Fan-Out Analysis Is the New Keyword Research That Most Teams Haven’t Started Yet

In this week’s episode of Voices of Search, we spoke with Karl Kleinschmidt, founder at Data Marketing Group and an 18-year SEO veteran focused on local SEO, large-scale data systems, and LLM optimization. Karl has been building custom tools and data workflows for years, and his approach to AI discovery is more technically grounded than most of what’s circulating in the industry right now.

The conversation covered three interconnected ideas: why intent matters more than keywords in LLM optimization, how fan-out analysis gives teams a structured way to map what users actually want, and what’s really happening in local search as AI models struggle to replicate what Google has built over two decades. Throughout, Karl was direct about what’s still unsolved and where teams are wasting time.

Key Takeaways From This Episode:

• LLM optimization is fundamentally about intent, not keywords. Stop thinking about what people search for and start thinking about what they actually want.
• Fan out analysis, the practice of mapping sub-intents within a primary query, is the most valuable research exercise teams can do right now, but there’s no standardized definition or perfect tool for it yet.
• In local search, LLMs are heavily dependent on business directories like Yelp. If your business isn’t ranking in the top three to five on those platforms, you’re likely invisible in AI responses for many verticals.
• Don’t replace your existing SEO tools yet. Give LLMs maximum context about your business, ask strategic questions, and see what they surface before investing in new infrastructure.
• Build your own reporting. Off-the-shelf dashboards can’t capture what’s unique to your business. Foundational data is worth buying; reporting is worth building.
• LLM manipulation, repeatedly prompting models to argue they should recommend you, produces short-term results at best and no durable value.

From Keywords to Intent

The shift Karl is describing isn’t subtle. Traditional SEO was built around matching content to what people search for. LLM optimization requires understanding what they actually want, the need behind the query, not the query itself.

“It’s no longer the right approach to say, okay, they’re going to search ‘I want a red running shoe,'” Karl said. “They want a shoe that helps them while they’re running, that looks stylish, their favorite color is red. You need to figure out what the intent is of what people are actually searching for. And then you can figure out how to best write the content to fit that intent.”

The practical implication is a more deliberate research process. Karl’s starting point with any client is understanding what LLMs currently think about them at different levels of specificity. 

He used Allbirds as an example: they’re not known as a running shoe brand, they’re known for lifestyle shoes and waterproof footwear. That LLM perception is the baseline. From there, the strategic question is whether to move up a level (can we become the best running shoe brand?) or go wider by dominating the waterproof footwear niche and building authority around adjacent use cases.

What Fan-Out Analysis Actually Is

“Fan-out” refers to the sub-intents that branch out from a primary query when an LLM processes it. When someone searches for something in an AI model, the model doesn’t just match the surface phrase. It generates multiple related queries to build a more complete answer. 

Understanding what those sub-queries are, and whether your brand is relevant to them, is a fan-out analysis.

Karl is direct about where the field stands: there’s no perfect tool or standardized approach yet. “Every person does it a little differently. Every LLM does it a little differently.” But the exercise is valuable regardless, because it surfaces assumptions about where you fit that may or may not match how LLMs actually understand your brand.

His process starts with a primary topic keyword, generates around 10 fan-out queries, and then runs four questions against each one:

• Why are they searching this?
• Who are they?
• What triggered the search?
• What type of content are they looking for?

The answers shape both the content strategy and the diagnosis. If you’re highly relevant to 60% of the fan-outs and irrelevant to 40%, that’s useful information. If you’re irrelevant to 70%, that’s a signal the LLM’s understanding of your brand’s intent territory is fundamentally misaligned from your own, and something upstream needs to be addressed.

Connecting Fan Out to GSC Data

Once the fan-out analysis is done, Karl overlays Google Search Console data to ground the intent map in real performance signals.

“I pull in all the GSC data and connect each fan-out with semantically related GSC queries,” he explained. “Then I have average ranking, impressions, and clicks for every single fan-out. And I can see here’s a complete blind spot for me—zero clicks, zero impressions for this thing. Let’s look more into it. Am I just not the right person to talk about it? Do I have no content for it?”

The output gives teams a prioritized list of where they’re invisible, where they’re weak, and what type of content intervention is actually needed: a new post, an FAQ, a page refresh, or additional citations. It’s a workflow that bridges the intent research with actionable execution rather than leaving it as an abstract exercise.

What’s Really Happening in Local

Karl has done deep research across several local verticals, and his findings are both specific and sobering. 

Two distinct problems are making local AI discovery difficult:

• The location problem. LLMs don’t know where you are the way Google Maps does, especially on desktop. Google has spent two decades building location signals into its core product. LLMs are starting from a much less rich base. “If you give the LLM enough context about where you are, they’re pretty good. But no one searches like that.” The behavioral pattern Google trained into users—just say “near me”—doesn’t translate cleanly into how LLMs interpret queries.
• The content problem. Local websites tend to be among the least structured and least content-rich sites on the internet, which means LLMs fall back heavily on business directories. “Some of the verticals I’ve looked at, Yelp is just the source of truth,” Karl said. “If you’re not in the top three or top five on Yelp, or if you’re not paying Yelp, you’re not showing up in AI overview or in the LLMs.”

That dependency on third-party directories has real strategic implications. Optimizing your own site matters less if the aggregators are the authoritative source for your category. Getting and maintaining top placement in the directories that LLMs cite for your vertical may be the highest-leverage local SEO activity available right now.

The longer-term picture is that local SEO and traditional organic SEO are converging in ways that will introduce complexity the local SEO community may not be fully prepared for. Fan-out analysis at the local level has to account for geographic intent on top of everything else. How far are people willing to travel for this service, and does that vary by cluster?

How to Actually Use LLMs in Your Workflow

The most practical section of the conversation was Karl’s framework for using LLMs as tools without overhauling your entire stack.

His recommendation for most teams right now: don’t replace anything. Don’t build new tools yet. Just start giving LLMs as much context about your business as possible and asking them questions.

“You’re not replacing any tools. You’re not building tools. You’re just saying: I have all of this data, I am trying to do this, any recommendations?” Karl said. “You’re going to find that they have recommendations for you that will be helpful, especially as you give it more and more information.”

For clients with whom he goes deeper, Karl has developed a specific setup:

• Create a Claude project for each client and ingest all available business information in a standardized format
• Treat the project as a persistent context layer for all subsequent analysis
• Route different tasks to different models based on their strengths

On that last point, Karl has found that Claude and Gemini are better at different things. Claude handles large volumes of data and decision-making well. Gemini is better at distilling complex concepts down to a single word or phrase. When he hits a wall with Claude on a naming or framing problem, he asks Gemini, feeds that answer back to Claude, and the result improves significantly.

On the question of buying versus building: foundational data is still worth buying from established tools. But reporting, Karl argues, is something every team should be building themselves. “Reporting is so unique to you—you should be building your own because you know way more about your business needs than any SaaS company out there.”

The One Thing Worth Prioritizing

When Jordan asked for a single focus for AI visibility, Karl’s answer wasn’t a tool or a tactic. It was content briefs.

“The amount of inputs that go into a content brief are expanding weekly,” he said. They used to be two pages. Now they’re approaching 11, incorporating fan-out analysis, intent mapping, GSC data, LLM response data, and more. The SEOs who build those briefs well, giving writers everything they need in the clearest possible format, are the ones whose content will consistently outperform automated alternatives.

“People who write really good, unique content with unique data, unique perspectives, and unique expertise are going to beat you in the long run, no matter how good a prompter you are. And there’s always a better prompter out there.”

Fan-out analysis, intent mapping, local directory strategy, and custom tooling all exist in service of the same goal: giving the people writing your content the context they need to produce something genuinely useful. 

That’s what gets cited. That’s what builds durable visibility.

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.