site, the next word is builder. Wix is a website builder. It knows, it knows that because we've said that all over the web for like years and years and years. So, um, if it, if it's the case that they're constantly seeing, you know, know, um, Crystal Carter is a this. Crystal Carter is a that Crystal. They don't, then they don't know what to say about you the next time. And a really easy way to check, to check this is to ask a couple of questions about your business. You know, like who is, you know, who, who is your brand. And essentially if you ask it to ChatGPT, ask it to Perplexity, ask it to Claude, um, or whatever, you can see how different they are. If they're roughly the same same for all of, for all of these tools and you're probably doing all right. If they're inconsistent, then, then they probably aren't. An example I did the other day was it was a company that was an, that had an acronym, but their acronym was the same as an institute somewhere else. So Gemini was convinced that they were this institute. Perplexity and chatgpt understood, but Gemini was convinced that they were this institute. What? And so they need to make sure that they always say certain things to make sure that they're clarifying who they are and that they're not this other thing. Thing. Um, so you just need to make sure that there's certain things that make sure that like no, no, I'm this person, I'm not that person.
Speaker A: Yep. No, that's good. And again across website, PR and social and everywhere else, you know, so it's got that consistency. Right? Okay, that's great. God, we've got quite the to do list there. Uh, Kristal, I really appreciate it. Before you go, I'd love you to just give us a handful of your top tips that people can action straight away after listening to this to improve their visibility in AI.
Speaker B: Um, so the first thing I would do is check how much you're getting from AI right now. So go into your GA4, check out that, um, that looker, ah, Studio data, ah, Studio that Will was talking about. Um, so have a look and see what traffic you're getting from there. The second thing that, that I would do is I would make steps to make your website more actionable. So adding in a search bar, adding in more conversion points, adding in more forms, um, making sure that there's at place to convert or take action uh, across every page on your website. And finally I would take steps to to see how LLMs are referring to you now. So ask a few questions, the same set of questions across a few different LLMs and see if you're getting a consistent response about who you are, what your brand does, what you're known for, what you're offering is. And then you'll be able to to identify whether or not you need to take steps to, to improve uh, your visibility to, to disem and to be more visible on AI.
Speaker A: Amazing. Well, look, Crystal, thank you so much. Really appreciate your time. Hope I get to chat to you again soon. Thanks.
Speaker B: Thank you so much.
Speaker A: If you enjoyed this episode, subscribe wherever you get your podcasts. And for more information about transforming your marketing career through certified online training, head to digitalmarketinginstitute.com thanks for listening.
Other episodes covering the same guests and topics, from across The B2B Podcast Index.