Site Architecture for SEO

Click Depth and Crawl Efficiency

Click depth is the number of clicks from the homepage to a given page. Pages buried five or six clicks deep are crawled less often and typically rank worse, independent of their content quality.

The working target is three clicks to any important page. For large sites this requires deliberate hub pages, sensible category structures, and pagination that does not strand content.

A flat architecture is not automatically better. Thousands of pages all linked from the homepage dilutes the signal about what matters. Hierarchy with reasonable depth beats both extremes.

URL Structure

URLs should be readable, stable, and reflect hierarchy. /codex/seo/technical/page-speed/ communicates structure to both users and crawlers.

Practical rules: lowercase, hyphens rather than underscores, no unnecessary parameters, no dates in URLs for evergreen content, and no stop-word padding. Keep them short enough to display fully in results.

Most importantly, do not change them without reason. Every URL change requires a redirect and loses a small amount of signal.

Hierarchy and Hub Pages

A well-formed hierarchy has a homepage linking to section hubs, hubs linking to subsections or individual pages, and individual pages linking laterally to related content.

Hub pages do real work. They consolidate topical authority, give crawlers an efficient path to related content, and often rank for broader head terms that individual pages cannot.

Internal Linking as Architecture

Internal links are the mechanism through which architecture actually functions. Pages with many internal links pointing at them are treated as more important.

Deliberate practices: link from high-authority pages to pages you want to promote, use descriptive anchor text rather than "click here", link contextually within body content rather than only in navigation, and audit periodically for orphan pages with no internal links at all.

Diagnosing an Architecture You Inherited

Most architecture work is done on a site that already exists, and the useful first step is measurement rather than redesign.

Crawl the site and plot click depth from the homepage. Anything beyond three or four clicks is a candidate for promotion. The distribution matters more than any single page: a long tail at depth six usually means a category layer is missing.

Compare the crawl against the sitemap and against your analytics. Three sets worth looking at: URLs in the crawl but not the sitemap, URLs in the sitemap but not reachable by crawling, and URLs receiving traffic that appear in neither. Each set is a different defect.

Look at log files if you can get them. They show what search engines actually requested rather than what you think is important, and the gap between the two is usually the finding.

Count internal links per page. Pages with one inbound internal link are relying entirely on the sitemap, which is a weak signal and a fragile one.

Changing It Without Losing What Works

Restructuring is the highest-risk routine SEO work because the downside is immediate and the upside is gradual.

The sequence that limits the damage: map every current URL to its destination before touching anything, including the ones you intend to retire, and decide for each whether it redirects, stays, or returns a gone status. A redirect map built after launch is built from server errors.

Redirect to the closest equivalent, not to the homepage. Bulk redirects to the root are treated as soft errors and discard the signal you were trying to preserve.

Update internal links to point at final destinations rather than relying on the redirects. Chains dilute, slow crawling and accumulate.

Keep the old sitemap available for a period so the redirects are discovered promptly, then retire it.

Expect a dip. Recovery on a well-executed migration typically takes weeks, and the most common mistake is reversing course inside that window on the basis of a fortnight's data — turning one disruption into two.

Sources

What each claim on this page rests on. Entries are typed so you can see which are primary.