How AI Models Decide Which Brands Get Visibility with Kevin White (ScrunchAI)

I do think a lot of people think of marketing as like, oh, we're going to generate [snorts] content. We're going to generate images for ads and stuff like that. But the other side of the equation is I think there's a Peter Dreer quote. If you can't measure something, then you can't improve it. And so the starting point actually of how do we improve our reach on AI platforms or how do we use AI is actually kind of like first let's benchmark how our brand is showing up here. Then it leads to what kind of content can we generate. Hi, my name is Demetri Bonichi and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI Agents podcast brought to you by Jot Form and featuring our very own CEO and founder Idkin Tank. This is the show where artificial intelligence meets

innovation, productivity, and the tools shaping the future of work. Enjoy the show. Hello and welcome back to another episode of the AI Agents podcast. In this episode, we have Kevin White, the head of marketing at Scrunch. How you doing today, Kevin? >> Hey, Demitri. Uh, good, dude. Doing really well. Uh, yeah, excited to get into some conversation here. >> Yeah, for sure. Excited to chat with you a little bit. So, just to kick things off, Kevin, obviously the landscape of search is uh, you know, it's kind of shifting beneath our beneath our feet uh, feet. Uh, the world of marketing in general is shifting beneath their feet with all the different things happening with AI. um you know you guys are kind of at the forefront of um seeing your brand through AI's eyes. And what I'd really like to know is you know what

was the specific moment or realization um in in the work you were doing that kind of led you to find um a company like Scrunch? >> Yeah. Uh so I it was kind of like doing the same the work myself and trying to you know be curious about where organic traffic and things like that are what's happening to our organic traffic at the company I was leading marketing at previously which is a company called common room and um yeah we saw that you know our traffic wasn't like disappearing but it was kind of leveling off on the organic side of things but then we were also getting um quite a lot of mentions of um we found your brand or we found your product through chat GBT or perplexity and starting to see some referral traffic coming from those sources. And that just kind of

like got me um exploring what what can we possibly do to influence these influence our results on these channels and um how can we grow um with this this new channel. This was uh to date things back in probably early January 2025. So about a year ago that we started seeing this um this these trends. And so that actually got me into, you know, uh what is this uh new user behavior? Um how are how are we showing up in these products, these LLMs like chat GBT or Plexity or Gemini? And um kind of got me uh on the like growth kick of like, oh, this is a new channel we should invest in. And then um you know, fast forward to uh to like later on in the year um saw the opportunity at Scrunch and what Scrunch was doing. um with not only

giving you visibility into um what's happening on these platforms but also the tools to take action and then also optimize your site for um you know accuracy, visibility, reach and so um yeah decided that uh I couldn't um fun story is that I I I feel like whenever I make a make a move to a new job I'm um you know if I if I if I hear about it and I can't stop thinking about it or like I lose sleep over thinking of ideas of how to market or grow the company, it's like a good sign. I really had that over the weekend after talking with Chris, our founder at Scrunch. And so, um, yeah, I decided to make the move after, um, you know, having that kind of like inside feeling of like I can't I can't ignore this. And so, that's how

I ended up, uh, heading leading up marketing at Scratch. >> You know, actually, it's interesting to me, you know, I work in marketing as well. And I just kind of want to know what what is it kind of like working not only at a company like Scrunch um just because of what you guys are doing and what's cool kind of in the um you know the way that you know CHBT Gemini and other models kind of deal with rank. I find that that fascinating in the first place. But secondarily, you know, what was it kind of like getting into the world of going from analyzing or doing a because you're in marketing and a lot of people are correlating, let me try to refine this question. Correlating marketing with AI in the output, right? Output of stuff you guys are, you know, obviously analyzing how

things are coming through from that output, right? in a sense from the search capacity. What is it like being on that side of things versus the generative side of things? >> Yeah. Um that Yeah, that's interesting. I I do think a lot of people think of generate content, we're going to generate um images for ads and stuff like that. But um the other side of the equation is I think you know if you if you can't measure something, I think there's a Peter Dreer quote, if you can't measure something, then you can't improve it. And so um the the starting point actually of uh of how do we how do we improve our reach on AI platforms or how do we use AI is actually kind of like um you know first let's benchmark how our brand is showing up here and then then it

leads to what kind of content can we generate um or how can we optimize our existing pages and things like that so that we do show up in this new growth channel. Um, I know that's like separate from a different like if you're creating an asset or something like that like you maybe don't need to use it as a distribution mechanism. It could be for um a webinar or like a inerson event or maybe you're creating like a one pager with AI. But um I think if you the way I was the way I'm looking at it is as more of like a growth channel of you know these platforms or or a new um a new platform where people are that are taking traffic from Google and Facebook and your traditional traditional um distribution platforms. So um so because of that uh you know

I um I think that that is like a let's see what's the right way to say this um I think that uh you know that is a a different way of looking at AI for marketing and that it's more around like you know can I use this as a as a uh way to get my brand awareness up versus like can I produce things for u my core audience. Yeah, fair. That's that's totally fair. And what are you guys kind of seeing is the you know Scrunch was founded in what 2023. >> Um yeah, we uh I mean we actually didn't make our product generally generally available until uh earlier this year till April this year. March this year. >> Okay. >> Um and so it was founded on a um as as a different company and then pivoted. And so um yeah, we've only

really been around for a year. uh and I was the first marketing hire um and that was you know 6 months after after that like general availability release. So um definitely moving fast and that you know my team is growing fast now. I'm hiring like crazy. Um that is definitely something that you know working at an AI um AI company uh there's definitely like it's fast moving the market's changing every uh day it seems like and you know it's definitely chaotic for in terms of like positioning messaging because it feels like you need to like update your website and all this stuff like every every other month. So it's it's fast-paced. >> Yeah, it's fair. It's uh it's very interesting to me kind of how uh quickly things uh keep uh moving not only model wise but companywise because of that but okay so since

March you've kind of been in the public realm what have you kind of noticed in you know this realm of uh I guess zero well actually first what were you guys then before this >> oh you know what I actually don't know I should know but uh I mean I wasn't like a founder or anything like that so >> yeah sure let's get to yeah no worries yeah we'll Uh we'll uh leave that for your conversation with the with the team later. Uh but for now, you guys are obviously kind of in this zero click search space to an extent, right? Um I I I would say that that's probably where you you would position yourselves as an analytics tool in regards to what's going on with how LLM see your search uh strategy. Is that correct? >> Yeah, I think that's the starting point,

but I I do um I would say that's not like your typical like semrush hrefs. is just like, you know, how is my brand or or like help me find like the keywords to um to prioritize, help me um uh you know, manage my ad spend and all that kind of stuff or give me analytics on search. Um I do think it's like a it's a different thing where you know where monitoring and analytics are the first part of it, but it's like a very commoditized thing I would say and you know the next step after that is you know how do you take action? How do you like change your actual website um experience? Um especially as like AI bots are crawling your site on behalf of someone prompting behind um an LLM. And so you almost need to like optimize your site and

experience for those AI chat bots that are crawling your site because they crawl it, they consume the right content, it's accurate, they'll represent a better picture and more accurate picture of um your your brand, your product, etc. uh to the user behind the prompt. And so that is like a different direction that's like out or it's in addition to monitoring and monitor is like the first part and then the second part is like you know how do we optimize our site so that the experience we we we kind of own more of the experience in the LM which we have zero control over when we have full control over our site for example. >> Yeah, speak to that a little bit more because I do think that's something that's uh often missed in a lot of these tools. Um I you know there's a lot

of really great uh things that you can try to do to optimize for something that exists that is outside of your control. And I I feel like this is a life lesson too, right? A lot of people focus on what's outside of their control versus inside of their control. And the LLMs to an extent are going to be outside of your control no matter what you do, right? You can optimize towards it, but you're saying you guys also help focus on the the site itself and what you can do to optimize like onpage experience and and whatnot. What does that look like? >> Yeah. Yeah. So we have a product and we call it uh agent experience platform and uh it will um it will uh identify uh bot traffic AI AI bot traffic um especially retrieval bots which is essentially like someone an AI

bot crawling your site after a prompt um it'll identify that and it will create um and fork a different user experience for that bot that is optimized for uh with markdown down and JSON and um consumability like low low light on tokens but high on dense on information. Uh and so that is like the lingua frana for uh these these a the AI traffic that's crawling your site whereas like the human experience is filled with JavaScript and images and things like that that is great if you're browsing and you want an interactive experience on on a site as a as a person. Um but it's uh it's not crawable. so so crawable for the AI bots right now. And so I think this is actually one um a premise or a thesis that we think is like different from others in the market now where

they're like, "Oh, we need to optimize our site for these AI bots, but then it could degrade the human experience if you just have like this long list of content that's dense and in markdown and like it's not like nice to look at or or skimable from the human experience." And so I think instead of trying to like force both together, um the ability to like split the the experience based on if it's bot traffic or a human web visitor, um actually is going to be the way that people um uh optimize their site in in the long run. Um and maybe even have like just completely like two different versions of the site. Maybe there's a CMS system that's like just for AI in the future. Maybe that's Crunch. Maybe that's another product. But like I I do think that um you know we

could see things go that way versus just having like one page. >> Okay. Yeah. No, that that makes a lot of sense. And it's interesting because I've been I I'm not actually too crazy, you know, sure how a lot of the LLMs do parse out what you're putting out. So could you actually let us know a little bit from your guys' perspective what and how that is? And you know, a little bit deeper into what you just said. like I I'm sure there's a lot of people who would probably benefit from a deeper explanation and how how you guys you know sort of reverse engineer analyzing that as well. >> Yeah. So um so there a few a few things there. One is just identifying that traffic in general and we have like an analytics product that will tell you you know what types of

what types of um LLM agents are are crawling your site. it's like a training agent, if it's a retrieval bot, if it's just like your normal search index crawler. um when you the way you identify that is um you can look at the traffic uh and there's unique identifiers for each of those different types of um that different type of bot traffic and then if you can identify at the um CDN layer or your CMS layer um which we have partnerships with like ai and burcell and um cloud cloudflare and all these sorts of CDNs um if you can if you can identify that traffic then you and one like say uh create analytics for it so that you know like what are the top models that are crawling our site what are the top pages that are being crawled um you know get a

real-time view of how much bot traffic which is kind of if you if you look at it most people are surprised by how much bot traffic you're getting um so that is just like insightful in itself and then um once you have once you are able to identify that traffic you can then serve different pages or different experience to uh that user in in the case of this user as a bot. Um, and the the the way to the the way right now at least to surface that content is typically through markdown. So you can take the the the human um optimized page maybe has JavaScript and video and all this kind of stuff on it that's heavy on um that's dense on like code and then convert it into markdown so that's much more crawable for these AI bots. And then um you can

also uh take the intent of the page and add more uh additional content to it like an FAQ section or something like that that you maybe like don't need a surface to the the human visitor. Um so it's just giving the the AI bots more context and more accuracy of like what's the intent of this page and um and so that's kind of uh the next step after identifying the traffic is you know providing the most accurate representation of what the intent of that page is. so that the bot can take that u return it back to the user and provide an accurate representation of what you're trying to express on that page. >> When you say bot traffic, what does that mean and how are you actually able to analyze that it is bot traffic? >> Yeah. Um I actually don't know like the

super technical details of it. I think when you um when you see referral traffic, there's like a um a referral like tag in the in the head of the uh I'm see I'm getting beyond my my like technical expertise here, but there's like a referral identifier from that traffic. um with if if you're getting like a chat GBTE uh retrieval bot that's visiting your site, you you're able to identify that that traffic um is uh is is from this bot from the the IP address and like the other um metadata that is being that is being identified um from from the visitor. So every every every person who visits a website, you know, you get an IP address that gives you a location, it gives you um all the sorts of stuff even if it's a human visitor. So like essentially we're doing that you

can do that but for bots and there's unique identifiers for the different types of bots that are out there. And so it's beyond my technical understanding of like what exactly like the the code and like the identifiers are there. We have a like guide that shows you how like kind of walks through like the technical side of it. But I think for marketers out there, you don't need to know those like super details. It's more like you can you can just identify it with tooling like uh your CMS or like a um uh CDN like Akami, Cloudflare, Fastly, all this kind of stuff. And all that work is just kind of like done for you. Um even even like Google Analytics, like Google Analytics 4 tracks this stuff for you automatically. So >> yeah, that's fair. So how do you uh primarily work with clients

then to help them on the uh different things that you're doing, right? Obviously, you know, as an AI product, it's interesting always to find the line between uh product and service, right? So how do you guys help uh companies specifically when you interact with them? What do you what is like the relationship look like? Yeah, I mean I would say we're more on the product side, but certainly provide a lot of guidance of what to do on with the once you uncover um or diagnose, you know, what your brand presence is like in these uh AI search um products. Um so yeah this we kind of think of it I like to think of it as like a maturity curve um where the first part is just like benchmarking uh monitoring side of things like you need to know what's happening and how your brand's

being represented on these platforms. Uh the next phase would be like okay we know what's happening we're watching trends like let's uncover lowhanging fruit for um you know how to take action. Maybe that's, you know, finding citations that are citing our competitors or um that we want to show up for as well or maybe it's creating new content or optimizing our existing content. Once you've identified that kind of like prioritization of um plan of action of what to do, then it's like actually taking action on it. So that could come in the form of creating content. It can come in the form of um optimizing or or uh changing like technical parts of your page to make it more crawable or indexable or whatever. Um but it all uh the other thing could be you know going this like route of you know creating creating

separate experience for for bots. So like there's three I would say there's three parts of the maturity curve. It's like monitoring getting insights and then taking action as the last step. So we we help with that whole end toend process but it always starts at the like beginning of benchmarking and then there's the services element of it which is like okay uh I have uh I have these insights like what should I do next? And that's where you kind of have like expertise come in and say like, "Oh, you should do this, this, and this." And we have like a services part of our organization that helps out there. Um, but for the most part, it's our customers who are like doing that like leg work of um, you know, we don't have access to like change their code on their their side or their

CMS or like making recommendations, they're taking action. >> And what would you pro what what would you say is the primary type of person or company that you see trying out your product? Yeah. Uh I mean it's kind of funny and also challenging as a marketer and that um this is a problem or something that like pretty much every company on the planet is looking to um understand and get get uh you know get their heads wrapped around. Um and so you know the that makes it challenging that we don't have like a single core like ICP. Um I mean I would say we have everything from fast growing startups to some of the world's like biggest enterprises um using our product. Um but uh the the the way like to to compartmentalize it would be um on the like self-s serve product side of

things. It's more um more like early stage startup type of companies or small businesses. We have also um uh a lot of agencies that are using our product to then do the content creation or the optimization for their clients and then we have like the bigger enterprise uh customers with like many brands and you know trying to do this all at scale. Um I would say like the common like persona would be um you know someone in growth or SEO is usually like the partic practitioner like hands-on using our products and then um there's like the board and CMO level of just like I think every marketer or head of marketing right now is getting questions from their board and C CEO like how are we showing up in these different products and so like that um we also provide like this kind of like

highle view for those like execs in sea level to get a to get an overview you like what's what's happening, but those are the typ typically not the ones doing the work. >> Yeah, fair enough. Okay. Um, what have been some of the cool results that you've seen out of uh you working with companies? >> Yeah. Um, I mean, uh, there's plenty of customer stories on our on our website. Um, I think uh, some of the like more fun wins are um, on the actual like revenue side of things. And so um the uh uh a lot of our customers are seeing a like flattening off or like a decline in organic traffic but then they are seeing that their brand presence is being shown and getting more share of voice in these LLM products but the more important thing is uh not only are

they getting a higher um uh share of voice but there's they're actually getting more traffic. that traffic is not exactly trackable because unless they click on unless the person visiting is clicking on a link directly from you know chat GBT orlexity which like doesn't happen all that much um they're probably like opening a new tab registering in their mind for later. This is kind of like the zero click uh space that you're or like zero click um trackability that you were kind of referring to earlier. Um and so they're seeing a correlation with share voice in these platforms and um direct traffic and then that traffic is converting at a much higher rate um and leading to a lot more revenue. So we've had um we had customers everywhere from like you know 3 to 9x um improvements in in traffic uh sorry in conversions

from traffic from um LLMs and then also um uh they mix that blend that with like direct traffic too so they kind of get like the full picture because it's usually direct traffic that's represented on the like zero click side of things. >> Yeah. Wow. Um I really that's a that's actually a pretty large percentage increases. Uh so I'm pretty Yeah, >> there's interesting there's an interesting other kind of like tidbit if you kind of like inspect the the user behavior here. It's like you're getting less website traffic because the the user experience is collapsed. It's like it's all happening on the the LLMs themselves. So instead of them visiting your website to do that research and then bouncing and then coming back, they're essentially the the LLMs are doing all of that research for them, making a recommendation. And so when they do hit

your website, um they're converting at a lot higher rate because they've done all the research up front and they're just like, I got it. I want this product. I'm just going to go to this website and transact. And so that's why you see like typically a much higher conversion rate from LM specifically, but then also higher in direct traffic as well. Hm. What do you uh love most about doing uh what you're doing and what like do you think feature- wise is the most exciting that you have? >> Uh yeah, feature- wise um uh we have this new product called AI search trends. And so what we do what we'll do is kind of give you a picture of what are the topics that people are talking about. Um what's the volume of traffic? What is your brand presence within those different topics? And I

think that is kind of like a thing that everyone's trying to like understand and try and wrap their head around um and so that they can prioritize the right way because you know the these um it's not like like SEO where you have like a pretty established um library of like keyword volumes and stuff like that. And so I think getting more to that so you can prioritize what's actually being talked about um is something that's on top of everyone's mind. I mean I would take that data with a grain of salt because it's not like perfect and the LMS don't publish it themselves but you know having a nice approximation directionally is uh something that I always look forward to um prioritize my you know where I'm going to focus my efforts for um you know getting for new channel and distribution um and

so uh yeah feature- wise I think I think that's a nice like call out um yeah and then you know I I love just working at early stage startups where there's you have direct impact on on the and like can ship stuff fast. And so that is definitely something we take pride in and do at at Scrunch. And so um yeah, that's what I think I like most about the the company, just like the ability to make an impact and like work and get stuff out the door fast. >> You know, I that makes a that makes sense. I I I'm actually interested following up on that kind of where do you think that tools like yourself will um help um [snorts] serve as sort of a lift for your small companies and uh because I I'm actually thinking that for this obviously you know

can work with any size company but where do you think you guys help serve companies of small and medium sizes to improve their ability to you know make more money off of their traffic and then not have to worry about like hiring out like multiple marketing people and and whatnot. What do you like? Because I think your your type of company, you stack a couple of those together, you can you can help really bridge the gap uh between needing to make that next hire. >> Yeah. I mean I think actually the entry points for uh for smaller businesses or like early stage businesses um to uh to show up to get your brand to show up in these in these channels on on um you know the different LLM platforms there's actually like a pretty low barrier to entry whereas in the past with SEO

the barrier to entry there was pretty high um just because everyone's competing for 10 blue links but the an interesting thing about um what we're seeing with um the way the way consumers and and people use these products is that they have much longer longtail prompts and they um the the LLMs themselves are really good at going and finding information from the like exact piece of information from a longtail prompt. um crawling the web and extracting and pulling it back to the user versus you know a keyword on Google you know you're going to get like the pages with the highest authority but not necessarily like answer get the like direct answer that you're looking for and so I like to call this like the long long tail of of search or prompt where like if you actually create content around those really longtail prompts

um in the past as a small business you probably wouldn't have gotten visibility on a search engine like Google. Uh, but now you're able to um the LLMs are able to sus out the people who have that intent and actually like surface your content to them and it and it doesn't and you can compete with like the larger companies who have like a really strong SEO um authority and footprint. And so um it's almost like if you build the content uh they kind of will come like like like Field of Dreams. Um but uh but you still need to have like you know you still need to to um you know think about what people are actually searching for and making sure your problem your product is actually solving for like a a core pain point. Um but if you create this longtail content there's

a higher chance that like you'll get seen you'll have less traffic but higher conversions kind of like we were talking about before. >> Yeah. Very cool. Um [snorts] what do you think is kind of the outlook of zero click search in the next uh couple years? Do you think there's going to be any major changes that uh shake the landscape up? Where do you think it's going to go? >> Yeah, there's like a big hubbub about like, you know, SEO is going away that's that um user experience is dead and like actually um interesting thing about you know new new changes in user behavior is oftentimes it's uh additive and so at least at first it's additive. So I don't think SEO or anything like that is going away. I think it's going to incrementally grow and then Google also like adapts you know get

like AI overview answers and stuff like that. Um so I think people will still use search but they will find new use cases for um and they already are finding many use cases for um the like AI chat experience and by by that I mean more like chatbt claude etc and Gemini I'd buck it in there and so um I think it's just like opening up more getting more attention from users and then also opening up more use cases especially on the generative side so I don't think it's a zero sum game. I think it's, you know, people are going to use both products more and more. I think the the chat experience will continue to accelerate a little bit more, but and then search will kind of slowly grow, but I don't think search is going to die off by any means. >> Yeah,

fair enough. I think that is a concern that's a little bit unwarranted at the moment. Um, a lot of lot of freak out there. Um, I I And where do you think if you had to say things are going to change? I I have some opinions on the way that agents continue to like scrape stuff, right? What do you think is going to be the percentage of traffic that comes from a lot of these like scraping agents? Uh >> I mean I think at at some point um many things there but like at some point uh the you'll be able to transact on these different platforms and so like you might get even less and less traffic but you'll still be able to >> oh they can purchase through a >> yeah you'll be able to purchase through purchase through you know whatever whatever um

product you're using. um they there's been some things like shopping experiences and and app um app marketplaces, but haven't like quite got to that like place where users are actually transacting. Um, and then on the content side of things, I actually think that the um these AI platforms are not going to reward AI generated content and they're going to actually reward and um maybe even pay you for like if you're Reddit or something like that, pay you for this um human content that is uh serving to like train their model. And so, um, I think, uh, I think that's something that, you know, there's a thing called model collapse where if you have a bunch of AI generated content and AI is training on its own content, it's going to like degrade the quality of that content. Yes. >> So, um, so I think that,

you know, the these platforms are trying to find ways to incentivize people to come up with their own original research so that they can train a model on it. >> Yeah. No, that's fair. Model collapse. Is that the word you used? >> Yes. >> Okay. I'm gonna start using that term because I It's funny enough. I I work in content. I hadn't heard the term yet, but I had been using the explained issue uh to try to explain to everyday people that I talk to about AI when they're like asking me about whatnot. And I was like, you the big issue, guys, is that you have the adventation of internet access and like last fall and there hasn't been like a counteraction to the quality assurance of it being true. one and and B there has been and B now we have a bunch of

outputed stuff from the lack of checked like true things that then are outputed by AI and then regurgitated. I think that report that came out that was like 50 or 60% of like sources in chat at GPT were Reddit at some point and ah that's bad for [laughter] >> yeah I think it just um if everything is reference self self referential it almost like lends itself gets to like mediocrity even if it is accurate data um so I can certainly see that happen and yeah I like to think of it as like the snake eating its own tail kind of thing um if if they rely if these models will run. But the people that work there are very smart like they know this is a thing and they're trying I my prediction is that they will um find a way to um continue to

have human generated research from experts like power their model and incentivize you know people to do that. >> Yeah. No, that's that's totally fair. I think they will essentially figure out what is the right way to allow for creative freedom to to really take root there because I agree with you. It's a probably a both end, right? Mediocrity is a thing, but also inaccuracy becomes self- reential, self- re referential at some point. And what's funny is that web search unlocked the possibility of referencing things that are outside of the LLM's data. But the LM's data originally is probably going to be is kind of some of the only accurate stuff unless you like go page by page and parse and check at a high level. >> Garbage in garbage out kind of thing. >> Yeah. [snorts] >> All right. Well, with that with that being

said, uh Kevin, tell us where everyone can find you guys at. >> Uh yeah, I'm uh you know, you can use Scrunch, find scrunchcrunch.com. Uh we got that uh you know, read uh dictionary domain, which is nice. Um and then uh >> uh you can find me on LinkedIn. Um I talk about this stuff a lot and post about it and show demos and all that kind of stuff. So yeah. >> Awesome. Well, Kevin, thank you so much for being with us today. Appreciate it. And we'll see you guys in the next one. Peace.