Hub & Spoke: The Content Architecture That Transforms Clusters Into Rankings
In the previous piece, we said that topical authority is built through clusters, not individual articles. Good in theory. But when you open that blank document and need to decide how to organize that content
Hub & Spoke: The Content Architecture That Transforms Clusters Into Rankings
In the previous piece, we said that topical authority is built through clusters, not individual articles. Good in theory. But when you open that blank document and need to decide how to organize that content — which piece is the hub, how many spokes you need, when a spoke becomes important enough to promote to hub status — theory isn't enough. — Topical Authority in 2026: Why Google Rewards Semantic Coverage Over Individual Keywords (see BeKnow pricing).
At BeKnow, we've codified a very precise model, because the AI that generates editorial plans needed deterministic rules, not inspirational guidelines. It's called Hub & Spoke, and it's how we transform strategic intuition into editorial structure that ranks.
What a hub really is
A hub is an article that does something very difficult: it covers the entire intent of a parent keyword, broadly enough to answer the general question, but deeply enough to avoid being superficial. It's not a landing page, not a long-form guide, not a pillar page in the classic 2018 sense.
A well-crafted hub answers these four questions:
What the topic is (operational definition)
Who needs it and when (use contexts)
How to approach it (overview of main solutions)
Where to start (call to action toward spokes)
The hub should not exhaust every subtopic. It should map the territory and direct traffic. If it tries to say everything, it becomes an unreadable monster and the spokes lose their purpose.
What a spoke is
A spoke is a vertical article that deep-dives into a single subtopic or answers a specific question within the hub's territory. The four recurring types are:
How-to: "How to do X" — instructional, practical.
Comparison: "X vs Y" — comparisons between approaches, tools, methods.
Problem: "Why X doesn't work" — diagnostic.
Solution: "The best way to X" — prescriptive.
Every spoke links to the hub (never the implicit reverse: the link is explicit and contextual) and links to at least one other spoke in the same cluster. This creates the network that LLMs read as an expertise signal.
The rule of 6: why no more than six spokes per hub
One of our most debated internal decisions was setting a limit. We could have allowed infinite freedom. We chose maximum 6 spokes per hub, and the reason is empirical.
Above six spokes, three things happen, all negative:
Semantic dilution: the model starts reading the cluster as "many vaguely related things" instead of "a topic covered in depth."
Cannibalization: spokes start overlapping. Two articles on the same sub-intent steal signals from each other.
Impossible maintenance: updating 12 spokes every quarter is theoretically possible, practically nobody does it. Oversized clusters die from staleness.
Six is the sweet spot where the cluster is rich enough to build authority but compact enough to stay maintainable.
When a spoke becomes a hub: the 15-query rule
The pattern we've observed and codified in BeKnow is very clear: when a spoke starts ranking for more than 15 distinct queries (read via Google Search Console), it stops being a spoke. It's attempting, on its own, to answer a broader intent than originally planned. Two options:
Promote it to hub of a new cluster, and start writing its spokes.
Narrow it down if the 15 queries are the result of semantic accidents, not real expansion.
In 90% of cases, promotion is the right move. It's the most reliable signal that the market is telling you "there's territory here, come claim it." Ignoring it means leaving a competitor the cluster that your own audience was building for you.
Internal linking structure
Three rules, that's it:
Every spoke links to the hub within the first 30% of the text, contextually (never "click here").
Every spoke links to at least one other spoke in the same cluster, chosen for intent affinity.
The hub links to all spokes in a dedicated section, but also inline where the discussion requires it.
No forced reciprocal links, no "you might also like," no generic link footers. The network must be semantic, not mechanical.
How to recognize a poorly built cluster
Three unmistakable symptoms:
The hub is a list page. If the hub is just a list of links to spokes with two intro lines, it's not a hub. It's a menu. LLMs treat it as navigation, not authoritative content.
Spokes answer each other. If two spokes address the same sub-intent with slightly different angles, you have a cannibalization problem. Operational threshold: if keyword overlap between two spokes exceeds 85%, they should be merged.
No spoke has grown in 90 days. If after three months no piece of the cluster has gained organic traffic, you don't have an SEO problem: you have an intent problem. The cluster is answering questions nobody asks, or answering them poorly.
Hub & spoke vs classic pillar page
For those coming from 2010s SEO: the fundamental difference is that the classic pillar page was designed for ranking the pillar itself. The modern hub is designed for ranking the entire cluster. It's a philosophical difference with enormous practical consequences:
The hub can be relatively short (1500-2500 words), because it doesn't need to exhaust the topic.
Spokes are the real traffic workhorses, each optimized for its specific query.
The value isn't in any single article: it's in the structure.
This also means measuring success by looking at only the hub's ranking is wrong. You need to measure the cluster's aggregate ranking, queries covered, overall traffic potential.
In summary
Hub & spoke isn't a marketing framework, it's an architecture. It has precise rules — six-spoke limit, 15-query threshold for promotion, 85% threshold for merging — because without rules, a cluster becomes organized noise.
If you have a blog with 100 scattered articles and want to recover authority, the right move isn't writing another 100. It's mapping what you have, identifying 5-10 potential hubs, understanding which existing articles are already spokes (even if they didn't know it), rewriting them to align with the cluster, merging duplicates. Six months of this operation is worth two years of blind publishing.
In the next article, we'll dive into the most underestimated problem for content managers: cannibalization, and how to detect it before it causes damage.
Related reads
Ready to Transform Your Content Strategy?
Start creating SEO-optimized content with AI-powered semantic intelligence.
See pricing