Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

SEO › Technical SEO

Site Architecture for SEO

How to structure a site so search engines and people find every page: hierarchy, URL design, click depth, hub pages, navigation and pagination.
Guideintermediate8 min readUpdated

Why structure matters to a search engine

Crawlers discover pages by following links, allocate attention by apparent importance, and infer importance largely from how a site links to a page. Three consequences follow.

Depth decides discovery and freshness. Pages close to the homepage are found early and recrawled often; pages many clicks away are found late, recrawled rarely and, on large sites, sometimes not at all. Click depth — the minimum number of links from the homepage to a page — is the single most useful architecture metric.

Links decide authority. External links give a site authority; internal links decide which pages receive it. A page linked from the homepage and every hub is treated as important; a page linked once from a footer is not. The internal linking guide covers the page-level craft; this guide covers the structure that internal links run through.

Grouping decides understanding. Pages that are linked together, share a URL path and share a navigation section are understood as a topic. A site whose pages on one subject are scattered across unrelated sections makes the engine work out the topic on its own, and it often concludes the site has no particular authority on it. The topic clusters guide applies this to content planning.

The shape that works

Most sites that rank well share one shape, however the surface navigation looks:

  1. Homepage
  2. Section hubs
  3. Sub-hubs where needed
  4. Pages
  5. Links back up and sideways
A hierarchy for discovery and understanding; cross-links within sections for authority and context.
  • The homepage links to every major section and to the handful of pages the business most needs to rank — not to everything.
  • Section hubs (categories, topic hubs, service areas) list and link to every page in their section, grouped so the list is readable. A hub is a page in its own right, with its own content and ranking target, not a bare list.
  • Sub-hubs exist only where a section is too large for one hub to list; each adds a click of depth, so they are used sparingly.
  • Pages link up to their hub (breadcrumbs do this automatically), sideways to the pages they are most related to, and outward to other sections where a reader would want to go next.
  • Everything is reachable through links a crawler can follow: plain anchor elements with href attributes, not click handlers, not elements that appear only after interaction (the JavaScript SEO guide explains why).

The shape is sometimes called hub-and-spoke, pyramid or silo. The names differ on one point: whether cross-links between sections are allowed. The strict "silo" idea — never link across sections — has no documented basis; linking to a genuinely related page in another section helps readers and does no harm. What should be avoided is the opposite: so many cross-links that sections stop being distinguishable. How the homepage's links serve visitors as well as crawlers, with routes ordered by what visitors come for, is covered in the guide to the homepage as a conversion page.

URLs

The URL is part of the architecture because people and engines read it. Documented guidance from Google is modest: keep URLs simple, readable and logical, use hyphens between words, avoid long parameter strings where a readable path would do. Beyond that, practical rules:

  • Mirror the hierarchy where it helps (/seo/technical/core-web-vitals) and keep it shallow; a path with six segments reflects a site with six levels, which is the real problem.
  • One URL per page. Variants — trailing slash or not, upper and lower case, parameters that do not change content — are duplicates and need redirects or canonicals (see crawling and indexing).
  • Do not encode things that change. Dates, categories and product attributes in URLs force redirects whenever they change; a stable identifier does not.
  • Change URLs only with a reason and a redirect map. The migrations guide covers the process; the reason is rarely "better keywords in the URL". The redirect map template records each changed URL with its one-to-one destination, status code and test result. When only one page moves, the SOP for changing a page URL runs the change from approval to the logged result, including the ads, emails and profiles that still carry the old address.

Where query strings are the source of the variants, the tutorial on auditing URL parameters lists every parameter the site emits and sets one rule per class.

Main navigation is the strongest internal-linking signal a site sends, because it appears on every page. Three things to get right:

  1. It links to hubs, not to everything. A mega-menu with hundreds of links spreads authority thin and buries the important pages among the trivial ones.
  2. It is crawlable. Menus rendered only on hover or click, by JavaScript that builds the links at runtime, may not be followed. The links should exist in the HTML; styling can hide them until interaction.
  3. It matches the hierarchy. If the menu says a page belongs to one section and its URL and breadcrumbs say another, the engine has to guess.

Breadcrumbs, footer links and in-content links complete the picture. Breadcrumbs are the cheapest architecture win available: they put a crawlable path from every page back to the homepage, and with BreadcrumbList structured data they appear in results.

Pagination and lists

Long lists — category pages, blog archives, search-like listings — raise two architecture questions: how do crawlers reach the items on page nine, and which page represents the list? Documented position: Google no longer uses rel="prev" and rel="next" as indexing signals (it announced this in 2019), and treats each paginated page as a separate URL. Practical approach:

  • Each paginated page is indexable, self-canonical (not canonicalised to page one, which hides the items on later pages) and linked to the pages around it with plain links.
  • Items that matter are also linked from somewhere other than deep pagination: sub-category hubs, "featured" lists, related-item links.
  • "Load more" and infinite scroll are implemented so that each batch also has a real paginated URL, otherwise the items beyond the first batch are invisible to crawlers.

For e-commerce the related and larger problem is faceted navigation, covered in its own guide.

Diagnosing an architecture

A crawler (the kind that crawls your own site, not Googlebot) produces the data; four reports do most of the work:

ReportWhat it revealsAction
Click depth distributionHow many pages sit at each depth; important pages deeper than threeAdd hub links; flatten sub-hubs; link from the homepage
Inlinks per pagePages with one or zero internal links; pages with hundredsLink the under-linked from hubs and siblings; prune the over-linked menus
Orphans (crawl versus sitemap versus server list)Pages that exist but are linked from nowhereLink them or remove them
Crawl frequency by section (from server logs or Search Console crawl stats)Sections Googlebot visits often versus rarelyCheck depth and links for the neglected sections

Compare the result with business priority: the pages that should rank should be the shallowest, most-linked pages. When they are not, the architecture is working against the strategy.

Illustrative example — not a real result

A publisher whose best guides nobody could find

A publisher's crawl shows its evergreen guides — the pages meant to rank — sitting five clicks deep, reached only through a date-ordered archive that pushed them further back every week, while the homepage linked to sixty recent articles. The fix was structural: a topic hub for each main subject, linked from the main navigation, listing the guides first and the recent articles after; a related-guides block on every article; and the archive left in place but no longer the only path. The guides moved to two clicks from the homepage and their internal link counts rose from one to dozens, which is the condition for rankings to follow once the pages are recrawled.

Designing an architecture from scratch

  1. Start from keyword mapping: the list of topics and the page that will serve each defines the pages, and the clusters define the sections.
  2. Group pages into sections by topic, not by content type or department. A reader looking for everything about one subject should find it in one place.
  3. Give each section a hub with a ranking target of its own.
  4. Draw the depth: homepage, hubs, pages. Add a sub-hub level only where a hub would list more than can be read.
  5. Define the URL pattern per level and apply it consistently.
  6. Specify the cross-links: breadcrumbs, related pages within sections, deliberate links across sections.
  7. Check every page has at least three crawlable internal links and no page is more than three clicks from the homepage, before launch.

When the design turns into new templates, writing SEO requirements for developers converts its URL patterns, linking and indexing rules into a specification developers can build and test.

When a blog, shop or help centre needs a home, the subfolder versus subdomain comparison explains how each option affects authority, crawling and the effort needed to run it.

Deep, poorly linked pages often appear in Search Console as known but never fetched; the troubleshooting page on discovered URLs that are not yet indexed shows how to confirm and fix that.

Whether adding contextual links to deep, under-linked pages pays on a particular site can be measured with the page-group internal linking experiment.

Common mistakes

  • Chronological structures for evergreen content. Archives bury the pages that should rank.
  • Navigation that links to everything. Authority spread evenly is authority spent on nothing.
  • Sections by department instead of by topic. The engine sees fragments of a topic across the site.
  • Pagination canonicalised to page one. Hides every item after the first page.
  • Orphans created by redesigns. Old pages still live, no longer linked, slowly dropping out.
  • URL changes for cosmetic reasons. Every change is a migration with migration risks.
  • Hubs that are bare lists. A hub with no content of its own rarely ranks for the section's head term.

The internal linking guide is the page-level companion to this one; category and product pages applies the same structure to a store; and the crawling and indexing guide explains what the engine does with the links the architecture provides.

Frequently asked questions

Does URL structure affect rankings?

Google documents that a simple, descriptive URL structure helps users and crawlers understand a site, and that words in URLs can be a minor factor. The large effects of architecture come through linking and depth, not from the characters in the path. Change URLs only with redirects and only for a reason.

How deep is too deep?

There is no documented limit. In practice, pages more than three or four clicks from the homepage are crawled less often and receive less internal authority, and pages reachable only through pagination or filters are at risk. Important pages should be within three clicks; the number is a rule of thumb, not a rule.

Should every page link to every other page in its section?

No. Links carry weight in proportion to how selective they are; a page with hundreds of links passes little through each. Link the pages that belong together, in both directions, and let hubs do the broad linking.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Part of these learning paths

Sources

  1. SEO Starter Guide: Organize your site — Google Search Central

Related