Every link is a vote

Inside a website, a link is an endorsement. When one page links to another in its body text, an author has said: this matters, go here. PageRank, the algorithm that started Google, is simply a careful way of tallying those endorsements, with the twist that a vote from an important page counts for more than a vote from an obscure one. Run it over a site’s internal links and every page gets a score: its share of the site’s total authority.

Most people meet PageRank as a ranking: which page wins. For structural diagnosis, the ranking is the least interesting output. What we read is the distribution: how the total is shared out across all pages. A website, it turns out, has a wealth distribution, and like an economy it can be broadly middle-class or grotesquely top-heavy.

Borrowing the economist’s ruler

The standard instrument for measuring inequality is the Gini coefficient: 0 means everyone holds an equal share, and values approaching 1 mean one holder has everything. It works on link authority exactly as it works on income, and it compresses a site’s entire circulation pattern into one comparable number.

Across the Fortune 500 sites we have measured, editorial-graph Ginis run from around 0.3 into the high 0.8s. The low end looks like a healthy federation: authority spread over dozens of section hubs, the strongest single page holding a few percent of the total. The high end looks like a monarchy: the top 1% of pages holding half the authority, and in the extreme cases a single page holding a fifth to a quarter of everything.

Rule of thumb from our cohorts: a top page under 5% of total authority with a Gini near 0.3 reads as healthy distribution. A top page above 15%, or a top-1% share above 40%, is concentration worth investigating regardless of what the rest of the site looks like.

What concentration actually costs

Why care whether authority pools? Three consequences, in increasing order of subtlety.

Fragility. A page holding a fifth of the site’s authority is a structural single point of failure. Redesign it, split it, retire it, or break its URL, and the shock propagates through every page whose standing depended on flowing through it. Distributed sites absorb changes; concentrated sites transmit them.

Distortion. Authority is a proxy for attention: crawlers, link equity, and any agent that weights pages by their position in the graph all follow it. When one hub dominates, the site’s own structure declares that page to be what the organization is about. If the dominant page is a press index or a careers hub, that is the identity the structure broadcasts, whatever the homepage says.

Starvation. Authority is conserved: the share pooling at the top is exactly the share not reaching the long tail of substantive pages. Extreme concentration and widespread orphaning are usually the same finding viewed from opposite ends: everything links to the king, so nothing links to anything else.

The shape behind the number

The Gini alone does not tell you why authority pooled, and the same score can dress two different diseases. This is where the distribution’s shape earns its keep.

One shape is the spike: a single page far above everything, then a cliff, with the second hub holding a fraction of the first. That is the signature of a site that routes everything through one index, typically a rolling news or product hub that every page links back to and that links onward to almost nothing.

Another is the twin peak: two pages holding similar, enormous shares. We see it on funnel-built commercial sites where every page points at the same entrance and the same destination.

A third is the plateau: a dozen hubs at comparable strength, then a gradual slope. That is what a genuinely sectioned site produces, and it can coexist with a moderately high Gini simply because enterprise sites have long tails. Context decides: concentration on top of healthy communities and real depth is a warning light; concentration on top of a flat, orphan-heavy graph is the disease itself.

Reading it in the wild

In practice we read Circulation, the second of the five lenses, as three questions asked in order. How unequal is the distribution (Gini)? Where does the inequality live (top-page and top-1% share, and the shape of the fall)? And what is the concentration made of: which pages, holding what role, linked from where?

The third question is where the diagnosis lands, because the fix for concentration is never “remove links to the strong page.” It is to give authority somewhere else to go: real section hubs, cross-links between related content, and body-text links that endorse specific pages instead of the same index every time. Authority flows through whatever channels exist. A site that only builds one channel should not be surprised where its wealth ends up.

Method note: PageRank is computed on the editorial (in-content) graph with navigation links excluded, so scores reflect deliberate endorsement. Gini is computed over the full PageRank vector. Thresholds quoted are empirical rules of thumb from Axiom Graph’s Structural Signals cohorts (2026), not universal constants; site size and archetype shift what “normal” looks like. No client-identifying data appears in this note.