Will Google's AI Experiment Replace Search Engines?

Summary

Podcast Summary: "Will Google's AI Experiment Replace Search Engines?"

Podcast Information:

• Title: Marketing Against The Grain
• Host/Author: HubSpot Podcast Network
• Episode: Will Google's AI Experiment Replace Search Engines?
• Release Date: November 12, 2024

Hosts:

• Kipp Bodnar: HubSpot’s Chief Marketing Officer (CMO)
• Kieran Flanagan: Zapier’s Chief Marketing Officer (CMO)

---

Introduction: Exploring the Future of Search Engines

In this episode, hosts Kipp Bodnar and Kieran Flanagan delve into the evolving landscape of search engines, focusing on Google's latest AI experiment, "Learn About." They discuss the potential impact of large language models (LLMs) on traditional search functionalities and what this means for marketers and business owners navigating this shift.

Notable Quote:

• Kip (00:00): "Thanks to AI, it is an action-packed show. Let's get into today's episode."

---

Google's "Learn About": A New AI-Powered Learning Tool

Kipp introduces Google's new tool, "Learn About," highlighting its unique approach to delivering focused AI experiences. Unlike broader AI tools, "Learn About" is designed for specific learning purposes, akin to Google's previous experiments like Notebook LM.

Functionality Overview:

• Interactive Learning: Users can type in topics to receive definitions, interactive diagrams, and detailed explanations.
• Dynamic Navigation: The tool allows users to drill down into subtopics, creating a composable web experience tailored to individual queries.

Notable Quotes:

• Kip (01:16): "...Google is creating these very, very focus-specific AI experiences where you're going to use it for very specific things."
• Kieran (05:55): "It knows what it's talking about. Clearly."

---

Comparing "Learn About" to Traditional Search Engines

The hosts compare "Learn About" to conventional search engines, noting its enhanced interactivity and personalized user experience. Kieran expresses curiosity about whether Google will integrate this tool into their main search engine or keep it as a separate application.

Key Differences Highlighted:

• Personalization: "Learn About" tailors responses based on user behavior and preferences.
• Interactive Elements: Incorporates multimedia and dynamic lists, offering a more engaging experience than traditional text-based search results.

Notable Quotes:

• Kip (07:28): "I find it way, way, way better than search. I find it much more interactive."
• Kieran (14:09): "I think this is an experiment to say, hey, can AI search be a better curator of information..."

---

Impact on Marketing Strategies

Kieran and Kip discuss how AI-driven search tools like "Learn About" necessitate a shift in content strategy. They emphasize the importance of creating unique, differentiated content that AI cannot easily replicate. This involves integrating customer insights, expert quotes, and original research to stand out.

Strategic Recommendations:

• Content Categorization: Assessing content based on its susceptibility to AI cannibalization.
• Differentiation: Investing in unique content that offers value beyond what AI can generate.

Notable Quotes:

• Kieran (21:46): "The opportunity right now is... you have to stay one step ahead."
• Kip (30:56): "You always want to be trying to figure out what your leverage is."

---

Predictions for the Future of Search Engines

The hosts project significant innovation and disruption in the search engine market over the next year. They anticipate improvements in AI search capabilities from platforms like Perplexity and ChatGPT, alongside Google's aggressive rollout of AI features globally.

Future Outlook:

• Continual Disruption: Ongoing enhancements in AI search tools will challenge traditional search dominance.
• User Experience Evolution: Emerging UX patterns will redefine how users interact with search engines, integrating multimedia and interactive navigation.

Notable Quotes:

• Kieran (19:15): "Next year is going to be a real year of continual disruption."
• Kip (16:10): "It's way better. Like I do like it much more."

---

Advice for Marketers: Preparing for AI-Driven Search

Kieran provides actionable insights for marketers to adapt to the changing search landscape. He categorizes content into three buckets—most at risk, maybe cannibalized, and still fine—for strategic prioritization. The focus is on enhancing content quality and uniqueness to maintain visibility in AI-driven search results.

Actionable Steps:

1. Categorize Content: Identify which content is most vulnerable to AI replacement.
2. Enhance Differentiation: Invest in creating content that offers unique perspectives and insights.
3. Leverage Community Presence: Build a strong online presence through community engagement on platforms like Reddit to enhance brand visibility.

Notable Quotes:

• Kieran (21:46): "You can't do anything about the most at risk bucket."
• Kip (28:24): "For my brand to appear larger online, both from a product and information perspective, I need to mobilize the people who do know about my brand from a community perspective."

---

Understanding LLM Prioritization: A Key Marketing Challenge

The conversation shifts to the mechanics of how large language models prioritize content. Kieran highlights the reliance on co-citation and word association in determining content relevance. He expresses concern over potential manipulation through mass content injection and emphasizes the need for quality over quantity in content creation.

Key Considerations:

• Word Co-Citation: Frequency and context in which brand-related terms appear together.
• Quality Signals: Analogous to Google's Page Authority, ensuring content credibility and authority to prevent manipulation.

Notable Quotes:

• Kieran (26:18): "LLMs are trained on the Internet and look at what are the most common words cited with a thing."
• Kip (28:39): "There will be a period of time where companies that might be smaller can have an outsized footprint on the Internet."

---

Conclusion: Navigating the AI Search Revolution

Kip wraps up the episode by reiterating the importance of maintaining robust content strategies while adapting to new AI search paradigms. He hints at future episodes that will explore how LLMs prioritize content and offer deeper insights into optimizing for AI-driven search engines.

Looking Forward:

• Series on AI Search: Upcoming episodes will track innovations in AI search and provide data-driven insights for marketers.
• Expert Interviews: Plans to bring in LLM developers to demystify content prioritization mechanisms.

Final Notable Quote:

• Kieran (31:55): "That's the thing we can do for our audience here."

---

Overall Insights: This episode of "Marketing Against The Grain" provides a comprehensive analysis of Google's "Learn About" AI experiment and its implications for the future of search engines. Kipp and Kieran offer valuable perspectives on how marketers can navigate this evolving landscape by focusing on content differentiation, understanding LLM prioritization, and leveraging community engagement. As AI continues to disrupt traditional search paradigms, staying ahead with strategic content and adaptive marketing practices becomes increasingly crucial.

---

Relevant Resources Mentioned:

• Created Podcast: Hosted by John Ushayak, featuring insights from creators like Logan Paul and Paris Hilton.
• a16z Podcast: Hosted by Steph Smith, covering a wide range of forward-looking topics with industry leaders.

---

Subscribe for More Insights: Kipp and Kieran encourage listeners to subscribe to "Marketing Against The Grain" to stay updated on the latest trends in AI-driven marketing and future episodes exploring the intricacies of AI search technologies.

---

Transcript

[Kip]

Hey everyone. I just came across a new AI tool that I don't think you're using and you're going to love it. I'm going to show you how to use it. I'm going to show you some of the really cool research use cases and learning use cases that come from it. And that leads right into an awesome discussion Kiran and I have about the future of search engines. How LLMs actually like prioritize information, what marketers or business owners can do to prepare for this search disruption that's happening right now. Thanks to AI it is an action packed show. Let's get into today's episode.

[Kieran]

Hey guys, real quick. You know we love building custom GPTs on the show and we love sharing it with all of you. Well, we wanted to kick that up a notch. We just developed this free guide that teaches you how to build your own.

[Kip]

Custom GPT on ChatGPT.

[Kieran]

We've taken the guesswork out of it. We've got templates, we've got a step by step guide to design and implement custom models. So you can focus on the part that's actually fun, the part we love actually building it. And if you want it, you can grab a link in the description below and go check it out now. Now back to today's show.

[Kip]

Okay, Kieran, I want to tell you about one of my favorite AI tools and it's one that nobody is really talking about. I would say of all the tools, it's kind of flying under the radar and it's from potentially an unlikely source. It's from our friends at Google. Let me show you what I mean. Google just released a new learning experiment and it is called Learn about and you can sign in to your Google account and you can just type in whatever you want to learn about. Sounds weird, right? What I think Google is doing really well is with Learn about and Google's Notebook lm, they're creating these very, very focus specific AI experiences where like you're going to use it for very specific things like Notebook lm, it's like, oh, you can upload those documents, ask questions about them and generate a podcast. Like that's very simple and straightforward. This is even more simple and straightforward, which is like, hey, just type in something you want to learn about. Which reminds me a lot of Search, right?

[Kieran]

Yeah, sounds like search.

[Kip]

Okay, so what I will say right now, Google Learn about is only in the United States and it has not been released outside the United States. So you have not gotten to use Learn about yet and you have not seen the experience yet. Is that right?

[Kieran]

Yeah. Yeah.

[Kip]

Okay.

[Kieran]

I have no access.

[Kip]

That's why I wanted to do today's show so that we could get Kieran's live reaction and get his feedback. So I want to start with showing you the learn about page here. What I'm showing you now is I typed in CRM to learn about just that, those three letters, CRM. And I was like, you know, HubSpot's obviously in the CRM business. I wonder if we just like give it the most basic query in the world, what would it tell us about CRM? So all I did was type in CRM and right away it gives me the definition of what CRM is and the goal. Like that's pretty good. Like that's very helpful. It gives me an interactive diagram. I can basically like. It highlighted sales for me. And when I click sales, it brought me down to this whole interactive list of why CRM helps sales. And what is fascinating about this, that I was kind of blown away with is that it's creating these kind of real time navs and kind of a composable web experience just specifically related to this. And so let's say I want to drill down and I want to understand how it helps with sales automation. It then moves this nav to the left. So it gives me like kind of the different artifacts. Right. I can go back to the diagram right here. It has this static list of why CRMs help sales. And then it's going to answer my question and it's going to prompt me with another interactive list. And so let's say I want to learn about task scheduling. Oh, what happens? That interactive list moves over here to the left hand side and it's going to dynamically generate a table all about why task scheduling in CRM is really important. And so it's no more missed follow ups, timesave, better organization and prioritization, improve consistency, enhanced accountability. And what I love is it's got these Chiclets here where it's like, oh, I need a simpler version of this, a deeper version, or I want images. Which one do you think we should do?

[Kieran]

Let's do images.

[Kip]

Images. All right, so we go get images. Here are some images showcasing CRM task scheduling in action. So it's actually finding screenshots on the web and showing you task scheduling. And if I open one of them and what's interesting, Karen, I open one of them, what's it take me, it takes me to an actual CRM website, HubSpot, unfortunately. But this is actually a referral traffic from this tool.

[Kieran]

So it's kind of just like a different search experience. Like they're Korean in a different user UX to navigate you around a topic.

[Kip]

Correct. But I find it way, way, way better than search. I find it much more interactive. And then what's interesting too, Kieran, is that it has a lot of context. So it's knowing that I'm looking at like features of a CRM product. So now it's like, how can I choose the right CRM system for my business? And it's actually going me from like I don't even know what CRM is to explain me what CRM is to now like, oh, if you wanted to buy a CRM, this is how you would think about choosing it. And it's given me this interactive list and it's got features, ease of use, integration, all of these things. And then can you recommend some popular CRM systems? Let's see if HubSpot comes up. Karen, you ready for this? Popular CRM systems for small business. HubSpot. We're number one, baby.

[Kieran]

There you go. It knows what it's talking about.

[Kip]

Clearly.

[Kieran]

I do wonder what they're doing with this experiment because it is kind of a just more curated search experience.

[Kip]

It's kind of wild, right?

[Kieran]

They talked about this when they were originally talking about how they would integrate AI across their search engine and that they would have these dynamically built interfaces, which is what this really is, is it's dynamically building an interface to navigate you around a specific topic. So I do wonder if what they're doing is using this as a separate app or they're just trying to learn some things and integrate it into their search experience. I suspect this will be integrated into search because really what it's doing is just like collating all the information and then present it to you and more like a course structure.

[Kip]

Look, I love this. Better than search GPT, better than perplexity. For me, this is like my most used form of AI search right now.

[Kieran]

Why?

[Kip]

Because I love the composability of it. And by the way, Kieran, it's super, super personalized. Remember, this is all getting generated per query, right? And what's interesting is Asia on our team, she and I were going back and forth about this and for example, I was getting way more videos and a lot of the things I was looking up than she and others were. And I'm like, oh, I wonder if it's like knows that I'm a big YouTube user, right? And like knowing that I'm A bigger video consumer and showing me those videos more. Right. So what's interesting is we've done enough back channeling to know that these are very much like unique to the individual. So it's almost like a custom generated search engine results page for everybody.

[Kieran]

Why would you go and use this versus search? Why would you separate out those two actions? I'm just curious, like how do you think about this and then how do you search?

[Kip]

Tell me more.

[Kieran]

You've searched here, how to, you know, learn about CRMs.

[Kip]

Yep.

[Kieran]

And in a world of this was a separate product, how would you kind of delineate between what you would just search on Google and what you would go to their learn about product? Because you could obviously just search for the same query on Google or the same information. It just wouldn't be as nicely presented.

[Kip]

Yeah. So first of all, I think I agree with you that I think this is an experiment to figure out how you integrate this into search. Because the actual answer to your question right now is that this is a better, just like straight up text based search experience. What it doesn't have is the integration with maps and YouTube and everything in the way traditional Google search does. That's a very solvable thing. The reason I imagine it doesn't is because this is a very focused experiment. Right. But that's the only reason you would use this and Google search in Parallels if you wanted some of the very specific search features. I think this is a better experience for most of my queries compared to traditional Google search.

[Kieran]

Yeah.

[Kip]

And what's interesting, Kieran, is I decided to do the reverse of what I just did. And this is pretty fascinating. And I haven't looked at this yet because I wanted to see what it would do and I wanted to look at it live with you. So I uploaded an image of the HubSpot Workflows product. I found a screenshot on the Internet and all I said was, can you explain to me what's going on? Is this a good way to use HubSpot workflows? What other tips or advice could you share? And it's literally just a screenshot of the workflows product. Okay. It creates all of these suggested topics around HubSpot workflows. Workflow triggers, workflow actions, workflow branches, linear versus nonlinear potential inefficiencies, best practices. It's pretty wild what it got from that one image. And it's like this image shows HubSpot workflow that uses a go to action feature. This feature allows you to jump to a specific action within the workflow, rather than having to follow a linear path, in this case, the workflow is checking the value of a multiple checkbox property.

[Kieran]

I think that's pretty cool use case. Like it can reverse engineer things.

[Kip]

That's pretty amazing.

[Kieran]

Reverse engineer subjects and teach you about them. Like, it is a pretty good tutor. And so maybe that's another use case that they're trying to look into, which is can you have a tutorial companion and have people learn through this, which I think AI is really good for.

[Kip]

Yeah, I mean, one of the things we've talked about on the show before is that like, AI, especially in the short term, is going to be great at helping us get unstuck. And like you could imagine just being in HubSpot or any product. Right. And just not quite know how to do something and just take a screenshot and use AI. In this case, learn about to basically tell you like, oh, this is what I'm doing. What am I doing wrong? This is what I'm trying to accomplish. And it will give you the context and the information and tell you what to do, but also give you kind of the next steps of drill downs of the things you might want to learn about.

[Kieran]

Yeah, yeah, I can see it being useful for going really deep on a topic. I just wonder if they will integrate those learnings back into search because it's just a evolution of search. It's kind of what they talked about their search product becoming. And it might be hard to have a separate product to kind of delineate between when do I use search for a query and when do I use learn about for a query.

[Kip]

Correct. The other thing, Kieran, that I have found it to be really good at is if you're doing like some hardcore research, I'm really enjoying it. So I had a separate tab on email marketing and you start to see I'm getting those YouTube videos. Right. So it's integrating these YouTube videos straight in. So one of my explainers here sends me off to YouTube. But one of the things that I think it's good at is like, can you give me some data and studies about the effectiveness of email marketing? Please give me the text and source links. So let's say you do all this research, you're interested in something, but you need to like understand what the data in the world is saying about this. Or I need to go build a presentation to explain to other people why we should do email marketing. It's very good at going out and getting you stats, attributing those stats and especially if like I was doing some stuff where it was looking at like scientific and academic papers and pulling out the relevant quotes linking me out to them so I could go read the full paper if I needed to in a much smoother way than traditional search.

[Kieran]

Yeah, it's just a more comprehensive deep dive into a topic, but it's still primarily using functionality they have in search and they are kind of having Gemini. So I think that's what's interesting to me about it is what are they trying to actually figure out here? And to me, what's different is none of the technology is different. It's using search, it's using AI, which is likely Gemini. What's different is actually the UX and how you can navigate someone around a topic.

[Kip]

I would actually argue there's one other core different thing here, and this is just one human's opinion. I have no information. Right. What do you also notice a lot in the learn about results that you don't get as much of in search GPT, perplexity, et cetera. The thing that I see here, Kieran, one of the big differences between this and search GPT and perplexity, everything. There's a lot more links. There's a lot more links. And like in each of these sections it's providing resources. Like, part of me wonders if this is an experiment to say, hey, can AI search be a better curator of information and still send people to other destinations a la for AdWords or for traffic referrals versus just like giving you the answer on the page.

[Kieran]

Yeah, right.

[Kip]

Cause I find myself really actually going off the search engine much more with this experiment than I do in like something like search GPT for example.

[Kieran]

I think in ChatGPT search engine they do have a lot of citations. They have citations and the answer, it gives back to you embedded within the answer. And then they have the button at the bottom that you can click on and gives you all of the sources that they use. Comes up in the right hand panel.

[Kip]

Yeah. I think because of the UX here, they're more front and center.

[Kieran]

Yeah.

[Kip]

And it's sometimes it's prioritized. Like it's giving me the YouTube thumbnail. I'm like, oh cool, I do want to go watch that video about email marketing.

[Kieran]

Right.

[Kip]

Like it's like the actual user experience of it is different for some reason I just find myself clicking through way.

[Kieran]

More here, which is what they want.

[Kip]

Right.

[Kieran]

Like that's what I think this is an experiment on is user experience patterns. I think that they are trying to Figure out what is the right conversational user experience pattern for users long term to get them to search more, to get them to use the product more. If you actually think about Claude, ChatGPT, all these other AI assistants, they have not really evolved the user experience on that conversational pattern. It's really just back and forth in the way you would talk to a human. There's some interesting things in the way ChatGPT search engine gives you the results that you can click out on citations. They obviously both now have the artifacts, which is a different type of UX experience. But I don't think any of them have really figured out like, well, what is the conversational experience that more people do through chat? I don't think it's just going to be this one all back and forth conversation. Like that's how you're going to interact with information. I suspect what Google is trying to figure out here is like, what is the right user experience for people to get them to click around and to engage with this stuff much, much more. And to me, the benefit of doing search through AI has always been speed of iteration. And you even mentioned it. And what you really enjoy here is like the way it dynamically builds different navigation for you to navigate around the topics, the way it pulls in multimedia, the way it actually links back out to research. That to me is what I think they're likely experimenting with, which is like, how do I get more users to click on more stuff and enjoy the experience more and to use our tool.

[Kip]

More well and to be able to navigate a topic much better. Right. Like, I think one of the flaws with just doing chat based search and research is that you have a long chat thread and it's really hard to go back and navigate that topic. Right, right. And what they've done is they've said, oh, let's experiment and say the future is both this kind of chat interface with this like dynamic navigation that lets you kind of go back and forth throughout that topic in a much easier way. It is way better. Like I do like it much more.

[Kieran]

Yeah, I think that's the interesting part is that you really like that much more. You can get the information you want much easier and it's going to be interesting to see if they start to integrate some of this back into search, the core search experience, which I assume they will totally.

I want to tell you about a podcast I love. It's called Created. It's hosted by John Ushayak, it's brought to you by the HubSpot Podcast Network. The audio destination for business professionals. Created gives an inside look at how to build businesses as a creator, improve your creative process and stay ahead of the latest content trends. John, an ex YouTuber and Instagram employee, breaks down the creator economy with guests Logan Paul, Paris Hilton, Jason Derulo, Jake Paul, George Lopez, and more. In a not so recent episode I loved was with Liza Koshe.

[Kip]

Remember her?

[Kieran]

She was one of the fastest growing YouTubers and Viners, growing a fan base well beyond 25 million subscribers.

[Kip]

But then she stopped uploading.

[Kieran]

So what happened? That's a smart woman who made a drastic career decision. You'd want to tune into that one. If you're looking for insights on entrepreneurship and creating a powerhouse business, then this is the episode you won't want to miss. Listen to Created wherever you get your podcasts hey, it's Kip. If you listen to this podcast, you.

[Kip]

Know how much I love keeping up.

[Kieran]

With the latest and greatest in technology, but very few podcasts actually give you a dose of the future. The A16Z podcast is the exception. It's hosted by our friend and frequent guest Steph Smith. The chart topping show brings on movers with a track record of being both early and right, like Apple co founder Steve Wozniak or the CISOs of OpenAI, Anthropic and DeepMind, even the very first CTO of the CIA. From the science and the supply of GLP1s to drone delivery to the economics of deepfakes, go ahead and eavesdrop on the future. Check out the a16z podcast wherever you get your podcasts.

[Kip]

And that's what I want to get into kind of for the last segment of the show. Here is where we're at in the war of search engines and let's give people a pulse and update of like where we think things are going. You know, I would set the table that we are in November of 2024 and in the last couple weeks we've had search GPT released by OpenAI. We have had Perplexity trying to do more and more partnerships, trying to scale their user base. We've had Notebook LM come out from Google, we've had learnabout come out from Google. It seems like we're actually getting real tangible innovation in search right now in a way that like we kind of talked about over the last year. But it seems like it's here. And the one other thing I would share is like we are starting to get some real traffic from perplexity and search GPT and some of these AI search engines to HubSpot.com so it's starting to show me that, like, search adoption is certainly shifting.

[Kieran]

Yeah, I think it's shifting within, you know, smaller groups. It's not really impacting Google's overall numbers. It is for some companies. You and I know there's companies who are really seeing some drop in traffic. I think we are still at the starting point. You have ChatGPT only launching their search engine a couple of weeks ago. They got 3.1 billion visitors in September. So there's a lot of people that AI search can be a front door for. For them, AI search is actually the best way for them to increase user adoption as a whole. Like, it's a front door to the Internet. It's a really common use case. It allows you to get all manner of users using AI without them even thinking about using AI. We are like really at the starting point of a ton of innovation here. Like you see Google roll out AI overviews to a hundred countries, so you're going to start to see it a lot more in your search results. I think their AI overview experience is really hit or miss. I still don't think it's like incredibly good compared to Perplexity and compared to ChatGPT's engine. And I think what we do not know is like, what is the user experience. There's no part of me thinks that historical blue links disappears completely. I think it gets interwoven into AI. What AI is, is really a layer on top of those links to be able to clearly and concisely give you information. But you probably in some cases still want to click around and do some of that yourself. But I think the interesting part is like, we don't know what the right user experience there is. We don't know how multimedia gets integrated into that. I think there's much more room for video, there's much more room for audio in how you can integrate this stuff. So where we are today, I think next year is going to be a real year of continual disruption. I think perplexity will get better. I think ChatGPT search engine will get better. I think Claude will try to get much better. I think Google will continue to roll out AI overviews very aggressively and try to improve. I think they'll have experiments like this where they try to figure out, well, what is the actual new experience for search? How do we actually integrate new user experience patterns to get people to use both the historical engine we have and the AI engine we have? Like, we kind of call that this would be incredibly disruptive for Google. And I think that's what we're seeing, right? I think that we are starting to see Google can really feel the pressure they're being put under.

[Kip]

And I guess my question for you, if you were a lowly marketer out there like we are, and you see all this change in search happening, what would you recommend that people do today to say, like, hey, we don't know where everything's going, but we kind of know that you should be doing these types of activities to properly prepare. What's your advice to everyone?

[Kieran]

Yeah, I don't think there's like great advice around this. Right. Because you can do what we did, which is categorize your content into these different categories to assess your risk. So what we did was put it into like most at risk, which is just these basic posts that you have, which is giving you informational content around a query. It's like really easy for an AI to just cannibalize. Then you have this really big bucket which is, well, there's some real effort went into creating this content. There's some unique customer insights, there's some unique research, there's some unique expert quotes. And so like that's in the maybe cannibalized, but harder for AI to cannibalize because there's some things that are actually differentiated and actually you have real uniqueness to that content. That's not just easy for AI just to say, well, here's the exact thing, but better. And then you have your stuff that is still for the most part fine today, but I don't think it's fine long term, which is just transactional. If you're searching for a product, if you're searching how to complete an action, we don't see any cannibalization of transactional traffic. It's all an informational. So if you are looking at this wisely, you would say, well, you can't do anything about the most at risk bucket. Right? AI is going to be AI, it's going to be better. Your at risk bucket. There isn't like anything you can do, it's just user adoption at the moment. We're still in the very much early adopter phase and so you might not be feeling much pressure there. But as user adoption increases and people realize AI is a much better experience, your at most risk bucket disappears and there's nothing you can do. So now you move on to the middle bucket, which is the bigger bucket, and you're like, well, what can I do to make this much, much better? If I Interweave AI into how content gets created and I reduce the amount of time that all of my people need to spend on these other tasks. Well, now they can spend 10 extra time on trying to figure out what actually makes content special differentiated. Something that AI cannot just replace. One of the examples I've used recently that seems to resonate with people is like we've obviously had Lenny on, we're big fans of Lenny and if you actually think about what did Lenny even do now? He's gone on to do lots of things, but where he started was he really just differentiated his content by doing this example based content which really he was a expert journalist for his industry. Right. That's the way I would think about him is like he had deep expertise in his craft and he became a journalist for people who wanted to learn more about that craft. And the way he separated his content from everyone else is he integrated examples from his network to make his content unique and to make his content differentiated. So when he writes a content piece about here's how you self churn as a SaaS company that if it's just an informational piece of content on how to do that, AI can replace that. But his was, well, here's how you solve churn and here's benchmarks from my network and here's experiments and the upside benchmarks from those experiments and here's where they actually experimented and here were some of the experiments they actually run. AI can't replicate that because it did not have that information. Now if Lenny has allowed them to crawl his content, it has that information in future iterations. And so you have to stay one step ahead. But that's the thing I would figure out is like the middle bucket, what is the thing that we are going to invest in that is going to differentiate our content from the AI just being able to replace this and from all of our competitors. And I actually think that is just a wise thing to do anyway because you always want to be trying to figure out what your leverage is.

[Kip]

I agree with all that. I would also just say, you know our friend and co founder at HubSpot, Brian Halligan, he always had this cartoon that he liked which was of a dog on the computer and all it said was like on the Internet nobody knows I'm a dog. I actually think that that really relates to AI search, which is AI search is trying to take a proxy of the information about you or your company that's on the Internet and give people answers and kind of estimate priority. There is going to be A period of time where companies that might be smaller can have an outsized footprint on the Internet. You know, like if you're really, really aggressive in content, product documentation, YouTube videos, community forums, you might not be the biggest company, but you are going to show up in all of these AI search experiments in the future of AI search. So where I think it used to be for traditional Google search, it's like, hey, I need to create these text based contents around this certain topics now it's like for my brand to appear larger online, both from like a product and information perspective, I need to mobilize the people who do know about my brand from a community perspective so that on sites like Reddit and everything, I'm represented and if so, I'm going to show up probably more favorable in the models than I would otherwise. Right, okay, well, do you agree with that or not?

[Kieran]

I don't disagree with the premise actually. We have the product leader for Claude coming on. But what we really want to know is, well, why do large language models pick one thing over another? Which is the thing that you are, you know, asking and for what I know at the moment and what we've talked to some folks on, and even we had Ethan Smith on this podcast is there's a couple of ways to think about it and the one that makes most sense to me is the large language models are trained on the Internet and they are going to look at what are the most common words cited with a thing. Like what is word co citation? So in that case it would say hey, every time I see CRM small business HubSpot is the next word or the word before. Right? And I'm sure someone who is much more scientific at this would say that's completely wrong. Or there's a different variation of that. But word co citation makes a lot of sense is what Ian said, which is, hey, the more times you're co cited with the words that someone asked the LLM, the more likely it is you're going to appear in an answer. And that means the opportunity right now is which is the thing you're making the point on is hey, if you're a small company and you can get a bunch of material on the Internet that basically co sites you with these other words, then you're more likely to appear in some of these answers today. The thing is, I still think that's like a text only thing. I don't think the core engines are training on YouTube and audio as much as I know they've all kind of trained their video models on scraping YouTube, which is a whole problem that they're trying to figure out. But I think that on the text based Internet, if you are mentioned a bunch of times with words that are related to your product or services, more likely you are going to appear in the search engine. So what actually you see search happening is if more people gravitate towards AI search and these AI assistants, the way you appear in those engines is just through good marketing and how to get more awareness of your brand and all of the optimization stuff and the stuff that used to do as a search engine optimization person disappears. None of that is actually that relevant for how you need to appear in these other AI search engines.

[Kip]

I think there will be a period of kind of new school SEO where like maybe even something as simple as like, hey, I've got this software product and I'm going to like post screenshots with annotations of how to best use this product so that LLMs will take advantage of it and cite me and include me.

[Kieran]

Well, the reason I was like saying I mostly agree is because when I first started to read into how you could potentially appear in LLMs, because when they first came out, really the thing I just did was like really try to figure out how it was recommending one thing over the other because that's really the thing that you really care about as a marketer. And so I spent a ton of time trying to figure that out. We had people on the podcast trying to help us figure that out. And if it is just as simple as that, which is like, hey, the amount of mentions you get in relation to words or services related to a product, you're going to appear in those answers, obviously I went straight to like, well, what's the hacked version of this? And the way you could hack this is to inject a bunch of content onto the Internet using AI itself to publish mass amounts of content to poison the training model to make sure that you are going to appear. And so that would seem too easy. But I don't know like historically, like why did Google win? Well, they had Page Authority and Page Authority used backlinks to try to assess like quality over quantity.

[Kip]

Yeah.

[Kieran]

And I don't know what the version of Page Authority is for the LLMs, but they must have something or they're going to have something because I, I don't think they're going to want an Internet where you can just mass inject billions of pages of rubbish to sway the model to actually have your service or product appear because you just have so Many more mentions than anyone else. But all those mentions are on like single page sites that you've just set up to actually get more content in the LLMs training model. Yeah.

[Kip]

So I think we're talking about a couple things here to kind of frame it up for everybody. One, I think we're saying, Kieran, you don't want to slow down on whatever content strategy you have, because having a good content strategy and online presence is just going to get you included in the LLMs, you know, at the most basic level. Right, right. And as search evolves, you're going to have a better shot. You then brought up this issue of like, how does it pick you versus something else? And that's where there's a lot of uncertainty. Right. And so it's like if you have a good content playbook, you're going to be in the game. And then it's like, how do I win the game? We don't know the answer to that yet. But what I can tell you for certain, in the history of marketing, when new platforms come out, there's always cheat code, ways to game the system and opportunities, especially in the first few years, right?

[Kieran]

Yeah, there's a temporary cheat code.

[Kip]

Yeah. We're going to learn those together and share those together here. So hit that subscribe button if you haven't already. And that's what we're focused on. And then I think the last part of it is how fast does the user behavior switch to these new search engines and what is kind of the old Google market share look like relative to the new search engine market share? And I think we're going to learn a lot more about that over the next six months. And one of the things that we're going to do is track data from some of these new search experiments and search engines and share that on upcoming episodes so that we can share that with all of you. All right, so that was Google learn about that was a little bit about the future of search and how we as marketers might be able to take advantage of it and kind of how we should lay the right foundation right now. Kieran, anything else before we close out today?

[Kieran]

No, I think what people really want to know is like, how do these LLMs actually choose something over another? And I think we can go find the person who naturally builds LLMs and knows how that works.

[Kip]

It sounds like a good follow up show.

[Kieran]

So that's the thing we can do for our audience here.

[Kip]

We'll be back with that. We'll kind of try to do a series over the next six months of AI search and the different components of it and kind of LLM prioritization is a good next episode to follow up with that. Thank you so much for watching or listening to today's show. We'll see you real soon on the next episode of Marketing against the Grain.