Episode 32: From Blue Links to LLMs in Local SEO

Key Takeaways

• LLMs fetch info based on web search, their own training data, and an orchestration layer
• Schema is how machines read and understand your site
• LLM responses vary by location, time of day, and device
• Evergreen content still matters for LLM training, but pages must show recent activity
• Internal linking still matters to the index, but the LLM itself does not care about it
• Brands need to be mentioned by others across the web

This year is shaping up to be even more unpredictable in search. As LLMs become the primary interface between brands and customers, the old rules of ranking in a static index no longer apply.

In this episode of the Local Marketing Beat podcast, host Christian Hustle sits down with James Hocking, Co-Founder of Hocking Digital and developer of Get Heard, a GEO analytics and tracking tool, to break down what SEOs are misunderstanding about LLMs, why schema matters more than ever, and how brands should think about discoverability in a world where blue links may disappear entirely.

‍

‍

Timestamps

00:00 Introduction and what’s changed in AI search

01:05 James’s background in tech, data, and neural networks

03:49 What SEOs misunderstand about LLM architecture

07:02 Why brand tracking alone isn’t enough

08:35 How location, time of day, and device affect LLM responses

11:41 Content freshness and the role of evergreen content

13:48 What happens if Google removes blue links

15:49 Internal linking, structure, and machine readability

17:16 Why being mentioned across the web is the new link building

19:06 Three actionable tips for brands

You’re Not Interacting with a Single Index Anymore

James Hocking: “You’re no longer interacting with a single index. You’re now interacting with multiple parts. The large language model doesn’t know anything about SEO. It doesn’t know anything about ranking. It doesn’t really care about the web. It’s a language model. Its goal in life is to interpret inputted text and give you a summary or a sequence that you ask for.”

James explains that when you interact with ChatGPT or Google’s AI, you’re not talking directly to a large language model. You’re interacting with a website that acts as a proxy, communicating with multiple layers: the LLM itself, a web search component, an orchestrator, and potentially their own emerging indexes. This is fundamentally different from the static, single-index world SEOs have operated in for 20 years.

This means that the old KPIs built around index ranking need to evolve. Brand tracking in AI search is a good starting point, but because LLM outputs are nondeterministic and change constantly, it’s not enough on its own.

‍

LLM Responses Change by Location, Time, and Device

James Hocking: “If you’re looking for consumer clothing in the US, the AI is unlikely to do any search on the internet to support its answer at 1 a.m. in the morning. However, if you search around 3 p.m. through to 6 or 7 p.m., it will. The bit that wraps around the LLM is clever enough to know your time zone.”

James shares data showing that the orchestration layer around LLMs adjusts its behavior based on geography, time zone, and user context. An AI search for clothing at 1 a.m. Eastern may not trigger a web search at all, while the same query at 3 p.m. will. The system also localizes results: A query from Arizona returns different results than one from New Jersey.

For multi-location brands, this means it’s not enough to track AI visibility at a country level. You need location-specific tracking that accounts for these variations. This aligns with the local-first approach in Location Performance Optimization (LPO), where visibility is measured and managed at each individual location.

Evergreen Content Still Matters, But Freshness Is Critical

James Hocking: “You could have the best evergreen content in the world, but if the page doesn’t reflect that you’ve assessed it within recent time, you’ll probably not get used by the AI at all, even though what you’ve written on your page is perfect.”

James addresses a common question: Should brands scrap their evergreen content and focus only on what’s current? The answer is no. LLMs are trained over years, so evergreen content builds the authority and knowledge that feeds into the model’s training data. But the web search layer that grounds AI responses in real time does check how recently a page was updated.

So, keep your evergreen content, but make sure pages show recent activity. Review and refresh regularly, even if the core information hasn’t changed. A page that hasn’t been touched in years will likely be skipped, regardless of how good the content is.

Structure and Schema over Persuasive Copy

James Hocking: “Make sure your site is as easily machine-readable as possible. These machines don’t care about persuasive copy. They just say: Can I use it to answer the question that I’ve been asked?”

When it comes to internal linking, James draws a clear distinction: The LLM itself does not care about internal links. But the index that feeds the LLM still does. So internal linking remains important for traditional SEO, but brands also need to think about how a machine reads their pages. Clean structure, proper schema markup, server-side rendered HTML (no JavaScript reliance), and content that directly answers questions are what the AI layer is looking for.

James also warns against relying on FAQ dumps. That worked briefly, but the real opportunity is structuring your brand’s message so AI can easily extract and use it in responses.

Being Mentioned Across the Web Is the New Link Building

James Hocking: “It’s about your brand or your message being repeated in lots of different places by lots of different people. That’s the equivalent of internal linking — that you are SEO everywhere, in a sense.”

James describes a fundamental shift: Where SEO used to be about internal links and backlinks, the AI equivalent is being mentioned and discussed across the web. The LLM doesn’t follow links the way a crawler does; it looks at whether your brand’s message is being repeated and validated by multiple sources.

For multi-location brands, this means investing in digital PR, active presence on platforms where your customers are (including Reddit and social media), and ensuring your listings and reviews consistently reflect your brand message. The more places AI encounters your brand saying the same thing, the more likely it is to surface you in its responses.

‍

Three Actionable Tips for Brands

Asked for one tip, James delivered three:

1. Find out what questions people are actually asking. Go to Reddit, Google’s “Also Asked,” and the AI tools themselves. These questions are a gold mine for understanding what content your brand needs to answer.

2. Make sure your site is clean, compliant, and has good markup. Content should be server-side rendered in HTML. AI systems won’t wait for JavaScript to load.

3. Make sure you’re ranking well in Bing. Many AI tools use Bing or Google to ground their responses. If you’re not on the first page of those results, AI agents won’t even find you. They don’t paginate.