Site architecture and internal linking: the invisible ranking lever

Every few months a marketing director sends us their analytics and asks the same question. We publish more than our competitors. Our articles are better researched. Our pages have higher word counts and cleaner writing. So why are we losing to a site that posts half as often and has half our domain authority?

The answer, almost always, is structure. They've built a brilliant content library on top of a fragmented site, and the architecture is leaking equity faster than the content can earn it. Google can crawl their pages, but it can't tell which ones matter. Their internal links scatter PageRank like loose change. Their categories are an accidental taxonomy that emerged organically over five years of CMS edits. The content is fighting the site.

This article is for the SEO leads and technical marketers who already know what anchor text is and have read the basics. We're going to walk through how site architecture actually moves rankings, why most large sites have the same six structural problems, and how to audit and fix them without breaking anything.

Why content alone never wins

Search engines do two things before they ever read your copy. They crawl your site to discover URLs, and they evaluate the relative importance of those URLs by looking at how they're linked from inside and outside the site. Content is what gets ranked once those two steps are complete. If your crawl is broken or your internal link graph is incoherent, the best content in your category will sit on page three of the SERP while a thinner competitor outranks you.

The trap is that content is visible and architecture isn't. You can read a blog post. You can show a client a beautiful longread. You can't show them a healthy link graph in a board meeting. So content gets the budget and architecture gets ignored, until someone finally pulls the data and realises that 40% of the site's pages have zero internal links pointing at them.

Site architecture fundamentals

Flat versus deep architecture

A flat architecture means every page on your site is reachable in two or three clicks from the homepage. A deep architecture pushes content five, six, seven clicks down. Conventional wisdom says flat is better, and for most marketing sites it is, but the reason isn't the click count itself. It's the link distance from your most authoritative URL.

Your homepage typically has the most external backlinks. Every internal link from the homepage passes a slice of that authority to the next page. Each additional hop dilutes the flow. A page sitting six clicks deep on a site with no strong internal links pointing to it is, from Google's perspective, almost orphaned. The page might be magnificent. Nobody will ever see it.

The three-clicks principle and when it breaks

The old rule said no page should be more than three clicks from the homepage. It was useful in 2008. Today it's a guideline, not a law. Large ecommerce catalogues, news archives and documentation sites can't physically obey the rule. A retailer with 80,000 SKUs is going to bury most product pages.

The principle still holds for pages you want to rank. Your top-tier commercial pages, your pillar content, your service pages, your highest-converting category pages, those should be within three clicks. The rest can sit deeper, provided you've engineered other ways for crawlers and equity to find them: well-structured XML sitemaps, contextual links from articles, related-content modules, breadcrumb trails.

Siloing versus topic clusters

Siloing was the dominant architectural model for a decade. You group related content into rigid sections of the site, link generously within each silo, and avoid linking across silos to keep topical authority concentrated. It worked, and it still works for some site types.

The modern alternative is the topic cluster model, popularised by HubSpot but rooted in how Google's entity-based ranking actually behaves. You build a pillar page that covers a broad topic comprehensively, then cluster of supporting articles that go deeper on subtopics, all linking back to the pillar and to each other. The semantic relationship matters more than the URL hierarchy.

Topic clusters tend to outperform pure silos on informational queries because they mirror how Google's knowledge graph thinks about entities and relationships. Silos still win for transactional architecture where category specificity is the ranking signal. Most mature sites end up doing both: silos at the commercial top of the funnel, clusters in the editorial middle.

How Googlebot navigates your structure

Googlebot doesn't visit your homepage and read your navigation like a human would. It discovers URLs from your XML sitemap and from links it finds on other pages. It prioritises crawling based on PageRank signals and historical update frequency. Pages it can't easily reach, or pages with no internal links pointing to them, get crawled less often or not at all.

Two implications follow. First, your XML sitemap is not a substitute for internal linking. It tells Google a URL exists. Internal links tell Google it matters. Second, render-blocking JavaScript navigation that hides links from a non-executed crawl will hurt you. Most modern crawlers execute JS, but they do it slowly and with limited frequency. Server-rendered links remain the safest default.

Architecture is mostly a decision about where equity should flow. Source: Pexels

PageRank and link equity

How PageRank actually flows

The original PageRank paper is twenty-five years old and Google's algorithm has evolved a hundred times since. The fundamental mechanic, however, still holds: a page passes a fraction of its authority to every page it links to, and that authority compounds across the link graph.

Practically, this means three things. Pages with many strong inbound links accumulate equity. Pages with many outbound links dilute the equity they pass per link. And pages with no inbound internal links accumulate nothing, regardless of external signals.

You don't need to model PageRank mathematically. You need to think about it directionally. Where is the authority on your site concentrated? Are you pointing it at the URLs that need to rank?

Orphan pages: the silent killer

An orphan page is one that exists in your CMS, appears in your sitemap, but has no internal link pointing to it from anywhere else on the site. Google will discover it, crawl it occasionally, and treat it as an unimportant outlier. Even if the content is excellent, it has almost no chance of ranking competitively.

Orphans accumulate naturally. A blog post gets unpublished from the index but stays live. A landing page from a 2022 campaign is no longer linked from the navigation. A product gets retired but the URL persists. Audit your site and you'll find dozens of them.

Fixing orphans is one of the highest-ROI tasks in technical SEO because it's quick. Find them, decide whether they deserve to live, link to the ones that do, and 301 or remove the rest.

Crawl budget on large sites

For sites under 10,000 URLs, crawl budget is rarely the binding constraint. Google will get to everything that matters. For sites in the hundreds of thousands or millions of URLs, crawl budget becomes a real bottleneck. Googlebot allocates a finite number of fetches per day based on your server's response times, your site's perceived authority, and how often content changes.

When budget is tight, low-value URLs starve high-value ones. Faceted navigation, parameter variants, paginated archives and tag pages collectively eat a disproportionate share. Tightening crawl budget is a structural exercise: noindex the noise, canonicalise the duplicates, block the dead ends in robots.txt, and make sure the URLs you actually want indexed are the ones with the strongest internal link signals.

Internal links are the lever most teams haven't pulled yet. Source: Pexels

Internal linking strategy

Anchor text: the three flavours

Internal anchor text tells Google what the destination page is about. There are three useful categories and you want a mix of all three.

Exact-match anchors repeat the target keyword precisely. "Site architecture SEO" linking to a page about site architecture SEO. They send the strongest topical signal but used at high density they look manipulative.

Branded anchors use your brand or product name. "Read the Nexus guide." They build entity recognition and look natural at scale.

Descriptive anchors describe the destination in natural language. "How internal links influence ranking." They're the workhorse: they read well, they pass context, and Google has been weighting them more heavily for years.

A healthy internal anchor profile leans heavily on descriptive, uses exact-match selectively on the most important commercial pages, and lets branded anchors emerge naturally from in-line writing.

Pillar pages and the cluster model

Picture a wheel. At the centre, a pillar page covering a broad topic: "B2B content marketing strategy." Around it, twelve cluster pages, each a deep dive on one subtopic: lead magnets, content audits, distribution, SaaS case studies, and so on. Every cluster page links back to the pillar with a contextual anchor. The pillar links out to every cluster page. Cluster pages link to each other where the topics overlap.

The pillar accumulates authority from the cluster, ranks for the broad head term, and feeds equity back into the cluster for long-tail queries. The whole structure reads as a coherent topical entity to Google.

The mistake teams make is treating the pillar as a glorified hub page with a list of links. It needs to be substantive: 3,000+ words, ranking on its own merits, with the cluster expanding rather than substituting.

Auditing internal links with Screaming Frog

The fastest way to see what your link graph actually looks like is Screaming Frog. Run a full crawl, then look at three reports.

Inlinks per page. Sort by inlinks ascending. Pages at the bottom with zero or one are your orphans and near-orphans. Cross-reference with your sitemap and analytics: any page with traffic potential and fewer than three inbound internal links is a priority.

Anchor text report. Export all internal anchors. Pivot by destination URL. Look at the anchors pointing at your top commercial pages. If your "/services/web-design" page is being linked with "click here" and "find out more" 70% of the time, you have an anchor-text problem you can fix in a week.

Crawl depth. Filter for pages four or more clicks from the homepage. Anything that should be ranking but sits at depth six needs to be pulled up with new internal links from higher-authority pages.

This audit takes a competent SEO half a day on a 10,000-URL site. The fixes can move rankings within weeks.

Priority linking on a new site

When you launch a new site or section, internal links are your only ranking lever for the first 60 days. External backlinks take time. Internal links work immediately.

Decide your top ten priority URLs before launch. Engineer at least five contextual internal links pointing at each of them from existing content. Use varied descriptive anchors. Place links high in the page, in body copy, not buried in footers. Footer links are weighted lower than contextual ones and Google has been explicit about this.

🎬 Watch: Mapping out a site before the build

Pillar pages and cluster routes, sketched before they become URL trees.

Navigation and taxonomy

Category structure as a ranking signal

Your main navigation tells Google which pages you consider the most important on the site. Pages linked from the global nav inherit equity from every other page on the site, since the nav appears everywhere. That's enormous leverage.

It's also a finite resource. A nav with 18 top-level items dilutes the signal. A nav with five well-chosen items concentrates it. Ruthless prioritisation of what gets nav placement is one of the cheapest SEO wins on a mature site.

Faceted navigation and the duplicate content trap

Faceted navigation lets users filter products by attributes: colour, price, size, brand. Useful for users, catastrophic for SEO if uncontrolled. Each filter combination generates a unique URL. A catalogue with five facets and ten values each can theoretically produce 100,000 URL variants, most of which are near-duplicates of each other.

The standard fix is a combination of rel=canonical pointing filtered variants back to the parent category, robots.txt or meta-robots noindex on the deep combinations, and selective indexing of high-value facet combinations (e.g. "red Nike trainers") that match real search demand. Done well, faceted nav becomes a long-tail ranking machine. Done badly, it tanks crawl budget.

Breadcrumbs as a ranking signal

Breadcrumbs do three useful things. They give users a sense of place. They build a structured internal link from every deep page back up the hierarchy. And, marked up with schema.org BreadcrumbList, they appear in Google's SERP, replacing the URL line with a readable path and lifting click-through.

Breadcrumbs aren't optional on any site with more than two levels of depth. They're a fifteen-minute implementation that compounds value across thousands of pages.

The site map you ship is the one Google actually reads. Source: Pexels

Common mistakes and how to fix them

Too many top-level categories. Sites that grew organically often have 12–15 main nav items. Audit what each one actually serves. Most can be consolidated. The target is five to seven.

Pagination handled incorrectly. Old advice said use rel=prev/next. Google deprecated support for that markup in 2019. Modern practice is self-referential canonicals on each paginated page (page 2 canonicals to page 2, not page 1) and indexable paginated pages so Google can reach the deep content. Don't noindex pagination unless you have a very specific reason.

CMS-generated URL fragments. WordPress, Shopify and Magento generate URL structures that fragment equity across tag pages, author pages, dated archives and category combinations. Audit which of those URL types are indexed and prune ruthlessly. If you're not actively maintaining "/tag/" archives as ranking pages, noindex them.

Duplicate URLs with and without trailing slashes, or with query parameters. Decide a canonical format and enforce it with 301s. Pick one capitalisation, one slash convention, one parameter order. Inconsistency leaks equity quietly.

Header navigation hidden behind JavaScript. If your nav requires JS to render the link tags, you're betting on Google's renderer. Server-render the links.

Internal links going to non-canonical URLs. Audit destinations. Every internal link should point at the canonical version of the target page, not a redirect. Redirected internal links waste a fraction of equity per hop, and on a large site that fraction adds up.

How Nexus architects for SEO

When we plan a new site or a migration, architecture sits in week one alongside brand and content strategy, not after. We map the priority URLs first, work backwards to the navigation, define the silo or cluster boundaries, and engineer the internal link graph as a deliverable, not an afterthought. We use Screaming Frog and the live analytics from any predecessor site to model what equity should flow where.

For migrations specifically, the architecture review is the single highest-stakes piece of work. Most ranking drops after a relaunch trace back not to content changes or technical regressions but to internal links being silently restructured by the new CMS. A pre-launch crawl comparison catches it. A post-launch one is reactive damage control.

For ongoing growth, we run quarterly internal link audits on enterprise accounts. Three hours of work usually surfaces a dozen fixable issues that move keyword rankings within the following month.

Want an architecture review before you build or migrate?

If you're planning a relaunch, a migration, or just suspect your current site's structure is leaving rankings on the table, we'll do a paid architecture audit and give you a prioritised fix list. Most clients recover the cost in the first quarter from rankings that move once the structural drag is removed. Book a 30-minute call and we'll scope it from there.