Why Google Is NOT Dead & How To Dominate Ai-Driven Search ft. Ethan Smith

Is Google Search dead? Maybe not. We are going to tell you why Google Search may be the predominant search engine for many years to come and how you can get your brand to show up in ChatGPT. All in that and more on the future of search with our very special guest, Ethan Smith. I'm Kieran Flanagan, here as always with my good friend Kit Bodnar. Let's get into today's episode. All right, we are here with Ethan Smith, the founder of Graphite, which is an international growth agency. Ethan, welcome to the podcast.

Thank you so much for having me.

You are brains behind some incredible search strategies. I know you have the masterclass honey notion upwork like a who's who of search, but you are 100% and I'm not just saying this because you are on the show, the most knowledgeable person I've ever talked to in search. And so what we would love to do is try to pick your brains on the present and the future of search and SEO. Because we are wearing, we and lots of other marketers who are love search as our primary channel way for SMBs and mid market companies to really scale and grow are trying to figure out what is the future with all of the AI. AI changes come in but the place I would love to start and maybe could we just go to the start on the present, like present day and maybe take into context the two updates that have just been done for Google because they seem to cause a lot of thrash in the market. A lot of people are, you know, worried that their entire businesses built around search are going, are actually going to go away. Where do you think we are today like Google Search? Are you still very bullish on it? How is AI impacting it? Maybe could you just give us the perspective of like, where are we today?

I actually think where we are today is very similar to where we were a year ago and two years ago and not much has changed. I don't think a significant amount will change. Two reasons. Reason one is that I don't, I don't see people moving away from Google en masse. We have seen people, we have seen Google be disrupted somewhat by TikTok search, Instagram search, YouTube search and some of these other services. But Google search hasn't gone away or even gone down. It's just that there's been an increase in new searches on these other platforms and with these chat bots. We haven't seen mass exodus from Google going to these other services. And even if we did see people moving there, I don't think that Google search would go away. And part of the reason why is because of what you mentioned, which is people are not always looking for answers. A lot of times they're looking to do something like to find a restaurant or to buy something and an answer doesn't fulfill the intent. Actually if you look at answerbox it's true that a candle bear takes away some clicks from the publisher, but publishers who show up in Answerbox get twice the click through rate as position two. And still you also see lots of people going to destination sites with AnswerBox. And Answer Box doesn't show up all the time, it shows up when it should show up, which is, I don't know if you have data on that, but it's definitely less than half of the time, which means that answers are not fulfilling the intent on their own most of the time. Most of the time you need to go somewhere else. So for all those reasons, I don't see Google search hasn't changed dramatically and I don't think that it will anytime soon.

So in that answer, do you think Google will not be disrupted by AI chat because AI chat serves a different use case or that Google is so entrenched in how people think about search and find information that it's just incredibly hard to disrupt something that's so like deeply entrenched about how user uses or how user users think about solving that use case.

I think that search engines are the best experience for a large number of questions and queries, but not all of them. So for the questions and queries that have a better user experience, kind of like Instagram is a better user experience for things where you want a visual result and YouTube and TikTok are a better experience when you want short videos. And so for the queries and questions that have a better their intent is will feel better with that, then they'll move over there. But for other stuff searches, Google search is the best user experience. And for those I don't see that they would be disruptive. The only reason why that would be disruptive is if somebody had built a better search engine, which is possible, but very hard.

So you're saying right now that to you the search landscape is more continued search diversification where it's it's YouTube search, it's TikTok search, it's Instagram search, it is chat, GPT and other LLMs. But so far those are all additive that like humans are just searching and consuming more and there's more transactions and you're still seeing growth in the overall Google search traffic. Volume for like commercial keywords and, and all of those things. Is that, is that accurate?

Yes. And what I would say is that in order for that to be the case, the pie needs to be bigger. And usually when these new services start growing, there is a share of the pie that goes somewhere else. But the pie is growing at roughly similar rates. So you don't see Google's total Google searches going down. It's that there's this shift slow and then the pie is just getting bigger and bigger because more and more people are using search and you know, the Internet and things like that.

Before we get back to today's show.

Here'S a quick word from HubSpot. If you're a marketer, one thing I know for sure is you love data and boy do we have data for you. The 2024 State of Marketing port is chock full of data and insights around the current trends that are shaping the marketing industry today. Things like artificial intelligence, you know, I love AI tools, personalization, influencer marketing, all of the topics that are key to getting a competitive advantage year. It's going to make sure you're not stuck in old strategies and old tactics. So click that link in the description and go get your free copy of the 2024 State of Marketing report today. Now let's get back to today's show.

It's interesting and maybe I'm just trying to push on this to try to win because I've done so many LinkedIn posts about the death of Google search but and maybe it's because I'm in a tech bubble. I do like, I do always try to, I always, I'm nervous about that where like just ad hoc so many people I know are using perplexity versus like the Google and even my own Google usage has gone way down in favor of AI chat assistance. And the thing that I mostly find it valuable for is that it it's basically an assistant to like I think of it as an assistant to do the search, go through all the blue links and then formulate the answer. The answer after it has done the research for the blue link. So like a layer on top of the search engine and that's basically what perplexity is, right? It's like AI, AI agent that sits on top of the search engine, goes through the blue links for you and then synthesizes the answer and formula and formats it and so you don't actually have to do the research yourself. And that's the part that I still am stuck on that I feel like that's a better experience for the user. And anytime software makes something easier, allows the user to do it in a much more lazier way and is faster, I feel like that that experience always wins. And so do you buy any of that? That like there is something to the fact that it kind of trawls through the blue links and like formulates the answer for you so you don't have to do that or you think that you mentioned like Google search is just the best experience for many searches, that there's just like certain, certain category of searches that you have to do that research yourself.

I do think that some when the case is something where you want to summarize large data sets, like an example would be best digital cameras, right. What would be great is if you could a bunch of listicles and rank lists and then summarize it into a single one. And I do think that that's a better experience and if it is, then Google will just build that themselves. And actually the first company that I worked at when I was doing SEO spam literally did that where we were summarizing other lists and leaderboards and things like that. And then we served it back and then Google released Google Panda where they said we don't want any scraped content, we don't want any of this. If someone's going to summarize it, it's going to be us. I don't have Google shopping, we don't need any of these other summarizers. So I do think that for many cases summarization of large data sets is useful and when it is, then Google will just do it themselves.

Like the thing I've been paying attention to is like number of search queries over time because you can't actually get. Google doesn't release any real data around daily active users and monthly active users. And basically you expect a number of search queries that are done in search engines to pretty stay stagnant if not grow over time. And then there would be another category of searches that is like AI chat searches and you still expect to see many of those done by Google like Google Gemini and then maybe some of these other players like Perplexity and these other kind of players get some sort of subset of those. But it's not going to be any real disruption to like Google's core search engine.

I hope that there's disruption and I am rooting for Perplexity and all these other search engines. However, if you think about how hard is it for Google to build a summarization, how is how Hard is it for Google to be or to build any of these things that are competing with it? It is hard to build TikTok. It's hard to build Instagram and YouTube where you have a content moat. But if it's something, if it's an algorithm like summarization of a data set, if people are using that, then Google will just do it themselves.

Right. What are your thoughts then on AdWords as an AD model, then I suspect that you probably think the same, that there's no real disruption in Google's Future. And the AdWords model, if their search engine stays pretty relevant and stays the same amount of queries, because that was. The other one is like, I don't think these. My kind of thinking was that AI chat would, would consume a subset of searches and even a subset of searches would be. A lot would be real problematic for Google's AdWords model. And when you go through these AI chat experiences, like Perplexity is a good example, actually, I think, I think that's a good example because they started as a subscription model to create revenue and now they're like, oh, we're going to add ad revenue. Because ad revenue is the only way you can really become a big business in that kind of search model. And in that ad revenue model, I just don't think anyone ever clicks on any of the sources cited. Right. So I don't think the adword, I don't think the ads model is going to work anywhere near as well as like it does for, for Google search. So do you think that the AdWords model doesn't get impacted at all either by AI chat or AI search?

I think the AdWords might get impacted, but ads generally I don't think will be reduced. I think that you'll just have different surfaces like product listing ads and map ads and things like that. You can have chat ads. And I agree that I don't think anyone's clicking on the publisher in a chat response. And if that's the case and you're taking ad revenue away, then they'll find a way to have clicks go to a publisher so that they continue to make money there. And so it'll just be a different type of an ad.

All right, cool. Well, I'm going to go back and edit all my LinkedIn posts. But, but before then, one thing I would love to quickly ask is, okay, so you have a bunch of clients, like some of the best clients who have incredible search strategies because you do them and they're coming to Ethan and they're Saying even like, I'm hearing a lot of chatter on LinkedIn from these people talking about, you know, AI and how AI is impacting search. Is this one person, Kieran, he keeps telling me there's a demise of Google coming. He's probably wrong. But we want to figure out like, how do we, how should we integrate AI into our SEO strategy? What are you doing for your, your best clients right now? Like, how have you, have you used AI? Are you starting to integrate AI in some ways? And we would love to get like, what are the interesting ways that you are integrate an AI into your, your client store strategies?

Yes. So there's many forms of AI. The most talked about currently are large language models and text generation. But there's also AI that helps understand things and synthesizes large data sets and things like that. So most of the questions that I receive are specifically about text and content generation. So how I respond to that is first to say what, what works? And is this working? So we did a study where we looked at a large representative set of keywords and URLs and found that most things that are ranking are not AI generated text only.

Do you want to show that data, Ethan? You can, you can share that data if you want.

Little, little preview.

This is, we took about 10,000 URLs and keywords across a bunch of different categories. And this chart here, can you slide.

Show it so it's bigger for everybody.

So what you'll see in this chart in the middle is the X axis is how much AI content. So bucket one is zero to 10%, 10 to 20, 20 to 30. And what you'll. And this is using originality AI, which is an AI detector. It usually has slightly, it's usually overcounted slightly, so maybe discounted by about 10%. Like NAV and footer will go in there. Basically what you'll see is that of Pages with between 0 and 20% AI content, it's 90% have almost none, 90% have less than 20 detected and 10% have more than 20%. So there's very little, very few URLs that are mostly AI content. The second thing I can show you is that you'll see the AI generated content also has a much higher SERP ranks, meaning it's ranking on page two instead of page one. And you'll see a pretty strong correlation here. And the third thing is looking at categories. So some categories have more AI content than others. And you'll see that food. And we want to look at the blue line Here. And you'll see that food has no AI content, whereas something like commerce and local, which is probably going to be, you know, product descriptions and things like that are specs and local is probably also going to be specs, has more AI content. So depending on the category, it makes more and less sense to have AI content. But in general you'll see that AI content is just not prevalent yet. And so the takeaway here is that it probably doesn't make sense to generate a million AI generated pages. And you also saw from the update that Google made a couple weeks ago or three weeks ago, they basically said we don't like bad unuseful AI content. They did not say that we don't like useful AI content. My hypothesis is that they will not like useful AI content. That's really the question. Will useful AI content perform? My guess is no. It might. But you'll remember from Google Panda and the scrape content days you mentioned synthesizing other people's content, that is a useful scraped content strategy and it does not work. If I scrape a bunch of people's reviews and synthesize them, I will get blacklisted. So Google doesn't care about scrape content, useful or not useful. And my guess is that Google will take the same position with AI content. I don't care if it's useful or not. If it's useful, I'm going to make it. I'm going to not let you make it.

And it seems like in the data you have, Ethan, you're like, you're basically already seeing that, right? That like Google even in the early days is getting pretty good at detecting AI content and then not blacklisting it, but definitely ranking it lower compared to non AI content, which is probably a good leading indicator that your hypothesis is right.

Yes. However, what really I would want to do would be like a controlled randomized study where I take a thousand pages and I have a really high quality useful AI version and a high quality human written version and maybe a mixed version where it's an edited version. And I test that on and test that across categories. And we did not do that. And it's possible that, you know, if you did that, then that would show that there is a path to some category and you know, specifically something like finance where summarize a stock price or a sporting event, describe what happened in a game. Like I could, I could see that performing reasonably well. So we have some indications that generally AI content is, is dubious. But I wouldn't say that we've proven that there's no scenario where it could possibly work. And there are people doing interesting things around useful AI generated content. And you know, I do think that there's a place for companies that help, help people make really high quality useful AI content.

The problem is that most people start with the lazy use case, right? They, they start with the just regurgitate something that is easy, easily detectable. I do wonder but like maybe not models now, but let's say GPT5, which apparently is going to be a real leap up and you, and you give the agent some pretty like a really good prompt to create a piece of content about how to do product marketing. And then you had a writer create how to do product marketing. And the LLM isn't regurgitating and maybe this is like something that I'm just wrong on, but the LLM isn't regurgitating content. It's like trained on content and then we'll guess the next word. So it should be original as in like it's just not copy and pasted from somewhere. So how does, how does Google detect if the model is really good? Who wrote? Because it's not a hard, that topic is just an educational topic, right? It's not point of view, thought leadership. It's hard to have like any kind of differentiation and how to do product marketing. It's more like an educational piece of content. So like in the future, how does Google detect the difference between those two things?

I don't know how they're detecting that. I can tell you that there are multiple companies that are less of a scare than Google that already have AI detection and the accuracy is pretty high and they're just built. You know, you build a model that's trained to specific chat algorithms, anthropic and OpenAI and there are already detectors that do that really well and I expect that Google will be able to do that.

Yeah. Okay.

In the case of a human versus an AI writing we see the results be similar and chatbots actually can come up with structure and outline, which I think is what you're asking about. It's not just what the next word is, but it's what's the general layout and structure of the content. And chatbots are pretty there. I would say they're decent at that. However, actually humans and chatbots are not as good as what we do. And what we do is the outline of the subtopics are based on looking at the different keywords that users are searching for. So we can see that there's different sub themes and follow up questions in the search queries, if you look at all the queries that you want to rank for that neither the human on their own or the chatbot actually knows. And so if you, if you have special data about what people are searching for, you can actually have an even better outline than a human alone or than a chatbot alone. So I do think that there's ways to have better outlines. The other thing that I would say is that chatbots don't add personal experience or unique, unique perspectives, whereas humans can add unique perspectives and personal experience and expertise.

That's where I think you can differentiate content. When you have personal stories like uniqueness to your, the way, the way that you would describe something. But there's a lot of content which is just like this is an educational piece of content, you know, it is what it is kind of thing. So it's hard to be much more creative with it. But I just want to come back to. We kind of, we were kind of really talking about how you're integrating AI into your current strategy. So we talked a little bit around the content and that was a really great study. It kind of shows maybe a little bit away from all the hype that actually AI content is not as prevalent as maybe we thought. It's very, it's in some very specific categories and Google is already pretty great at detecting that AI content and it ranks lower and they had an update very recently to actually get much more of that content ranking lower than like original content. But what are some, maybe other things that you are integrated into how search is done today for your clients?

So, so there's two things. There's first thing is how do we build strategies for Google search today and then how do we build strategies for chat? So in terms of how to build strategies for Google search today, the job of an SEO person is generally data curation, data entry, like a glorified data entry. And, and the reason why is because tools are not able to make sense out of the data. So you have to cluster your own keywords, you need to figure out what they mean. You need to figure out what the Persona is for that. You need to come up with the outline yourself, the title tag, all this stuff you have to personally do yourself. Internal linking, where do I put the link in the body? Like all this stuff is fairly low level work done by someone who's very high skilled. And the reason why is because there's just not tools and algorithms to do this. Well, they're very noisy. But AI can do data synthesis and data analysis much Better. It's just very hard to build these algorithms. So we've built our own algorithms to do things like keyword clustering and prioritization of the keywords, which internal links are optimized. These are hard to build. We basically built our own set of these algorithms so that. And similar to HubSpot. So 30 or 20 years ago, someone working in email marketing had to build all the infrastructure and plumbing to make emails work. Forgot password, IP warm ups, all this stuff. People don't need to do that anymore. You have a strategist who comes in, turns some knobs. But all this plumbing and infrastructure happens magically in the background with a platform and then they can focus on strategy. So somewhere with SEO, I think that that's where it's going. Where a lot of this data curation infrastructure stuff is, you don't need to worry about it. You can focus on strategy. So that's, that's the other thing that we've used AI to do. And I think that AI will get better and better at that.

Could you maybe give our listeners a quick lesson on what you mean by clustering? Because I think it's a really important thing for search strategies and you've done it from, you've done it from some, some of the largest websites there are on the Internet. So maybe a quick lesson on why that's an important thing and why you decided to build, you know, an AI algorithm for it.

Yes. So if you take something like five year plan and actually five year plan is one of my favorite examples because HubSpot also ranks for five year plan. But we worked with better up on five year plan. And when we worked on five year plan, there were, there were no pieces of content around five Year Plan. And five Year Plan is in my career. I want to plan the next five years and be purposeful about how I design my career. No one had an article about this and all the results were about Stalin's five Year Plan for the Soviet Union, which did not fulfill the intent of the user, hopefully. So when we create that piece of content, the reason why people were not competing on that is because the search volume for an exact match of five Year Plan is not that high. But there's a thousand different ways to search for five Year Plan. Five Year Plan examples, Personal five Year Plan, Professional five Year Plan, ten year versus five Year. Like there's all these subtopics and we need to know, do we need a thousand pages or do we need one page? So like personal versus Business five Year Plan, do we need One page for that or two different pages for that. And if you use standard NLP algorithms, it does not know the answer to that, whether or not there should be one or several. And what Google's basically doing is they're using NLP to figure out whether or not a piece of content is relevant for a given search query. But then they'll look at the click through rate. And so if a user is having a high click through rate for both personal and business, Google will say, I'm going to show the same result for these two queries. And if the click through rate's low, then they won't. And so we basically built an algorithm to look to see which keywords have the same set of search results. And if it has the same set of search results, they should be grouped. And again, this is essentially Google crowdsourcing the entire world telling me, you know, based on click through rate, which keywords and URLs should show up at the same time. So that's what our algorithm is based on. Other algorithms are not based on that. Which is why you need to do all of this cleaning work like it's a recommendation.

Then you create content around those keywords, like one piece of content for one or two keywords and then you interlink them like is that the, is that the output for the customer, how they use that data?

No. So it's the opposite of that. It's you make one piece of content for all thousand of those. Oh, and there's all these subtopics and sub themes. So for five year plan, there's examples, templates, personal versus professionals, 10 versus 5. And you did cover all these. And you should not make articles for all these. If you do make articles for all these, you're wasting money, you only need to make one. And if you make an article that's just for five year plan and you don't have examples, samples, personal versus professional, you didn't fulfill the intent and you have these content gaps. So this keyword. And also it's hard to go through a thousand different keywords manually. And you would be wrong, you would be going through and you would be making assumptions about that. So there's this mass amount of inefficiency for bad keyword clustering and it takes a ton of time. And so that's why you would want an algorithm to do that, because it's not only faster and cheaper, it's significantly better than a human can do it on their own.

Okay, and what's it, could you maybe just like give the summary of what the. I know it probably sounds like overly simplistic of a question, but just what the AI is doing within that process.

The AI is going through billions of search keywords and URLs and finding which ones co occur the most frequently. So which keyword URL pairs appear the most. And what Google's doing is they're looking to see for all the URLs that I'm showing for this keyword which ones get the highest click through rate. And I'm going to show the URLs that get the highest click through rate. And these thousand are all getting a really high click through rate on the same set of 10 search results. So I'm going to show all these 10 search results for these thousand different keywords because the click through rate so high.

And you mentioned there is three. So that was one of the algorithms, one was internal links. What was the middle one?

Content scoring. So we've taken five year plan, we've gone through all these different subtopics we want to look and Google's looking for those. So they're saying are you comprehensive? Did you cover everything or did, did you have a gap? You didn't talk about, you didn't give an example of a five year plan. So that's a gap. So we take all this information about subtopics and then we can score the content and we do that based on which of these subtopics have appeared. And then we also built a machine learning model to basically look to see the patterns that correlate with rank. And that's sort of an abstract thing, but basically it's looking for strings that co occur with or that correlate with rank. And there's a feedback loop and then you can predict the probability of this piece of content ranking for this thing and what are the gaps. And if there's a gap, like you didn't use personal experience, okay, add that and then you're, you'll be more likely to rank for that.

So Ethan, it sounds like one of the really cool things is, and this is related to a conversation Kip and I recently had with Jason Lemkin about strategists and ICs. But it sounds like everyone like what, what you're doing is given Armin, a search person with like a pretty high caliber, like data analyst person who can like really pattern match data, provide a bunch of recommendations. And you mentioned something really interesting, which is, you know, the AI can take a bunch of that work and move the search person to a strategist and One of the things that we talked about recently, Kip, is like, we actually fear that a lot of great individual contributors want to move to strategy. But we kind of talked about it as a manager.

Right.

They are like, I've done my job for eight months, and now I want to go be a manager and work only on strategy. But in this case, it sounds like the strategic part will be managing the AI software and AI agents. Can you maybe talk to how you see the elevation of the search role or how you see the search role changing, like, to be more strategic because you have these AI tools?

Yeah. What I'm seeing. The best search strategists that I'm seeing now coming up, are really good at content and marketing generally. Like the people that I work with at Masterclass. Many of them are great SEO people. And because they're so good at content strategy and editorial strategy and marketing generally. So what I think will be the future of SEO people are people who are good at identifying ideal customer profile. What do they care about? What's content that will resonate with them? How do I make it engaging and sell the product that I'm selling? I think that that's the future of SEO people. Kind of like if you think about advertising and Facebook, you don't need to think about how the Facebook algorithm is working. You need to think about what's compelling content, what'll get people to buy something more so than the mechanics of what keywords and what the bid rate is and stuff like that.

That's pretty interesting, right? It basically is taking the kind of real, deep technical expertise and doing that. And so that person can. The nature of the skill set that person needs to have actually changes pretty significantly towards needing to understand marketing a little more, needing to understand customer Personas a little more, need to understand content a little more. And you know that deep technical expertise can be managed by the. The AI software itself. So, like, it feels like if you were in search and you were like real technical, deep in the weeds on that side of things, that was your career. Like, you might want to start to figure out how to learn some new skills.

Yes. And I don't think we're there yet, but we're almost there, and I think we'll be there over the next one or two years. The other thing that I would add is that getting people to. To convert, I think is very important. So you buy an ad, it's optimized for your audience. The content needs to be optimized. But then you need to get somebody to actually convert and so one thing that I think that is missing a lot with marketing and content is the conversion optimization piece. That would be one other skill that I would add.

We've talked about like the continued fragmentation of search and people searching in lots of different places for different types of formats. If you were a business trying to build a search strategy is today is it from and you're just a startup, for example, are you starting on Google or are you actually taking a beat and saying oh you know, maybe TikTok, maybe YouTube, maybe some other more specific form of search is a better place to start? Like is Google still the default place to start if you're a startup out there?

I think it depends on the channel. So if I'm doing Beauty, for example, I might start on TikTok and Instagram I would probably and if I'm doing finance, I'm probably not going to do TikTok, I'm probably going to do Google. So I think it depends on the category. I would also say that I typically if I'm starting I want to try a few channels lightweight, dip my toe in the water and then figure out what I think might work and then double down on that. So I wouldn't have five different focuses and focus on all five. I would try a few, find one that seems to be working and then double down on that one to two channels.

What did you see from the recent Google updates? They seem to be like kind of Panda Penguin esque blowback. Maybe just give us your $0.02 on what you saw happen and what you've seen happen among like all of the kind of portfolio of companies that you work with.

Yeah, we saw no client go down in traffic and I think that the sites that went down in traffic are the, were the gray hat AI con, you know, pure AI not useful type type companies. I think that this is basically like a, a pre Panda notice. Like we're about to launch Panda and we're letting you know ahead of time and we're gonna shut down low quality spam AI content. Kind of like with Panda it was scrape content is we're gonna get rid of that. They're basically saying clean up your AI content. We're not gonna, we're gonna, we're gonna penalize you.

I would say what we've seen Ethan, both at HubSpot and I've talked, I've talked to a lot of companies out there, big some of household brands that people would know and some early stage startups that they're definitely seeing erosion on the informational keyword side of things. Not transactional but informational text based blog articles, how to those types of things and none of them are doing, you know, deep AI driven content. We're not doing any crazy AI driven content. Like what do you subscribe to that change? Because it does. If you read, you know, the SEO subreddit, you talk to a bunch of folks. There is, there are some subset of sites out there that are experiencing that.

I haven't seen a discrete jump in traffic. You're describing low intent, informational, simple to answer queries like yeah, like what is two plus five or like how many teaspoons in a tablespoon?

No, more, more like how do you, how do you write a great blog article? Like how to articles information articles that have real depth but maybe are getting deprioritized for community rankings and you know, Reddit threads, other other types of content.

Quora was a big winner.

Quora was a big one.

For example, stuff that I've seen.

Yeah, Quora and Reddit went way up in traffic. I think Quora tripled in traffic or something huge. But that's not, that's Google prioritizing ugc but it's not Google prioritizing AI responses in place of in depth articles. I haven't seen a huge change across the board for what how to content, but I have seen both Reddit and Quora dramatically increase. I think that that makes sense and you know it's up to Google whether or not they want to prioritize UGC content but that's what I've seen.

Anything else like if you're a business listening to this, thinking through how you would use AI within your search strategy and you mentioned a bunch of ways that you're using it for clients. Anything else we've missed that you think is a no brainer for someone to implement through AI or even look at.

Start to look at through AI AI assisted content. So there's 100% AI written content which I think could be useful but I'm skeptical about that being a search Strategy. And there's 100% human written content and then there's AI assisted. I think the future of content is AI assisted where the AI is sort of your co pilot, it makes suggestions, it could even write initial versions of some text but then the human is editing it, adding their own personal experience and their own wisdom. That I think is the future of content. So I think embracing that workflow and building that out where AI is helping you but there's still a human in the loop that's what I would focus on.

Totally agree with that. I think having AI write first drafts, have an AI give you very. Like the thing that people, I just think underestimate is the AI doesn't get bored. Like, if I ask the AI to write a hundred variations of a headline, it will do it right. If you ask a human to do that, like on board at like number 10, like, it's just much better at these kind of ways to spark ideas. But to your point, I think your tone and voice has never been more important or never will be more important than now because it's going to be what makes you different online. Right. That you can get these first drafts and then add your tonality and your voice and your way of looking at the world. But it's really good for first drafts. So we would be remiss not to ask you, Ethan, about future Playbooks, right? Like not the here and now. I know you're even launching a platform to help companies do search and maybe you can talk through some of the future ways that you've built in future Playbooks within that platform. But like if you're, you're a business and you're listening to this and maybe actually you feel a little bit better because Ethan's told you, like, hey, Google search is not going away. It's as important as it ever has been. Don't see it going away anytime soon. But when you kind of pull ahead maybe two years, three years, anything that you are thinking through in terms of how search will change or how your Playbooks might change to take advantage of what's coming.

Yeah, I think the main thing that I'm thinking of future facing about is how to optimize for chat answers or for AI answers. And so we talked about the summarization of listicles and I think that if you think about either Google's chat for what's the best camera or some other surface like ChatGPT, you ask what's the best camera you can actually optimize for that? And so building strategies to optimize for chat answers. And there's, there's different cases. So case one would be what's the best X what's the best camera? And so for that I think listicle optimization. So I want to show up on 50 different listicles and have my camera show up over and over again so that the language model says best camera co occurs. You know, next word is Sony Alpha A3 or whatever. So optimization for that. I think that there's probably also other things where you just generally Co occur with search term and product. Like what is a good project management software? And maybe separate from listicles it's just many articles with project management software and notion appearing and then next word prediction predicts notion as a result of that. So I think that there's a whole field of chat optim or chat answer optimization and then there are spam versions of that and there are non spam versions of that. So spam versions of that is where we. Let's say that we could create a whole affiliate marketplace where we allow publishers to sell having a co occurrence of a keyword and a product so that they're that the next predicted word. It's kind of like link building. So you know, co occurrence of or next word optimization is the new buying buying links which could be spam or which cannot be spam. But I think that there'll be a whole field of how do you optimize for. For your company to show up in a chat response?

I will tell you that everyone listening to this has just like glued their ears to what we're talking about which is how do I optimize for AI assistance? How do I optimize for ChatGPT? So you mentioned one of them which I think is really important for these large language models. If you are trying to appear in like best digital camera the more times you're listed in the top 10 all across the web the language model learns that you are in the top 10 repeatedly and it's more likely to give that response when you ask it for a top 10 digital camera. You mentioned a couple of others. Can you make. Can you just make sure you spell them out for the listeners how you're thinking about optimizing for these kind of chatgpt experiences? Because everyone is taking notes right now so I want to make sure they get them.

There's listicle optimization which is best digital camera. Show up a bunch of listicles and actually you'll. You'll note on Bing which I think and for some of these chat responses I think chat GPT either is today or will use search engines in the background. So ChatGPT obviously is related to Bing and if you search for stuff on Bing they're actually explicitly summarizing listicles and it'll even mention the listicles that they're showing for many of these things. So listicle optimization is one. The other is mentions or like co occurrence. So keyword plus product name and actually Google will also has mentions where it's the number of times I see your brand mentioned even without A link then that'll give you an authority score. The other I think will be quantity of mentions and quantity of listicles. And this is actually an interesting thing where I think Google might be able to massively outperform ChatGPT. So you'll recall link farms where I'm gonna make a million websites and link to myself a million times. I had one and that used to work. Very effective. Fun, fun days. That used, that used to work and it doesn't work anymore because those million websites have a low trust score. Right. And I can see the patterns of these million links. So clearly these are fake but. And you could do the exact same thing with, with listicles and articles. So I can just make a million websites with a million listicles that all mention my brand and all have mentions with my keyword and my product. And Google will be able to detect that because they already have algorithms to say whether or not I trust these domains or not. I don't know the ChatGPT necessarily has that. Or let's make maybe they do with Bing, let's say anthropic, which does not have their own search engine. Will they know that those million listicles and million articles should not be trust or not? I don't know. So I think that there'll be a whole link, you know, AI chat link farm version. And then I also think that there's discussion about whether or not the large language models know what large language models you would want to train on non chat responses. So if the large language model is trained on its own answers, then you have this infinite loop of derivatives of derivatives converging on a single perspective, which would be bad. You don't want training data of your own responses. And we talked before about whether or not you can detect AI content. I think that you can, but we don't know. Let's say that you don't know. Then again you can just generate infinite amounts of content that then gets fed back as training data into the language model for zero cost. So that would be bad. But anyway, those are a few initial ideas.

Yeah, the second one is really interesting to think through which is the future of link building is like what you call, you know, cohort citations or mentions, which is, you know, how do I get my word, how do I get my brand mentioned beside this word? So the large language model figures out how to associate those things together. That's actually really smart. Like the. I hadn't thought that, you know, you mentioned an affiliate marketplace where you can actually have affiliates go create content and co site those things together. I hadn't thought about that. I had read a thread on Twitter, obviously. I also was really curious about how you hijack the results and like there is this kind of example of how you could poison the LLM, which is like very similar to what you mentioned, which is like create millions and millions and millions and millions of pages around this, you know, whatever content will help train it on the things that you want to help it train to be trained on. So, you know, it's a, it's a kind of wild new world, like a little bit similar when search was coming up. There's going to be lots of ways to like figure out how to do it the right way, the wrong way. But you give two incredible lessons for listeners there, which is the list. List goes. One is something that people can work on right there and then actually HubSpot are a little bit ahead because that was something that we worked on many years ago. And then the second one is like, how do I start to think about my link building as co word citations? Like get my, my brand mentioned alongside the words that I want the LLM to associate me with. Like they're very actionable and I haven't heard them.