Topic Clusters: Hub, Spokes, and the Honest Test for a New Page

Two senses of one word, then the structure this page actually means: one hub, distinct spokes.

The idea got popular, then got cargo-culted.

"Build a topic cluster" became "publish thirty posts and link them in a circle" — which produces thirty thin pages, a tidier sitemap, and nothing else.

This is the version without the mysticism. It sits inside our content marketing guide, which is the hub of the cluster you are reading a spoke of — so the worked example below is this website.

Before any of it, one word needs untangling.

First: "pillar" means two different things in marketing

The word carries two unrelated meanings, and people arrive here having read the other one. Both are legitimate. They just collided.

• Content pillars, the social sense. The handful of recurring themes a brand posts about on Instagram or LinkedIn, so the feed stays coherent and the calendar stays fillable. That sense belongs to our guide to content pillars, which is where to go if it is what you came for. We are not going to redefine it here.
• A pillar page, the SEO sense. One page covering a broad topic at a high level, sitting at the centre of a group of narrower pages that link back to it. Also called a hub page. This is the only sense used below.

The tell: if the unit is a theme you post about, it is the social sense. If the unit is a URL other URLs link to, it is the SEO sense.

What a topic cluster actually is

A topic cluster has exactly three parts, and the third one is the part people skip.

1. The hub. One page on the broad topic, useful on its own, linking out to every spoke.
2. The spokes. Narrower pages, each owning one sub-query someone genuinely types, each answering it more completely than the hub could at its width.
3. The links between them. Hub to every spoke, every spoke back to the hub, and spoke to spoke wherever a reader would actually cross over. Without this you have a folder, not a cluster.

HubSpot popularised the model, crediting Anum Hussain and Cambria Davies' 2015 "Topics over keywords" research (HubSpot blog, accessed September 2026) — which is why the vocabulary is marketing vocabulary, not Google's.

The worked example is this site

Our own architecture, not a diagram: Digital Hangover runs this structure across nine live hubs. Click any of them and check — that is the point of using ourselves as the example.

Three more hubs run the same way: link building (one tactic per spoke), social media marketing (one platform), marketing channels (one channel).

Note the analytics row. Two spokes, not twelve. Clusters need not be symmetrical, and forcing symmetry is how thin pages get made.

A cluster is also not a content map. The map — see our content mapping guide — lists questions by persona and stage. The cluster is those rows grouped by topic.

Why it works, mechanically

Clusters work because they resolve two specific failures that afflict libraries built one post at a time. That is the entire mechanism. There is no cluster bonus.

The orphan half has a plain Google statement behind it. Google's Search Central documentation on making links crawlable, last updated 10 December 2025, says: "Every page you care about should have a link from at least one other page on your site." A hub is the cheapest way to satisfy that for a whole topic at once.

Google's SEO Starter Guide, same date, adds that grouping similar topics into directories helps Google learn how often each directory's URLs change, and that links "may help demonstrate your knowledge on a topic". Note the hedge in Google's own sentence — may. We are not going to upgrade it.

The third benefit is editorial: an empty spoke is visible in a way a missing blog post never is.

How to choose the hub topic

Pick a hub you could defend as something your business is actually about. A hub is a commitment to keep publishing under it, so the test is less about volume than about whether you will still care in a year.

• It has real sub-queries. Five or six things people search separately underneath it. Only three, and it is a spoke on somebody else's hub.
• It is commercially connected. There is a service or product at the end of it. Hubs with no destination become traffic you cannot use.
• You can be specific about it. You have examples, client patterns, numbers you may share. Topics you know only generically produce encyclopedia entries.
• It is not already a hub on your own site. Two hubs on overlapping topics is cannibalisation at the architecture level, and far more expensive to unwind than two competing posts.

On width: the hub is the broadest term you could rank for in two to three years, not the broadest that exists. "Marketing" is not a hub. "Content marketing" is, if content is what you sell.

Run a content audit before the architecture — most sites find they already have four accidental half-clusters and one topic with nineteen overlapping posts.

The honest test: own page, or a section?

This is the decision everyone gets wrong, in both directions. Split too eagerly and you manufacture the cannibalisation the cluster was meant to prevent. Split too rarely and you get a 6,000-word hub that ranks for one query and buries fifteen answers.

Four questions. A sub-topic needs yes to at least three before it gets a URL.

1. Do people search for it in words that are not synonyms of the hub? "What is content marketing" and "content marketing meaning" are one query in two coats — one page. "Content brief" and "content audit" are different jobs with different vocabularies — two pages. If every phrasing rewords the hub keyword, you have a section.
2. Does the intent differ from the hub's? The hub usually serves someone learning the subject. A sub-topic serving someone comparing options, or trying to complete a task, is a different job and a different page. Our search intent guide covers how to read that from the SERP rather than guessing.
3. Would a complete answer run past roughly 800 words? Not "could you pad it to 800". If the honest answer finishes in three paragraphs, it is a section with a heading and an anchor. A section that keeps growing tells you when to promote it.
4. Would the two pages want the same title tag? Write both. If you cannot make them meaningfully different, Google will not be able to either, and you have built the cannibalisation problem on purpose.

Three noes means it lives on the hub as a section. Not a demotion — a well-anchored section on a page that already has authority often outranks a new URL with none.

What we see most often in our own audits, stated as a pattern rather than a measured figure because we have not measured it: the splits people regret were made for volume. Someone finds a keyword row with a number beside it, publishes a page, and six months later that page ranks for nothing while the hub ranks slightly worse. The splits people are glad about were made because a section had outgrown its heading.

The habit that prevents it: write the sub-topic's content brief before it gets a URL. If the brief cannot fill itself without repeating the hub, you have your answer — and you spent an hour, not a quarter.

The internal linking rules that make a cluster real

Four rules. A cluster without them is a directory with ambitions.

• The hub links to every spoke, in body copy — not only from a menu or a grid of cards. A sentence explaining why you would click does more than a tile.
• Every spoke links back to the hub, high up — first two or three paragraphs, anchor text naming the hub topic. This page does it in its third paragraph.
• Spokes link to each other only where a reader would cross over. Sharing a decision, yes. Merely sharing a folder, no — manufactured cross-links are the commonest cargo-cult symptom.
• Anchor text describes the destination. "Our content audit guide" beats "click here". Google's SEO Starter Guide makes the same point about descriptive link text.

The mechanics belong elsewhere: anchor text patterns, click depth and how link value moves around a site are in our internal linking guide, which owns the topic. This page only specifies which links a cluster requires.

One caveat, because it is where clusters silently rot: publish a tenth spoke and the hub needs editing the same day, or the new page is an orphan the moment it goes live.

If that is where the plan usually stalls, it is the work our content marketing services do. Directionally, content marketing in India runs about ₹25,000 to ₹1,50,000 a month depending on volume and depth — a market range, not a quote.

Topical authority: what it plausibly means, and what it is not

"Topical authority" is useful shorthand for a plausible idea: a site that covers a subject thoroughly, and is linked to and cited on it, tends to do better on queries in that subject than a site with one post. That much is reasonable.

What it is not is a Google metric. There is no published score, no Search Console report, and no number anyone can show you. Two things worth having straight:

• Google's guide to Google Search ranking systems, last updated 10 December 2025, names the systems Google will name — BERT, RankBrain, neural matching, link analysis and PageRank, freshness and passage ranking among them. Neither "topical authority" nor "topic cluster" appears. That is an absence from a document, not a denial — but it means nobody can point you at the setting.
• Search Engine Journal reported on 13 November 2023 that Google's John Mueller, replying on Reddit, agreed with a description of topical authority as essentially a rebrand of relevance and authority, and told people not to worry about it.

Use the phrase if it helps you get budget approved. Do not build a plan whose success criterion is a number that does not exist. And note the inversion it invites: forty pages create no authority if thirty are sections promoted to URLs. Coverage means answering questions, not page count.

How to judge a cluster, and on what timeframe

Judge a cluster on four things, in this order, and give it two to three quarters. Organic results from new or restructured content typically take about three to six months to become meaningful, and clusters are slower than single pages because the internal links need crawling and recrawling.

1. Coverage. How many of the sub-queries you identified now have a page that clearly owns them? The only metric you control outright, and available on day one.
2. Query separation. In Search Console, does each spoke have a distinct set of queries, or are two spokes showing impressions for the same one? Overlap is the early warning; merging fixes it cheaply.
3. Hub performance on the broad term. The slowest thing to move. Check the trend is up, not that a position was hit.
4. Enquiries, by entry page. Spokes answering decision-stage questions often convert long before they rank well. Judge those on enquiries, or you will kill your best pages for looking quiet in a tracker.

What not to judge it on: page count, average words per page, or whether the diagram looks symmetrical. Review quarterly — re-check which spoke owns which query, merge anything that has drifted into overlap, and promote any section that has outgrown its heading.