Site architecture is how pages relate to each other through links and URL structure. It determines what search engines find, how often they return, and how authority moves between pages.
It is also the area with the best ratio of effort to result on established sites, because fixing it requires no new content. You are redistributing signal that already exists.
Click Depth Is the Metric That Matters Most
Click depth is the minimum number of clicks from the homepage to a page. The working consensus across technical audits is that pages beyond three or four clicks receive noticeably less crawl attention and rank weaker than equivalent pages sitting shallower.
The mechanism is straightforward. Crawlers allocate finite attention per site. Pages requiring many hops to reach are visited less often, which means updates take longer to register and new pages take longer to index.
Depth is also a proxy for internal signal. A page four hops down typically has fewer internal links pointing at it than a page two hops down, and internal links are how authority distributes.
Crawl your own site and pull the depth distribution. Most sites find a long tail sitting at depth five, six, or worse, almost always because the only route to those pages is paginated archive browsing.
Flat Beats Deep, Within Limits
The instinct on discovering deep pages is to flatten everything, which overcorrects. A completely flat structure where every page sits one click from the homepage means the homepage links to thousands of pages, diluting each link to near meaninglessness.
The workable shape is a shallow hierarchy: homepage, then category or hub pages, then content, with most content reachable in three clicks. That keeps depth low while preserving the topical grouping that tells search engines which pages relate to each other.
Category pages do double duty here. They shorten the path to content and they function as hub pages for a topic, which is the clustering structure covered in the topical authority guide.
Orphan Pages Are the Sharper Version of the Problem
An orphan page has zero internal links pointing at it. It exists, it may be in the sitemap, and nothing on the site references it.
Audits routinely find a substantial share of pages in this state, usually old content that fell off pagination, landing pages built outside the main template, and pages created for campaigns that ended.
A sitemap gets a page discovered. It carries no signal about importance, and a page with no internal links is telling search engines that the site itself does not consider it worth referencing.
The fix is either linking to it from somewhere relevant or removing it, which is the pruning decision covered in the pruning guide. Leaving orphans indefinitely is the option that helps nobody.
URL Structure Should Describe, Not Impress
A URL should be short, readable, and contain the primary term. Hyphens between words, lowercase, no unnecessary parameters.
Whether the URL reflects the site hierarchy is a genuine choice rather than a rule. A flat URL like /product-name is easier to maintain and never breaks when a product moves category. A nested URL like /category/subcategory/product-name communicates structure and creates natural breadcrumb paths.
My position: use flat URLs for content that might move between categories, and nested URLs where the hierarchy is genuinely stable. The failure mode with nesting is a product recategorised three times leaving a trail of redirects, and that failure is common enough to weight toward flat.
Breadcrumb markup restores the hierarchy signal without requiring it in the URL, which is covered in the breadcrumb guide. That makes flat URLs a smaller compromise than they used to be.
Stability Beats Optimisation
Changing URLs to improve them costs redirects and loses some accumulated signal. An established page with a mediocre URL is usually better left alone.
This applies to architecture generally. A restructure that moves hundreds of URLs is a migration, with all the risk covered in the migration guide, and it should be justified by more than tidiness.
Where restructuring is genuinely needed, do it once and properly rather than incrementally, because each partial change adds another layer of redirects to the eventual chain.
Faceted Navigation Is Where Ecommerce Architecture Fails
Filters for colour, size, price, and brand combine into an enormous number of URLs, most describing near identical inventory.
Left indexable, this creates thousands of thin duplicate pages competing with each other and consuming crawl attention. Large sites can lose a meaningful share of crawl coverage to URLs that should never rank.
The handling: canonical tags on filtered views pointing at the unfiltered category, except for specific filter combinations with genuine independent search demand, which get treated as real pages.
Blocking filter parameters in robots.txt is the common instinct and usually wrong, since a blocked URL cannot be crawled, so its canonical is never read, so no consolidation happens. Let crawlers reach them and let canonicals resolve them, per the canonicalization guide.
Pagination Architecture
Paginated archives are how most sites bury their older content. Forty pages of archive means page forty is forty hops deep and effectively unreachable.
Category and tag pages shorten those paths substantially. A post sitting on page thirty eight of the main archive might sit on page two of its category, which changes it from unreachable to routinely crawled.
This only works if the taxonomy is real. Thirty tags each containing one post creates thirty thin pages and no useful shortcuts. A handful of genuine categories with meaningful membership does the job, and the pagination guide covers the technical implementation.
Internal Link Distribution Is the Lever
Architecture is expressed through links, so changing where links point changes the architecture without moving a single URL.
Link from high authority pages to pages you want ranked. Most sites have a handful of pages carrying most of their external links, and those pages frequently link nowhere useful.
Link contextually within content rather than only through navigation. A link from within a relevant paragraph carries more signal than the same link in a footer.
Audit for pages receiving zero or few internal links and fix the worst offenders. This is the cheapest ranking work available on an established site and it is routinely deferred because it produces no visible artefact.
What This Buys Beyond Rankings
Every architectural improvement here also determines whether AI systems can find and understand your content, since discovery works the same way. Shallow, well linked, clearly grouped content is easier for any system to traverse.
That overlap means architecture work does not need a separate business case for AI visibility. It is the same work.
Subdomains Split Architecture in Ways Subdirectories Do Not
Putting a blog on a separate subdomain rather than in a subdirectory divides the site into two structures that accumulate signal separately. Whether that costs you depends on how strongly the connection is signalled elsewhere, and it is a decision usually made by whoever configured the CMS rather than deliberately.
The subdomain guide covers the full trade. The architectural point is that a subdirectory inherits the main site's structure automatically while a subdomain requires you to rebuild the connection through cross linking and consistent markup.
Navigation Built in JavaScript Is Not Navigation
A menu that renders only after JavaScript executes provides no crawl path in the initial HTML. The site appears navigable to a visitor and appears to have almost no internal linking to a crawler that does not render.
This is worth checking specifically because it is invisible in normal testing. Load your homepage with JavaScript disabled and see whether your main navigation still produces working links, per the rendering guide. Where it does not, your architecture exists only for humans.
Verifying What Gets Reached
A crawl tool shows you your own architecture. Whether crawlers actually traverse it as expected is a separate question, and the log file guide covers reading server logs to see real crawler behaviour rather than modelled behaviour.
The NotionCue AI Crawler Audit confirms reachability for specific crawlers on specific URLs, which catches the case where architecture looks correct and a WAF rule or rendering issue prevents traversal.
Start your free NotionCue trial and check pages sitting deep in your structure. Those are where access failures hide longest, because nobody tests them.
One number worth pulling from your next crawl: how many pages sit at click depth five or deeper. On most sites that number is larger than anyone expects, and the pages in it are usually older content that still has search demand.
Common Questions
Does URL length affect rankings directly?
Marginally at most. Shorter URLs correlate with better rankings largely because shorter URLs tend to belong to shallower, more important pages. The correlation is real and the causation runs through depth rather than character count.
Should I restructure an established site with poor architecture?
Fix internal linking first, since it addresses most of the problem without redirects. Restructure URLs only where the current structure is actively causing errors, and treat it as a migration when you do.
How many internal links should a page have?
Enough to be useful and not so many that each carries nothing. There is no threshold worth targeting. A page with three relevant contextual links outperforms one with forty scattered through a footer.