For years, site architecture served its purpose and served it well: help search engines crawl, understand, and rank your pages. While that still matters, AI search has assigned your site architecture a second job that most sites aren’t yet built for.
When someone runs a search today, search engines and AI systems rarely stop at the words they typed. They expand that query into a network of related questions and pull an answer from whichever pages cover those questions best. This is known as query fan-out and it changes what a well-built site looks like.
Ranking first for your head term no longer guarantees you appear because the system is looking across a whole cluster of sub-questions, not one keyword.
That nuance makes visibility an architecture problem as much as a content one.
A site structured to cover the full intent around a topic has many ways to surface. A site built around isolated keyword pages has fewer. Our goal as SEOs is to figure out how to adapt.
Anchor your architecture around entities
Search engines and AI systems organize topics around entities: the people, places, products, and concepts a subject involves.
When a query fans out, those systems move through related entities to build it. If your content doesn’t clearly establish which entities it covers, it won’t be part of that expansion.
Start by naming entities explicitly rather than dancing around them and support them with schema markup where it adds clarity.
Schema helps machines understand what a page is about. It won’t guarantee a citation on its own, but it does remove ambiguity.
“Jaguar speed” could mean the animal or the car. Clear entity signals tell search engines which one your page answers.
Naming entities is only half the work. Each entity connects to several intents. Take “truffle oil,” for example. Someone might want to know what it is, how it compares to fresh truffles, where to buy it, or which brands are best.
Content that maps an entity to the intents around it has far more surface area in a fan-out world and that entity-to-intent map becomes the blueprint for everything below.
Build content clusters that mirror how topics get organized
The cleanest way to put that blueprint into your architecture is a hub-and-spoke model. A hub page covers a topic broadly and establishes your authority on it.
Cluster pages branch off to address specific entity-and-intent combinations in depth, each linking back to the hub.
To build it, map the intent universe around your topic, list the key entities in your space, and create a page for each meaningful entity-intent pairing.
Then audit for gaps. The sub-intents you haven’t covered are where competitors surface instead of you, and each one is a candidate for a new cluster page.
This, in theory, may sound like a lot of work. But with the right prompt, AI can now help you build a solid base from which to work, cutting the time from concept to implementation.
Use internal linking as semantic glue
Clusters only work if they’re connected. Internal links tie the pieces together and tell search engines how your pages relate.
Anchor text does much of that work. A descriptive phrase like “our guide to faceted navigation for e-commerce” signals the intent a page serves far better than “click here.”
Cross-linking related cluster pages, not just linking each spoke to its hub, reinforces the topical relationships that fan-out systems rely on.
Structure pages for passage-level extraction
AI systems rarely lift a whole document into an answer. They pull the passage that best resolves a specific sub-question. So each section of a page needs to stand on its own.
Frame headings as the questions your audience actually asks and answer them directly in the lines that follow.
A page built this way hands search engines and AI systems clean, self-contained passages to extract, which is how a single URL can show up across many sub-queries.
Also, keep in mind that visitors to your site also appreciate well-structured content that’s easy to skim and understand. We live in a world where skimming happens a lot more often than reading a full article.
Measure what you’ve built with Content Lens
Now we get to the slightly tricky part. You can design all of this deliberately on a handful of pages. Across thousands of pages, you can’t tell by hand whether your architecture actually covers the intent it’s supposed to. That gap is what Content Lens closes.
Content Lens uses AI to score each page across five dimensions and they line up almost one to one with the practices above:
- Query fan-out coverage tells you whether a page answers the follow-up questions around its topic, a direct read on whether your clusters cover the intent universe you mapped.
- Heading structure reflects how extractable your content is, the passage-level readiness you built for.
- Relevance and user experience, meta tags, and grammar and spelling round out the quality signals that decide whether a page earns its place.
You get one global score per page, prioritized recommendations for what to fix first, and it runs across your whole site rather than a sample.
From there, the AI Search Lens shows whether AI systems are actually crawling and citing the pages you’ve structured, so you can connect the architecture work back to real AI search visibility.
What does this mean for you?
Good architecture used to be about crawlability and a clean hierarchy. That foundation still holds, but the target has moved.
What decides visibility now is coverage: whether your site answers the whole question and everything a searcher or an AI system asks next.
The takeaway: anchor your structure around entities, cluster and link your content around intent, write pages that extract cleanly, then measure your coverage and act on the gaps.
Do that, and you’re built for a query fan-out world, not just the search of the past.
For the full framework, download our e-book, Mastering SEO in a query fan-out world.
