Why AI Search Behavior Is Still Messier Than the Industry Wants to Admit

In this week’s episode of Voices of Search, we spoke with Garrett Sussman, Director of Marketing at iPullRank, about what AI search behavior actually looks like right now, why the hype doesn’t match reality, and how marketers can separate genuine opportunity from industry noise.

iPullRank is an agency at the forefront of rethinking how organic search works in the age of AI, building frameworks and research to help enterprise brands navigate a channel that is fundamentally changing beneath their feet. Garrett came to marketing through an unconventional path—teaching third graders in New Orleans, surviving Hurricane Katrina, getting into live streaming before Twitch existed, and eventually landing in San Francisco’s startup world before finding his way into SEO through local search and reputation management. 

He joined iPullRank in 2020 and has been fully immersed in the AI search transition ever since.

Today, Garrett broke down why conversational search is breaking traditional keyword measurement, how the relevance engineering framework helps marketers build a strategic roadmap, and why accepting that AI platforms are inherently unfair is the first step toward working with them effectively.

Key Takeaways From This Episode:

• 53% of users do not return to Google AI Mode after their initial attempt, meaning AI search behavior is still very much in flux
• Keyword volume as a measurement concept is eroding as natural language queries become increasingly personalized and one-time
• Measurement should be structured in three tiers: input metrics, channel metrics like share of voice and mentions, and performance metrics like traffic and revenue
• Relevance engineering is a five-pillar framework covering AI information retrieval, content strategy, digital PR, user experience, and channel presence
• Google still processes 16 billion searches per day versus ChatGPT’s 1 billion prompts, and a large portion of those prompts are work-related rather than discovery-focused
• AI search platforms are not fair and never will be, and marketers need to make deliberate choices about how they respond to that reality
• Understanding how AI fans out a single query into 20 to 30 synthetic related searches is key to building content that earns visibility

The Hype Is Loud, But the Behavior Is Still Evolving

Despite the loud hype surrounding AI search, the data tells a different story: according to clickstream research from iPullRank, 53% of users do not return to Google AI Mode after their first attempt. This is the sign of a channel still trying to find its footing rather than a sign of one that’s already taken over.

Garrett was clear that this doesn’t mean AI search doesn’t matter. It means the transition is messier and slower than the loudest voices in the industry suggest, and that marketers who treat it as already settled are making decisions based on a fiction.

The underlying shift, though, is real. Google activated AI Overviews in a way that users can’t meaningfully avoid, which means the search experience is changing whether users have fully adopted it or not. As Garrett put it, marketers are now responsible for a “probabilistic answer engine that can talk about you in a certain way based on everything that’s happened with your brand in the history of your brand.”

That’s a fundamentally different responsibility than optimizing a page to rank for a keyword.

Why Traditional Measurement Is Breaking Down

One of the most practically important parts of the conversation was Garrett’s explanation of why keyword volume—the metric SEO teams have relied on for decades—is rapidly losing its usefulness as a measurement tool.

The reason is simple: natural language search doesn’t produce repeatable queries. 

A 30-word conversational prompt entered into an AI search platform is likely to be used exactly once, by exactly one person, in exactly one context. There’s no search volume to measure. There are no keywords to optimize for.

Garrett’s recommendation is to restructure measurement into three distinct tiers:

• Input metrics: What you’re creating, how your content is structured, how bots are interacting with your site
• Channel metrics: Share of voice across AI platforms, mentions, and whether you’re being recommended or cited
• Performance metrics: Traffic, leads, conversions, and revenue, which the C-suite still expects to see

The challenge is weaving all three into a coherent story. Attribution in search has always been complicated, but as Garrett noted, “now it’s just so fractured.” The goal isn’t to find a perfect measurement system—it’s to build a framework that leadership can understand and that gives teams something actionable to optimize toward.

The Five Pillars of Relevance Engineering

The framework iPullRank has built to navigate this environment is called relevance engineering, developed by Mike King. 

It organizes the work of AI-era organic marketing into five core pillars:

1. AI information retrieval: Ensuring your content is discoverable by the AI search bots that generate outputs
2. Content strategy: Both on your own site and across every rented channel where your brand can appear, from YouTube to Reddit to email
3. Digital PR: Participating in channels you don’t own but can influence, so your brand gets pulled into the selection opportunity
4. User experience: Ensuring that when people do land on your site, they find what they’re looking for, because satisfaction signals still matter
5. Channel presence: Understanding which systems your audience is using and optimizing your visibility within those specific ecosystems

The framework is designed to replace the old checklist mentality of SEO with something more holistic—a strategic roadmap that accounts for where brand reputation, content, and discovery are all converging.

Google vs. ChatGPT: Where Should Marketers Actually Focus?

A recurring tension in the conversation was how to prioritize between Google’s massive scale and the emerging LLM platforms that get so much attention. The numbers say that Google processes 16 billion searches per day, while ChatGPT handles around 1 billion prompts—and a significant portion of those prompts are work-related tasks, not consumer discovery.

Garrett’s view is that this doesn’t mean ignoring LLM platforms, but it does mean keeping perspective. AI Overviews alone justify attention to AI-era optimization because they’re already embedded in Google’s results at scale. The LLM-native platforms matter too, but for most brands, they represent a future investment rather than an immediate traffic driver.

The more important strategic question is understanding that as AI assistants become more integrated into devices and operating systems, users will increasingly become dependent on whichever ecosystem they’ve chosen—Apple, Google, Microsoft, or others. 

Marketers will eventually need to think about optimizing visibility within specific LLM ecosystems, not just across search broadly.

AI Platforms Are Not Fair, and That’s the Starting Point

At one point, the topic shifted to the editorial nature of AI platforms. Garrett’s position was unambiguous: these platforms are not fair, they are not going to be fair, and pretending otherwise leads to bad strategy.

This isn’t a new problem, either. Google’s results have always had bias baked in—favoring large brands, rewarding content that mirrors what already ranks, creating virtual cycles where winners keep winning. AI search inherits and in some ways amplifies those dynamics, because the training data itself reflects existing power structures on the internet.

For marketers, Garrett said this means making deliberate choices about time horizons and tactics:

• Some brands will pursue short-term visibility through aggressive content generation and manipulation of whatever sources currently appear in AI citations
• Others will invest in long-term brand authority, knowing that the platforms will continue to evolve and that sustainable visibility comes from genuine authority
• The right answer depends on business goals, brand values, and an honest assessment of what’s actually achievable

“You have to sit down with yourself,” Garrett said, “and know what your north star is.”

Understanding Query Fan-Out Is the Practical Takeaway

Beyond the strategic frameworks, Garrett offered one of the most concrete tactical insights of the episode: the concept of query fan-out.

When a user enters a prompt into AI Mode or ChatGPT, the model generates 20 to 30 synthetic related searches to help ground its response. Those related searches are personalized to the individual user—which means no two people will get the same set. 

The practical question for content teams is whether their site has pages that satisfy all of those synthetic related searches, giving the brand as many chances as possible to appear in both the generative output and the citation layer.

This reframes content strategy entirely. Rather than targeting a keyword and its close variants, the goal becomes mapping out the full range of questions a specific audience segment might generate around a topic, then building content that satisfies that entire space.

The Channel Is Changing Faster Than the Budgets Are

One of the broader points Garrett made—and one that resonates well beyond any single tactic—is that organic search is functionally becoming an awareness and brand channel, but it hasn’t yet received the budget treatment that awareness channels typically get.

Brand and paid have always commanded large budgets precisely because they drive awareness at scale. Organic search has historically been underfunded relative to its actual contribution to customer journeys. As AI search blurs the line between brand reputation and organic visibility, that budget conversation is going to have to change.

“We can’t just keep calling it SEO,” Garrett said, “if you’re going to shift the value the channel brings.”

The teams that make that case now—before the shift is fully complete—will be in a much stronger position to resource the work properly and demonstrate impact at the scale the channel actually warrants.

Separate the Signal From the Noise Before the Market Does It for You

AI search is real, it is changing consumer behavior, and it requires a strategic response. But the response that works isn’t chasing every new platform, tracking every mention, or flooding the internet with content optimized for whatever source currently appears most in AI citations.

The response that works is building genuine brand authority across the channels that matter, structuring content to satisfy the full range of queries an audience might generate, measuring in tiers that connect inputs to outcomes, and accepting that these platforms will never be perfectly fair—and competing anyway.

Voices of Search is a daily SEO and content marketing podcast hosted by Jordan Keone and Tyson Stockton. The show delivers actionable strategies and data-driven insights to help marketers navigate the ever-evolving world of search engine optimization and content marketing. New episodes air weekly, covering everything from technical SEO to AI discovery, featuring industry leaders and practitioners sharing real-world frameworks and proven tactics.

Subscribe to Voices of Search on Apple Podcasts, Spotify, or your favorite podcast platform. Follow Previsible on LinkedIn for updates and subscribe to the VOS YouTube channel for video episodes and clips. You can also visit the official VOS site to explore the full episode archive and submit your SEO questions for future episodes.