7:05
Hello and welcome back to Equity Tech Branch's flagship podcast about the business of startups. I'm Rebecca Balon and this is the episode where we bring on industry experts to help us explore a trend in the tech world and dive deep. Search as we know it is getting a serious makeover. Google IO made it official last week, putting AI generated answers front and center. Brands have almost no visibility now into how AI is describing them to their customers. So SEO needs its own makeover in response. Startup Scrunch is positioning itself at the center of this shift, trying to monitor and optimize how brands appear in AI platforms. Today, we're joined by Matt Thompson, the VP of partnerships and channels at Scrunch.
Matt, welcome to the show. Thanks, Rebecca. Excited to be here with you. Yeah, excited to hear your perspective because it feels like your whole startup has been preparing for this moment, for the moment that search really changes. I mean, it's been changing as we've had AI throughout different platforms, not just Google, obviously, ChatGpt, Perplexity, take over different parts of search, but I'm wondering if you can just give our readers, anyone who's been under a rock for the last week, a little bit of an overview of like what was announced at IO and why it matters. Yeah, sure. And I was out there for Google IO and for Google Marketing Live.
The buzz was pretty palpable. There were a lot of things that were released. There were updates around Agentic Commerce. There were updates around Google's new TPUs. There were updates around Google glasses. But the notable one is that Google is going allin on AI search. Which is funny because a lot of people say may say, well, of course it's Google. Of course, they're going all in on AI search. But 18 months ago, this whole AI search space was a bit of an existential threat for Google. It is very different than traditional search. Google is an advertising and search juggernaut. It's one of the most unbelievable business models that has ever been developed in history. And it's
all built historically really around 10 blue links. And so Google 18 months ago was really scared when the early moments of that chat GPT moment happened and people started conversing with these search agents or these search chat bots and Google was really kind of late to the game there. But what I will say is they have caught up in a very meaningful way. Uh and they're going all in on it and most of last week's announcements were all about that. Yeah, it's really interesting seeing just like on a consumer side the way people are kind of freaking out about it. You know, there's a lot of people who are like this is the obvious next step, you know, to have overviews be the dominant
thing that you use, but then a lot of people just I've heard even walking down the street, I heard someone explaining the change to a friend or something on the phone and saying, "Well, Google basically isn't Google anymore and I started using duck.go," which I thought was really interesting. I think that we can all feel that the way that we search has changed, right? We're searching in like kind of full sentences. We want a conversational response. We want to be able to ask follow-up questions. And startups like yours, talk a little bit more about like what Scrunch does, how it helps brands or companies that have websites ensure that they're not getting lost to the winds. So, at Scrunch, we help our customers
get their websites agent ready so that they can show up more in AI search. And there's a few different components to that. One is first understanding at a baseline level how are you showing up in AI search? A lot of brands today don't know are you showing up for the right prompts or questions that you want to be found for or are you showing up for maybe the wrong ones or are your competitors being found instead of you. So there's an entire amount of intelligence that's needed just at a foundational level to understand what might be working and what might not be working in your search strategy. And that search strategy is pivoting a lot more toward AI search these days. So we really help inform our customers and
provide them with a lot of data through our platform to answer a lot of those questions. Then once you answer a lot of those questions, then the next question that any savvy marketer asks is, okay, well, this is all really helpful data that helps me understand my share of voice in the market or what my visibility may or may not be, but what should I do to optimize this? What should I do to get cited or mentioned more by AI? And so we then help our customers create very specific, we call them scrunched down versions of their website through our agent experience platform and we will optimize their site so that it can be more easily parsed and found by AI agents on behalf of a human who's on the other side of that prompt.
I have so many questions about how that works and how it works across different platforms. But before I dig into that, going back to how Google's changing in the postarch era, right? Because I think that will inform a lot of the way that you treat your customers. What is it about the ad side that has changed? Like what has changed and what also stays the same? Because I remember us talking about this previously and SEO is still kind of a thing apparently like it's not completely gone, but the way that AI agents will interact with it has changed. Yeah. SEO is definitely not going away. The search landscape, the total addressable market, the number of surfaces you can now search on is really
expanding. So while there's less web traffic, there's actually more search queries. And so this was initially when we I talked about that existential piece for Google, this is what they were initially worried about. their most of their search came through the web browser, came through google.com and that's where a lot of the these search ads were found. But now, if you think about it, there's a lot more surfaces for advertisers to advertise within. So, of course, there's still 10 blue links. Those aren't going away anytime soon. But now, there's AI overviews. There's AI mode, which has over a billion monthly active users, and it's doubling quarter over quarter. It's insane growth
in AI mode. And then there's Gemini. So Gemini powers both of those two, overviews and AI mode. But Gemini also has its own interface just like a chat GPT does. So there's a fourth surface. Marketers can now place ads and really programmatically place them in any of those four Google surfaces to try to reach their customer. So the search landscape has expanded dramatically. Now, you could argue a marketer's job is now also much harder because where do they intercept their customer at the right moment in time to give them the right message as they're searching? And this is where Google released a lot of
interesting tools for marketers. They released some updates to Pmax, Performance Max, to AIAX. That is basically saying allow Gemini to orchestrate and arbitrate your ad spend and we will find your highest value customers and we will place your ads in any of those surfaces where we're finding the customer. So like a jumped up programmatics like is that how you can think about it basically? Yeah. And it's the early days here, but the Google was releasing a lot of interesting statistics talking about how when this is done right, your rorowaz goes through the roof. And a lot of it is, you know, Gemini behind the scenes matching the right ad to the right person.
Can you explain to our audience what rorowaz is? Your return on ad spend. So, as an example, like a classic brand, a Coca-Cola as an example, spends billions and billions of dollars a year to try to get the right advertisement with a polar bear cracking a bottle of Coca-Cola and drinking a very refreshing sip in front of someone like you or someone like me who can go buy that Coca-Cola. It works every time for me. That little bear. There you go. Well, see, you probably like Coca-Cola. I actually don't drink much soda. So, this is a great example. Them putting that ad in front of you, you're saying it works every time.
I don't actually like soda. It just makes me want one. It's It's crazy. Well, well, that's the point of advertising. So, it to some extent it is implicitly working. Putting that ad in front of you is probably of high value. Whereas for me, I don't really drink any soda. Putting that ad in front of me is of much lower value. And what Google is basically saying is we will find for that ad a lot more Rebecca's for you and a lot fewer mats and we will do it for you. We'll find your ICP, your ideal customer profile, and we'll get those ads to them in the most efficient way possible. So, one of the big things in terms of Google IO's like search announcements was that, you know, everyone will have their own
personalized agent that will be doing a lot of this searching for items for you anyway, right? And we can talk a little bit about that later, but one thing that just occurred to me is that kind of part of the value proposition here for Google to advertisers, it's like, oh, we'll know exactly how to place an ad in front of the right person because we don't just have a history of their search results and what they do online, but we have a history of like, you know, their very personalized agent and what they search for. Yeah, that's that's exactly right. And this was some of the updates that were released around the commerce element. So, Google has more commerce listings via Google Shopping than any other
platform out there and they have for a long time. What they just released was this idea called universal cart. basically saying anywhere you are within the browser, whether it be in Chrome on your Gmail, so you're in a Google app inside of a Google browser, which is technically another Google app, or if you're just surfing Wall Street Journal, within the browser, you'll be able to have your cart of all the various different websites that you have been shopping at and adding things into and be able to check out anywhere you are. And so Google having all of that information stored in a cart to be able to do lookalike modeling and to be able to do some really interesting advertising modeling is going to be a huge
differentiator for them against a lot of other advertising platforms that you could go to. And I think this speaks more toward Google's advantage going forward, not just in advertising, but across all of AI. Google is one of the very few companies that own the entirety of the AI stack. They own and are developing their own silicon, their own chips, TPUs. Google has its own very robust and growing cloud platform with data warehousing and a lot of different AI workflow orchestration abilities built right into their cloud infrastructure. Google has models. They just released Gemini 3.5 Flash, which is a more efficient, smaller model that was just released last week. But they've got a bunch of different models, mostly now
under the Gemini brand family. And then on top of those models, they've got various different apps. Gmail is an app, your calendar is an app, Google Docs, uh Google Sheets is an app, but then the web browser like we're talking about here for universal cart. The web browser is an app. Now, I know a lot of other organizations like OpenAI are talking about we'll we'll have our own browser. Maybe that'll take off, maybe it won't. We'll see. Perplexity has come out with a Comet browser. So a lot of people are realizing that the future warfare between these companies may actually happen at the browser level. But at all
of those levels, chips, cloud, models, and apps, there are very few maybe one other maybe company in the world that can dominate it at all levels. So Google's well positioned going forward. Yeah. I mean, well, we're making the case for Google being a monopoly. And if those of you remember about a year ago, Google lost its um landmark antitrust case that it had monopoly in advertising. Some of the proposed remedies were breaking up its advertising business, forcing Google to share some of the data that it has. And you know, it's stuck in a massive appeals battle. And who knows if we'll ever have a solution, but something tells me that they're going to they're
going to keep chugging away making the most out of this monopoly that they have. So what does this mean then for an advertiser who just wants to know where their dollars are going? Yeah, a lot of that is going to be how this can be measured in Google search console which again will be not just traditional search but it will be all of AI search as well. Google Analytics is going to play a big part of that. But a really important thing for these advertisers, and Google made a big deal of this at their keynote stage at Google Marketing Live, is making your web properties agent ready. If you have the best content in the world on your website, but these agents
or bots that are coming on behalf of a human from chat GPT or from Claude or from Grock or from Perplexity are coming to your site and they can't read or an agent language. They can't parse that information on your website quickly and easily. It doesn't matter how pretty that web page looked. It doesn't matter that web page had a lot of great web copy on it. If all of that web copy was hidden inside and nested inside of JavaScript, it doesn't matter. And so this I'm actually curious about that. I do want to talk about that. So I know that one of the you know, you spoke a little bit about this already about what Scrunch does and just give us like the oneliner again.
Yeah. We help our customers get their websites agent ready so that they can show up more in AI search. Right. And so part of that I mean you said like nesting into JavaScript could be a bit of a problem. give me some examples of exactly what kind of a service you do because you've got a few products, right? And just as an aside, you guys have raised what was it? Raised a $4 million seed last March and then a series A last July, 15 million. Any thing that I'm missing there? Nope, that's that's about right. Yep. Okay. And then just landscape for our listeners. Um overall, AI search startups are really starting to blow up because of this, right? Exabs reportedly raised 250 million at a $2.2 billion
valuation. There's parallel that's another competitor. But anyway, you guys have a few products, one of which is AI search visibility monitoring across AI systems. Talk a little bit about that. Yep, that's one of our products. Our other one is called our agent experience platform. So, I'll talk a little bit about both because they are quite interlin. The first one basically involves step one for any marketer's journey, which is answering the question of how am I showing up in AI search? Am I showing up for the prompts that I want to own from a brand standpoint? Both the branded prompts I want to own and the unbranded prompts that I want to own and be known for. Am I showing up? Am I not
showing up? Are my competitors showing up or am I showing up for the right things? Am I accurately showing up? So, there's a lot of different elements that go into that. So, we've got an entire AI search analytics platform basically that helps you monitor, measure, and derive insights from how your brand is or isn't showing up in AI search. Can you as you do this um I thought it was really helpful when you explained the parallels between like you know a last decade's worth of like SEO SEM like what are the parallels between what you're offering for AI search versus what you know a company would have offered for just SEO search.
It's similar but different. Um SEO was a lot about keywords. This is Google's what Google's latest update really helped to unearth is that the days of typing in two to three keywords and generating 10 blue links. Every month those days are going away more and more. And what people really want to do is chat with search. Search is becoming much more conversational. Google was saying as much as AI mode is exploding and doubling quarter over quarter, there's a 30% larger growth on brainstorming queries above and beyond the growth that AI mode is seeing where people want to brainstorm. They want to
ask these many sentences open-ended type of questions. It takes a lot of compute and a lot of effort to get the semantic meaning out of those versus just keyword best running shoes. We don't have to say best running shoes anymore. We can say I'm going on a trail run that's about four and a half miles long with one of my best friends. It's going to be up in Maine. It looks like it's going to be a rainy day where my shoes are probably going to be all messed up with mud and water throughout this trail run. Give me the best shoes for me for this run. Like exact. Yeah. Exactly. And eventually you'll be able to say, you know, I'm a size 11 and a half and go purchase them for me. We're we're almost there. But those are the types of
queries that are being asked. The semantic meaning is a lot harder to understand. And so there's this idea of query fan out where that query that I just rattled the huff off the top of my head there, that'll fan out into maybe 10 other queries. And so that'll go out into the web to really understand what the best information is that should come back to me. Our platform through a variety of different data signals across a variety of different inputs is trying to understand all of those different query fanouts across API data across web crawling data across different panel data and we're aggregating all of that
to bring back that information for you. So it's a lot more complicated than what SEO was. The other thing is the middle of the funnel is very much collapsing because a lot of people are doing all of their research in like a chat GPT or a quad or a Gemini and they're getting an enormous amount of information there such that when they arrive on a customer's website, they're ready to convert. the research has been done. And so we're seeing that AI referrals are converting at 400% higher than traditional organic search because they're coming to the website, the sale is already made. They've already done all of the research. So while traffic might be going down, the amount of queries are going up and the
quality of the people that are coming to the website are generally much higher. So it's it's one of those things where the metrics that people looked at for SEO and the metrics that you need to look at now we're finding each month these metrics are not the same metrics. And so then how does that play into the second product that you offer the website agent readiness? So, this goes more into you could have your content from a copy web copy standpoint optimized in a really amazing way, but if it's not if your website isn't agent ready, if it's not built agent first, it may not and likely doesn't matter. And we're at this inflection point right now where AI search is growing so much and
it's becoming the dominant way that consumers want to search. However, the issue is basically up to this point in time, all websites have been built not for agents. They've been built for humans. They've been built for you and built for me. They've got beautiful imagery on them. They've got rich parallax scrolling. They've got a lot of JavaScript. They've got rich multimedia. All of these things are great for a human and they're great for my eyeballs to look at, which is fantastic. But what we're seeing with all of our customers is that there's actually more AI bot traffic on most of our customer sites
across industries than there is human traffic. And the disparity between the two is growing month over month where the human traffic is going down. Again, like I said, because people are doing all of their research in these chat interfaces and the AI bot traffic is going up, which makes sense because at the time of inference, when I type in, give me the best running shoe to go running in Maine or whatever longtail query I put in there, that query at the time of inference sends a retrieval bot over to Nike and Adidas and New Balance and to Reddit and to YouTube and to a lot of different websites basically to pull back information about my query. So it makes sense that there's less human
traffic because I didn't leave chat GPT to make that query. But when I made that query, it sent out a bunch of AI bots on my behalf. And so if your website is not able to have those bots hit your site, quickly parse the pages and come back to the user on the other end of that query with an answer that's accurate, timely, and relevant, then your brand is behind. And so what our agent experience platform does is it takes the key pages on your website that have a lot of like I said imagery, JavaScript, rich multimedia, things that agents can't read or parse and have no semantic meaning and we will strip those away and we will scrunch down basically the codebase or in other words the token base. Again these large language models
are called that because they want language. Language is tokens. We will scrunch down the token count on those web pages and then serve them at the edge via partnerships with a variety of different CDN players and then serve them at the edge to the AI bot when they come to the website. So to be very clear, a human still sees the human experience, but when a bot comes at the edge, we serve them an optimized scrunched down web page just for the bot. And this is going to be a really interesting paradigm shift going forward. There's all there has been this talk for the last 20 years about personalization for humans. We are now going to be shifting that personalization topic to creating personalized experiences for AI bots on
behalf of humans. Okay, I'm going to double click on that in a second, but a couple questions first. One, did you say this already about what like how many tokens are on a typical enterprise homepage? So on any given page, you may see 100,000 tokens plus. Most of those tokens are not the semantic meaning of what that page is trying to communicate to an end user. A lot of it is code, a lot of it is JavaScript, a lot of it is image tags, etc. And so what we can do is we can remove all of that, scrunch the token base down by 98 or 99%. So what was a page that had 100,000 tokens will now have maybe a,000 or 2,000 tokens. We find generally these agents aren't going to read more than 10,000. So we get well below that 10,000 token mark. The agent
can quickly come parse the page quite quickly, get the information in HTML that they need and then return back the right answer to the user on the other side. And when our customers are doing this well, they're seeing that their mentions of their brand on these scrunched pages are going through the roof. Okay, interesting. So, the data is showing that it does work. So, Google's guide to optimizing AI search, right? They talk about they're like, yes, make your, you know, website agent ready, but then at the same time, they have like this whole myth busting section at the bottom, right? So it includes stuff that you don't have to do, ineffective hacks they call them, like building machine only layers or rewriting content specifically for
agents to be indexed, content chunking, etc. Isn't this a lot of what Scrunch is offering? So couple things. First, Google is just saying what you should do to optimize for Gemini. Really, they're not saying what you should do to optimize for all of the other LLMs and chat GPT still having most of the traffic. So Google is coming at it with ChatGpt has most of the search traffic for AI search. For AI search, chat GBT today has I think 55 it might be 60% of AI search traffic goes through chat GBT. Gemini has come a long way. They're now in second. And then there's basically Anthropic or Claude Grock and Perplexity that kind of are all jockeying for that third spot. But yes, a lion share of the traffic goes through chat GPT. So number
one, the fir the first answer to your question is you can't just optimize for Google because you're going to miss most of the market, at least with the way the market is set today. Number two, a lot of what Google says around these SEO best practices are still valid in terms of you need to have content that's unique, that's helpful, that's authoritative. Those are all things that you should have. And how you chunk out your content to your point may or may not be relevant. But if the content is hidden inside JavaScript, it can't be found. If the content is on a very large page that a lot of these models when they come to can't find the information very easily and therefore will then go to a competitor's page, then also you're you're not doing
yourself any service here. So this is one of those things where the best practices again with SEO it's SEO has been around for 20 25 years. It took at least five you could maybe argue 10 years to have really concrete best practices around SEO. We are in the first inning with GEO or AI search to say that there are standard best practices today that everyone can concretely follow and know that those things will drive the results. That's not there yet and I wouldn't expect it to be just yet. But what I can say is that for the c our customers that are using our Asian experience platform, we will do AV tests and we will benchmark the pages that they're scrunching that they're deploying and how often those pages were mentioned before.
And if those pages were mentioned maybe 2% of the time in AI search, we're not just seeing that now they're mentioned 2.5% of the time. We're seeing that they're mentioned 10% 12% of the time. These are enormous, you know, 300, 400, 500% lift in mentions and it's within their branded ecosystem, which is really important because a AI search for the most part likes to reference third party content like on Reddit, on LinkedIn, on YouTube, things like that. And so these marketers are saying, well, how can I own the narrative more? Right? One of the ways you could own the narrative is to optimize your owned properties, your website, and make sure
it's really easy for an agent to parse. Right? One thing that I thought was really interesting about our chat was that Google says that they still need AI agents still need unique, interesting content, and that they're kind of like downgrading content that's all written by AI. That kind of almost makes the case for, you know, human copywriters to get their jobs back, maybe. Who knows? Yeah, exactly. I mean, it's never been easier to create content ever. Type in a prompt and tell it to write you a 30-page dissertation and it will write you a 30-page dissertation, which is a wonderful thing, but it's also
something that a lot of people are doing to create new web pages or create new blog posts or optimize their website. And a lot of times the copy when you type in a pretty generic prompt and so does a hundred other people who all have similar websites, the copy and content across those websites looks pretty similar and then your SEO gets dinged in a really big way. And this is another reason why our agent experience platform is so important is because a lot of other folks in the industry right now will say, "Oh, well, to help fix some of your visibility gaps or discoverability gaps, we can help you create a ton of AI content, right?" And while that may help in the short term, we deal with enterprise
customers, a lack of content has never been their issue. Their marketing team is like, "Oh, we have that piece of collateral. Oh, yeah, we have that in a PDF over here. Oh yeah, that sits in our DAM. That sits in our SharePoint file. They have the content. It's not about creating more content. It's about optimizing the content that they have and serving it in the right way. Yeah. Give me just really quickly, what are some things that startups can take away from this, right? In terms of whether there's a new um opportunity for startups in search space or if there's just some one thing they should take
away with how they should be structuring their own websites. The first thing I would say is understand how you're showing up in AI search. It's really, really important. Like I said, on all of these startups websites, they may or may not know it, but there is probably two or three times more bot traffic on their sites than human traffic. So, first off, understand what your bot traffic is looking like, and then understand how your brand is showing up. Scrunch can help. You can go right to our website. We've got a oneweek free trial. Go test it out. I think you'll be surprised how many insights you can pull about whether you are or aren't showing up really quickly and easily. And then
once you want to take step two, start thinking about how your website and ultimately the codebase of your website can be optimized for agent consumption. Cool. Well, we are out of time. Um, so our listeners can go to Scrunch to connect with you. Anywhere else they can connect with you personally? I'm on LinkedIn. It's Matt Thompson. It's a very common first and last name. So, put in Matt Thompson and Scrunch. The with the three of those, you'll find me on LinkedIn.
I think we had a reporter that used to be called Matt Thompson at Techrunch. I had a college lecture freshman year where there were nine other Matt Thompsons in the lecture. Oh my god. So, yes, Thompson scratch in that search bar, people. Okay. Um, thanks again for joining us on the show. Um, our listeners, you can find me on X and LinkedIn as well. And you can find Equity at EquityPod on Blue Sky and X. Talk to you next time.
