Site Architecture

TL;DR — Site architecture is the crawlable network of links between your pages — not just your URL folders. The goal is simple: every important page should have a link pointing to it, and related pages should link to each other. You’ll hear three names for organizing that network — silos, hub and spoke, and topic clusters — they’re overlapping practitioner labels for roughly the same idea, not three separate systems. None of this guarantees rankings or traffic by itself — it’s what makes pages findable and understandable. Google likes a pyramid: home page at the top, broad category pages in the middle, specific articles at the bottom.

What site architecture is

Site architecture is the network of crawlable links that connects your pages — the paths search engines (and readers) use to discover and move between them — plus supporting entry points like an XML sitemap. It’s not the same thing as your URL folders: a page’s path (/blog/category/post/) doesn’t by itself establish where that page sits in the architecture. It’s the difference between a tidy library where every book has a shelf and a sign pointing to it, and a pile of books on the floor.

It matters for search for two reasons:

1. Discovery. Search engines find pages mainly by following standard <a href> links from pages they already know about. If nothing links to a page, it’s hard for Google to find it at all.
2. Context. When a page links to another with relevant words around the link, it helps Google understand what the linked page is about.

Discovery isn’t the finish line, though — a page being linked and crawlable doesn’t guarantee it gets indexed or ranks. Architecture gets you found and understood; it doesn’t promise an outcome past that.

The three models you’ll hear about

• Silo structure — you split your content into separate topic “silos” and keep each silo’s pages linking mostly to each other. The old, strict version says never link between silos.
• Hub and spoke — one broad overview page (the hub) links out to a set of detailed pages (the spokes), and each spoke links back to the hub.
• Topic clusters — a “pillar” page covers a big topic, and “cluster” pages cover the smaller subtopics, all interlinked.

Here’s the part most guides bury: hub and spoke and topic clusters describe the same underlying structure. One name came from library/information-architecture people; the other was popularized by HubSpot in 2017 as a content-marketing idea. None of these three are Google-documented categories — they’re practitioner labels with overlapping implementations, not an official taxonomy. I’ll explain that more in the Advanced version.

What actually matters

You don’t need to pick a brand name. You need:

• A clear hierarchy: home page → category/hub pages → specific pages.
• Every important page linked from at least one other page (no “orphans”) — Google’s own documentation puts it plainly: every page you care about should have a link from at least one other page on your site.
• Real <a href> links, not click-only navigation Google can’t reliably follow.
• Related pages linking to each other with descriptive link text.
• Nothing important buried so deep that crawlers struggle to reach it.

None of this is a guarantee of crawling, indexing, or ranking — it’s what makes a page reachable and legible in the first place.

The thing most people get wrong

You don’t have to avoid linking between sections. No Google source requires banning links between topic areas. The strict silo rule — “never link between silos” — trades away links that would genuinely help readers navigate and help Google understand the destination page, for the sake of a rule that isn’t documented anywhere. Judge cross-links by whether they’re relevant and useful to a reader, not by whether they cross a silo boundary. The valuable idea from silos is grouping related content together, not walling it off.

Want the deeper version — how link equity flows, flat vs. deep hierarchies, and when each model fits — switch to the Advanced tab.

TL;DR — Silo, hub and spoke, and topic clusters are overlapping practitioner labels, not documented Google architecture categories — but they converge on one structure Mueller has described favorably: a pyramid / top-down hierarchy. Hub and spoke and topic clusters share the same underlying pattern more than they genuinely differ — one comes from information architecture, the other from HubSpot’s 2017 content-marketing rebrand. The part of silo thinking worth keeping is topical concentration of internal links; the part to drop is the strict “no cross-silo links” rule — no Google source requires it, so judge cross-links by relevance and user navigation instead. What Google leans on is internal linking and context, not URL folder structure. Keep hierarchies reasonably shallow, link related pages across clusters when it’s relevant, and remember architecture enables outcomes like topical authority and AI-search visibility — it doesn’t guarantee them, and it can’t rescue thin content.

Why SEOs argue about this at all

Site architecture is one of those topics where three communities invented overlapping vocabularies and then spent fifteen years insisting their word was the real one. I’ve spent six-plus years at Ahrefs on the product side around Site Audit, looking at what site structures actually look like in crawl data across a huge number of sites — and the gap between the dogma and what works is wide. So let me try to collapse the confusion.

There are three named models:

• Silo structure — content grouped into isolated thematic sections. The strict interpretation (popularized by Bruce Clay) prohibits cross-silo internal links to keep each silo’s “link equity” concentrated.
• Hub and spoke — a broad overview hub page links out to detailed spoke pages, which link back. The hub is sometimes called a pillar page.
• Topic clusters — HubSpot’s 2017 framing: a pillar page targets a broad keyword, cluster pages target long-tail subtopics, everything interlinked.

The clarification I’ll plant my flag on

None of “silo,” “hub and spoke,” or “topic cluster” is a Google-documented architecture category. They’re practitioner labels, and in practice their implementations overlap heavily. “Hub and spoke” is the older information-architecture term: a central overview page linking to detailed subtopic pages with reciprocal links. “Topic clusters” is HubSpot’s 2017 content-marketing rebrand of essentially that same pattern — a pillar page plus interlinked cluster content. “Pillar page” is just the marketing word for the hub. Calling them flatly identical overstates it — HubSpot’s framing leans more on planned topical coverage as a content strategy, while “hub and spoke” is older, more general information-architecture vocabulary — but if you’re choosing how to structure links, you’re solving the same problem either way: a central page, detailed subtopic pages, and reciprocal links between them.

And once you allow cross-silo links — which nothing Google has published tells you not to — a “silo” is a hub-and-spoke cluster in practice. So the three names mostly converge on one workable structure: grouped, hierarchical, internally linked, with sensible cross-links. Treat them as overlapping vocabulary for that structure, not as three competing systems to choose between.

What Google actually recommends

Google’s recommendation is a pyramid / top-down hierarchy: the home page covers the broadest topic, category/hub pages sit in the middle, and specific content pages live at the bottom. John Mueller put the why plainly: the top-down approach or pyramid structure “helps us a lot more to understand the context of individual pages within the site.” That’s the payoff — structure isn’t a ranking trick, it’s how Google figures out what each page is about and how pages relate.

Worth noting: Google uses the term “hub page” in its own documentation — describing how “a hub page, such as a category page, links to a new blog post” for discovery. So the hub-and-spoke framing isn’t an SEO invention; it’s Google’s own language.

Internal links beat URL structure

This is the single most misunderstood part of architecture. A lot of silo advice is really about URL folders — putting /category-a/ pages under one path and forbidding links to /category-b/. But Google focuses on internal linking signals, not URL path segments. Mueller has repeatedly noted that some SEOs over-focus on URL structure; Google works out hierarchy from how pages link to each other, not from folder names.

The implication is big: a page at /blog/technical-seo/site-architecture/ signals nothing to Google beyond what its content and inbound links say about it. Logical URLs are good for humans and for managing the site — but they aren’t the architecture. The link graph is the architecture. This is also why “virtual silos” (concentrating links by topic regardless of folder) work fine, and why strict “physical silos” (folder-based isolation with no cross-links) buy you crawl and UX problems for no offsetting benefit.

Google’s documentation is specific about what counts as a link it can reliably follow: a standard <a href> element with a resolvable URL. Navigation built only from JavaScript click handlers or non-standard markup doesn’t get you the same reliable discovery path, no matter how clean your URL structure looks. Get the crawlable-link contract right first — folder naming is secondary.

Where strict silos go wrong

The legitimate insight in silo thinking is topical concentration — grouping related content and linking generously within a topic area genuinely helps readers navigate and helps Google understand what a page is about. The failure mode is the strict rule: never link between silos. No reviewed Google source requires that prohibition — so judge it on its own terms, not as an assumed authority mechanic:

• it prevents natural, contextually relevant links readers would benefit from,
• it hurts UX by dead-ending users at artificial silo boundaries, and
• a page can still be relevant to more than one topic area, and a link that helps a reader find it shouldn’t be blocked because it crosses a label you drew.

Shari Thurow has been making this point for years (“stop the silo madness”), and Ahrefs’ own contrarian take (Joshua Hardwick’s “why it makes no sense”) lands in the same place. Keep the concentration; drop the wall — and make the call based on whether a link genuinely serves the reader and the destination page, not on unverified claims about how strictly Google weighs isolation.

Flat vs. deep hierarchies

Two failure modes at the extremes:

• Too flat — everything one click off the home page. Link equity and topical signal get diluted; the home page can’t meaningfully vouch for hundreds of equal children, and you lose the topical grouping that helps context.
• Too deep — important pages many clicks from home. Mueller’s framing: going too deep “makes it harder for us to crawl and harder for us to pass the signals around.” Deep pages tend to get crawled less and inherit less internal authority. Google hasn’t published a specific click-depth number; don’t treat any fixed count as a requirement — the right depth depends on your site’s size and how distinct its categories are, which is its own topic (see the flat vs. deep and crawl-depth deep dives in this cluster for the fuller decision framework).

The target is a shallow hierarchy with strong contextual linking: keep important pages reachable in as few hops as your site’s scale reasonably allows, grouped into hubs, with cross-links where topics genuinely relate. This is the same crawl-depth concern that shows up whenever you’re auditing how far bots have to travel to reach your money pages.

What internal linking actually does

Architecture’s real job is shaping the internal link graph. Three things to get right:

1. Linkability. Every important page should have a link from at least one other page — ideally several. Google’s own guidance is explicit about this minimum. Pages with no inlinks are harder to find and harder to rank, but a link doesn’t guarantee either outcome.
2. Context. The words before and after a link, and the anchor text itself, can help people and Google understand what the target is about. This is how a hub “explains” its spokes — it’s not a promise of a specific ranking effect.
3. Concentration. Linking densely within a topic cluster puts relevance signals where they belong — the kernel of truth silos were always reaching for. Whether that concentration translates into something practitioners call “topical authority” isn’t something Google documents directly; treat it as a reasonable hypothesis worth testing on your own site, not a guaranteed mechanism.

This site is a live example. patrickstox.com runs a pillar → cluster → article → sub-article hierarchy across a few hundred pages, using pillar, cluster, clusterSelf, subcluster, and subsubcluster taxonomy — and it deliberately cross-lists articles across clusters (the alsoIn pattern) rather than walling them off. That’s hub-and-spoke with intentional cross-linking: exactly the model I’m describing, not a strict silo.

When to use which model

Honestly? Build the hub-and-spoke / topic-cluster structure and stop worrying about the labels:

• Pick a pillar/hub topic with real breadth and search demand.
• Identify the subtopics that have their own demand — those become spokes.
• Interlink hub ↔ spokes and spoke ↔ spoke where relevant.
• Cross-link to other clusters when the context is genuinely related.

There’s no magic number of cluster pages per hub. The right count is however many distinct subtopics with real demand exist — not an arbitrary target of 5, 10, or 30. Wikipedia is the canonical example: broad overview pages linking deeply to detailed subtopic pages, all densely interlinked, no silo walls anywhere.

Architecture and AI search

AI Overviews and AI assistants use query fan-out — they decompose a question into multiple related sub-queries. The reasonable hypothesis, not a documented guarantee: a site with organized, interlinked coverage across a topic’s subtopics has a better chance of surfacing across more of those sub-queries, since more of the subtopics are covered by a findable, crawlable page. I haven’t seen controlled evidence isolating architecture as the cause here, so treat it as something to test on your own content rather than an established mechanism. Gary Illyes’ public position is that AI search optimization needs normal SEO — well-structured, crawlable, high-quality content — not a special architecture. The structure that already serves traditional search is the same one you’d build for AI search; there’s no separate playbook.

Auditing your architecture

A practical pass, mostly with crawl data:

• Find orphans — pages with no internal links in. Crawl the site (Ahrefs Site Audit, Screaming Frog) and look for zero-inlink pages.
• Find over-deep pages — important URLs buried many clicks from home; no fixed number is the rule, but if crawl data shows them getting crawled less, dig in.
• Map topics and find hub candidates — clusters of related pages missing a central overview.
• Run an internal-link gap analysis — related pages that should link to each other but don’t.

The honest caveat

Architecture is infrastructure, not a ranking shortcut. It enables crawling, controls how link equity flows, and helps engines understand context — but a beautiful hub-and-spoke structure wrapped around thin, low-value content still won’t rank. Illyes has noted it’s rare to see two results from one domain in a SERP; structure serves content quality and relevance, it doesn’t substitute for them. Get the structure right so that good content can do its job — that’s the whole point.

The neighboring topics in this cluster — how internal links pass signals, how crawl depth affects discovery, and faceted navigation on large sites — all plug into the same architecture decisions covered here.

AI summary

A condensed take on the Advanced version:

• Site architecture = the crawlable link graph, not the URL folders. It shapes discovery and how clearly Google understands each page’s context — it doesn’t guarantee crawling, indexing, ranking, or AI citations by itself.
• Three practitioner labels, one overlapping structure. Silo, hub and spoke, and topic clusters aren’t Google-documented categories; they converge on a pyramid / top-down hierarchy, which Mueller has described favorably: the pyramid structure “helps us a lot more to understand the context of individual pages.”
• Hub and spoke and topic clusters mostly overlap. Same underlying pattern, different communities — IA term vs. HubSpot’s 2017 content-marketing rebrand. “Pillar page” is just the marketing word for the hub.
• Drop strict silos. The valuable bit is topical concentration of links; no Google source requires the “never link between silos” rule, and it hurts UX.
• Internal links > URL structure. Google works out hierarchy from links, not path segments; Mueller has said SEOs over-focus on URL structure.
• Keep it shallow, but don’t chase a fixed number. Google hasn’t published a click-depth threshold; too deep hurts crawling and signal flow, too flat dilutes topical signal.
• No magic cluster count. Cover the subtopics with real demand; Wikipedia is the model.
• AI search needs normal SEO (Illyes) — organized, interlinked coverage is a reasonable hypothesis for helping with query fan-out, not a documented guarantee.
• Architecture is infrastructure, not a ranking shortcut — it can’t rescue thin content.

Official documentation

Primary-source guidance from the search engines on structure, links, and hierarchy.

Google

• In-Depth Guide to How Google Search Works — URL discovery via links, and Google’s own use of “hub page” / category-page language.
• Crawlable links — Make your links crawlable — every important page should be linked from at least one other page; links must be <a href>; anchor text and surrounding context for understanding linked pages.
• Importance of Link Architecture (2008) — Google’s long-standing guidance that internal link structure shapes PageRank flow and perceived page importance.

Bing / Microsoft

• Bing Webmaster Guidelines — a clear, flat, crawlable structure; reasonable internal links; heading hierarchy (H1/H2/H3); submit sitemaps.

Quotes from the source

On-the-record statements from Google. Where the source page supports it, each link is a deep link that jumps to the quoted passage.

John Mueller, Google — on hierarchy / pyramid structure

• “The top-down approach or pyramid structure helps us a lot more to understand the context of individual pages within the site.” — John Mueller, on site structure. Coverage (Search Engine Journal) · Coverage (Search Engine Roundtable)

John Mueller, Google — on URL structure vs. internal links

• Mueller has cautioned that some SEOs over-focus on URL/folder structure; Google relies on internal linking to understand hierarchy, not URL path segments. Coverage (Search Engine Roundtable)

John Mueller, Google — on crawl depth

• “I don’t think it would always have a negative effect. I do think if you make it too deep, then that makes it harder for us to crawl and harder for us to pass the signals around.” — John Mueller, on deep page hierarchies.

Google Search Central docs — on discovery via hub pages

• “Other pages are discovered when Google extracts a link from a known page to a new page: for example, a hub page, such as a category page, links to a new blog post.” Jump to quote

Google Search Central docs — on linkability

• “Every page you care about should have a link from at least one other page on your site.” Jump to quote

Gary Illyes, Google — on crawlability and AI search

• “MAKE THAT DAMN SITE CRAWLABLE.” — Gary Illyes (Reddit AMA), underscoring that discoverability via links is the priority over any specific architecture brand.
• Illyes has said AI search optimization requires only normal SEO — well-structured, crawlable, high-quality content — with no special architecture changes needed.

Site-architecture checklist

A pass to confirm your structure helps crawling, context, and link flow:

• Clear top-down hierarchy: home page → hub/category pages → specific pages.
• Every important page is linked from at least one other page (no orphans).
• Important pages aren’t buried needlessly deep (no fixed click-depth rule — check crawl data for pages getting crawled less as they get deeper).
• Each topic area has a hub/pillar page that links to its subtopic pages.
• Subtopic pages link back to their hub and to relevant sibling pages.
• Cross-cluster links exist where topics are genuinely related (no strict silo walls).
• Internal links use descriptive anchor text with relevant surrounding context.
• No important page relies on click-only navigation — links are real <a href>.
• URL structure is logical for humans, but you’re not relying on folders to signal architecture (links do that).
• Crawl data reviewed for over-deep pages, orphans, and internal-link gaps.

The mental models

1. The link graph IS the architecture. URL folders are for humans and housekeeping. Google works out hierarchy and context from how pages link to each other. Audit and design the link graph, not the folder tree.

2. Three names, one structure. Silo, hub and spoke, topic clusters → all converge on a grouped, hierarchical, internally linked pyramid once you allow sensible cross-links. Stop shopping for the “right” model; build the cluster and link it well.

3. Keep the kernel, drop the wall. From silos, keep topical concentration (link densely within a topic). Drop the isolation rule (never link between silos) — no Google source requires it, and it hurts users.

4. The depth dial. Too flat dilutes topical signal; too deep starves pages of crawling and link authority. Google hasn’t published a click-depth number, so aim for shallow-with-strong-linking judged against your own crawl data, grouped into hubs, rather than a fixed target.

5. Architecture is a multiplier, not a source. Good structure multiplies the value of good content by making it discoverable and contextually clear. Multiply zero (thin content) and you still get zero.

Architecture model reference

What each layer should do

Test yourself: Site Architecture

Five quick questions on silos, hubs, clusters, and what Google actually recommends. Pick an answer for each, then check.

Resources worth your time

My related writing

• The Beginner’s Guide to Technical SEO — where site architecture and internal linking fit in the bigger picture.
• Internal Links for SEO: An Actionable Guide — how internal links pass signals and shape the link graph that is your architecture.
• SEO Silo Structure: Why It Makes No Sense — the Ahrefs contrarian take on strict silos (the case for dropping the wall).
• How to Build a Topic Cluster — the practical hub-and-spoke / cluster build, step by step.

My speaking

• How Search Works (SlideShare) — my walkthrough of crawling, rendering, indexing, and ranking, the pipeline architecture has to serve. (Standing disclaimer applies: “This is my understanding of systems… not going to be 100% complete or accurate.”)

From around the industry

• Stop the Silo Madness: Effective Site Architecture for SEO and Findability (Shari Thurow, Search Engine Land) — the definitive case against strict silos.
• Site Architecture for SEO: Structure That Ranks & Scales (Search Engine Land) — a thorough structure guide.
• Complete Guide to Topic Clusters (Search Engine Land) — the cluster model in depth.
• Topic Clusters: The Next Evolution of SEO (HubSpot) — the 2017 origin of the “topic cluster” terminology.
• SEO Content Strategies: The Hub and Spoke Model (Botify) — the same structure under its information-architecture name.
• John Mueller Recommends Pyramid Site Structure (Search Engine Journal) — the source for the pyramid/context quote.

Site Architecture

Site architecture is how a website's pages are organized, categorized, and interlinked. It controls how crawlers discover pages, how link equity flows, and how clearly search engines understand each page's topical context. Silo structure, hub and spoke, and topic clusters are the three common models.

Related: Internal Link, Website Structure

Site Architecture

Site architecture (also called website structure or information architecture) is how the pages on a site are grouped, categorized, and linked together. A good architecture makes every important page reachable, concentrates internal links within topic areas, and gives search engines clear context for what each page is about. Google recommends a pyramid / top-down hierarchy — home page at the top, category or hub pages in the middle, specific content pages at the bottom.

Three organizational models dominate SEO discussion:

• Silo structure — content grouped into isolated thematic sections; the strict version prohibits cross-silo links. Popularized by Bruce Clay.
• Hub and spoke (also: content hub, pillar-and-cluster) — a broad overview “hub” page links out to detailed “spoke” pages and back.
• Topic clusters — a content-marketing model (popularized by HubSpot in 2017) where a pillar page targets a broad term and cluster pages target subtopics, all interlinked.

The key thing most write-ups miss: hub and spoke and topic clusters are functionally the same model — different labels from the information-architecture and content-marketing communities. And what actually moves the needle is internal linking, not URL folder structure: Google leans on contextual links to understand hierarchy, so strict physical silos that block relevant cross-links tend to hurt more than help.

Related: Internal Link, Website Structure

Changelog