에피소드 122: How Brands Benchmark and Grow Visibility Inside AI Tools

I do think a lot of people think of marketing as just generating content and images for ads, but the other side of the equation is that if you can't measure something, then you can't improve it.

The starting point of how we improve our reach on AI platforms or use AI is to benchmark how our brand is showing up, which then leads to what kind of content we can generate.

Hi, my name is Demetri Panici and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI Agents podcast brought to you by Jform and featuring our CEO and founder Aytekin 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 are you doing today, Kevin?

Hey, Demetri. Doing really well. Yeah, excited to get into some conversation here.

Just to kick things off, Kevin, the landscape of search is shifting beneath our feet, and the world of marketing in general is shifting with all the different things happening with AI.

You guys are kind of at the forefront of seeing your brand through AI's eyes. What was the specific moment or realization in the work you were doing that led you to find a company like Scrunch?

I was doing the same work myself, trying to be curious about what's happening to our organic traffic at the company I was leading marketing at previously, Common Room.

We saw that traffic wasn't disappearing but leveling off on the organic side, while we were getting mentions of finding our brand or product through ChatGPT or Perplexity, and referral traffic from those sources.

That got me exploring what we could do to influence our results on these channels and how we could grow with this new channel, starting probably early January 2025.

It got me on the growth kick of seeing this as a new channel to invest in, and later in the year I saw the opportunity at Scrunch with what they were doing.

Scrunch gives visibility into what's happening on these platforms and provides tools to take action and optimize your site for accuracy, visibility, and reach.

I couldn't ignore the feeling I had after talking with Chris, our founder at Scrunch, and decided to make the move to lead marketing there.

It's interesting to me as someone who works in marketing as well. What is it like working at a company like Scrunch, especially given what you guys are doing with AI models like ChatGPT and Gemini?

What was it like going from analyzing marketing to correlating marketing with AI output and being on that side of things versus the generative side?

Many people think of marketing as generating content and images for ads, but the other side is that if you can't measure something, you can't improve it.

The starting point of how we improve our reach on AI platforms is benchmarking how our brand is showing up, which then leads to what content we can generate or optimize.

I see AI platforms more as a growth channel where people are taking traffic from Google and Facebook and traditional distribution platforms.

This is a different way of looking at AI for marketing, more around using it as a way to get brand awareness up versus just producing things for the core audience.

Scrunch was founded in 2023 and made its product generally available in March or April this year. I was the first marketing hire six months after that release.

Working at an AI company means things move fast, the market changes every day, and it's chaotic in terms of positioning and messaging because you need to update your website frequently.

Since March, you've been in the public realm. What have you noticed in this zero click search space? What were you guys before this?

You guys are in the zero click search space to an extent. Would you position yourselves as an analytics tool regarding how LLMs see your search strategy?

That's the starting point, but it's not like typical tools such as Semrush or Ahrefs. Monitoring and analytics are the first part, but the next step is taking action to change your website experience.

AI bots crawl your site on behalf of someone prompting behind an LLM, so you need to optimize your site and experience for those AI chat bots.

If they crawl and consume the right content accurately, they'll represent a better and more accurate picture of your brand and product to the user behind the prompt.

Monitoring is the first part, and the second part is optimizing your site so you own more of the experience in the LLM, which you have zero control over, unlike your site.

Many tools miss this aspect. You can optimize for something outside your control, but you also focus on the site itself and what you can do to optimize on-page experience.

We have a product called the Agent Experience Platform that identifies AI bot traffic, especially retrieval bots crawling your site after a prompt.

It creates a different user experience for that bot optimized with markdown and JSON for consumability, low on tokens but high on dense information.

The human experience is filled with JavaScript and images, great for browsing but not crawlable for AI bots right now.

Instead of forcing both experiences together, splitting the experience based on bot traffic or human visitors is the way people will optimize their sites in the long run.

Maybe there will be two different versions of the site or a CMS system just for AI in the future, possibly Scrunch or another product.

That makes a lot of sense. I've wondered how LLMs parse what you're putting out. Could you explain a bit more about how you reverse engineer and analyze that?

We identify traffic in general with an analytics product that tells you what types of LLM agents are crawling your site, such as training agents or retrieval bots.

You identify traffic by looking at unique identifiers for different types of bot traffic at the CDN or CMS layer through partnerships with providers like Akamai, Cloudflare, and others.

You create analytics to see top models crawling your site, top pages being crawled, and get a real-time view of bot traffic, which surprises many people by how much there is.

Once you identify that traffic, you serve different pages or experiences to that user as a bot, typically surfacing content through markdown.

Human-optimized pages have JavaScript and video, which are dense and not crawlable, so converting them into markdown makes them more crawlable for AI bots.

You can add additional content like FAQ sections that may not be surfaced to human visitors, giving AI bots more context and accuracy about the page's intent.

When you say bot traffic, what does that mean and how do you analyze that it is bot traffic?

I don't know the super technical details, but referral traffic has tags or identifiers in the head of the request that indicate if it's from a bot like ChatGPT retrieval bot.

You identify bot traffic from IP addresses and other metadata from the visitor. Every visitor has an IP address giving location and other data, and bots have unique identifiers.

We have guides explaining the technical side, but marketers don't need to know all details. Tools like CMS or CDNs handle this automatically, and Google Analytics 4 tracks this too.

How do you primarily work with clients to help them with what you're doing? How do you find the line between product and service?

We're more on the product side but provide guidance on what to do once you uncover or diagnose your brand presence in AI search products.

I think of it as a maturity curve: first benchmarking and monitoring to know what's happening and how your brand is represented.

Next is uncovering low-hanging fruit to take action, like finding citations citing competitors or creating and optimizing content.

Then comes taking action, which could be creating content, optimizing technical parts of your page to be more crawlable, or creating separate experiences for bots.

We help with the whole end-to-end process, starting with benchmarking. There's also a services element where we provide expertise on what to do next.

Most customers do the leg work of changing their code or CMS based on our recommendations.

What is the primary type of person or company trying your product?

It's challenging because this problem affects almost every company. We don't have a single core ICP but have customers from fast-growing startups to large enterprises.

On the self-serve product side, it's more early-stage startups or small businesses. Agencies use our product for content creation or optimization for clients.

The common persona is someone in growth or SEO hands-on using our products, with execs and CMOs getting high-level views for board questions.

What are some cool results you've seen working with companies?

Many customers see flattening or declining organic traffic but increased brand presence and share of voice in LLM products.

More importantly, they get more traffic that's not exactly trackable because users often don't click links but register the brand mentally.

This zero-click traffic correlates with share of voice and direct traffic, converting at much higher rates and leading to more revenue.

Some customers have seen 3 to 9 times improvements in conversions from LLM traffic, blended with direct traffic for a full picture.

User behavior shows less website traffic because the experience is collapsed in LLMs, which do the research and make recommendations.

When users visit your site, they convert at higher rates because they've done research upfront and are ready to transact.

What do you love most about what you're doing and what feature is most exciting?

We have a new product called AI Search Trends that shows topics people are talking about, traffic volume, and your brand presence within those topics.

Everyone is trying to understand and prioritize the right way because unlike SEO, there's no established keyword volume library for these new channels.

I take the data with a grain of salt as it's not perfect, but having a directional approximation helps prioritize efforts for new channels and distribution.

I love working at early-stage startups where you have direct impact and can ship stuff fast, which we pride ourselves on at Scrunch.

Where do you think tools like yours will help small and medium companies improve their ability to make more money from traffic without hiring many marketing people?

There's a low barrier to entry for smaller businesses to show up on LLM platforms compared to traditional SEO, which had a high barrier due to competition for blue links.

LLMs handle long-tail prompts well, finding exact information from long-tail queries and surfacing content to users, allowing smaller companies to compete with larger ones.

If you build the content, they will come, but you still need to think about what people are searching for and ensure your product solves a core pain point.

Long-tail content may get less traffic but higher conversions, as we discussed earlier.

What is the outlook for zero-click search in the next few years? Will there be major changes shaking up the landscape?

There's talk that SEO is going away and user experience is dead, but often changes are additive. SEO and search will incrementally grow while AI chat experiences accelerate.

People will use both search and chat products more. Chat experiences will accelerate, but search won't die off.

Where do you think things will change, especially regarding scraping agents and the percentage of traffic from them?

Eventually, users may transact on these platforms directly, reducing traffic but enabling purchases through the AI interfaces.

AI platforms won't reward AI-generated content but will reward human content that trains their models, like Reddit content.

Model collapse happens when AI trains on its own content, degrading quality. Platforms are trying to incentivize original human research to power models.

I like the term model collapse. It's a real issue where self-referential content leads to mediocrity and accuracy problems.

Web search unlocked referencing outside LLM data, but LLM data is probably the only accurate source unless manually checked.

Kevin, where can everyone find you guys?

You can find Scrunch at scrunchcrunch.com and me on LinkedIn where I post about this stuff and show demos.

Thanks so much for being with us today. Appreciate it. We'll see you guys in the next one. Peace.