What a topic cluster is
The model has three parts and a rule. The pillar page covers the whole subject at the depth the broad query wants and introduces each part. The cluster pages each take one part and cover it fully, owning the specific query for that part. The links connect pillar to every cluster page, every cluster page back to the pillar, and cluster pages to each other where their content meets. The rule is that every page in the cluster owns exactly one query and that no two pages in it compete for the same one.
Why the structure works is documented only in parts. Google's SEO Starter Guide, the source cited on this page, describes links as the main way its crawler discovers pages, anchor textThe visible, clickable words of a link. as a signal of what a linked page is about, and a logical organisation of a site's pages as something that helps both people and the search engine understand how they relate. Google does not document a ranking factor called topic clusters; the model is an industry pattern that arranges content so those documented behaviours work in its favour. Describing the result as topical authority is shorthand, and the mechanism beneath it is relevance, coverage and link structure, which is the part of SEOImproving visibility in unpaid search results. this framework addresses.
As a thinking tool, the cluster makes gaps and orphans visible. Laying a subject out as a pillar and its parts shows which parts have no page, and listing the site's existing pages against the cluster shows which pages belong to no subject. What it hides is demand and quality. A beautifully structured cluster around a subject nobody searches for, or one whose pages are thin, is still a well-organised failure.
Pillar and cluster pages
| Attribute | Pillar page | Cluster page |
|---|---|---|
| Query it owns | The broad head term for the subject | One specific sub-topic query |
| Depth | Broad: every part introduced, none exhausted | Deep: one part covered completely |
| Structure | A section per part, each ending in a link to the cluster page for it | The format the sub-topic's results page demands: guide, comparison, tutorial, list |
| Links out | To every cluster page, in context | To the pillar with the subject as anchor; to sibling pages where content meets; to the next-step page |
| Links in | From every cluster page, from the site's navigation, from older content that mentions the subject | From the pillar and from the siblings that relate |
| Role in the journey | Entry point for people new to the subject; where external references land | Where a specific question is answered and where most conversions begin |
| Update rhythm | Each time a cluster page is added or retired | When its results page or its subject changes |
The pillar is not a table of contents. A pillar that only lists its cluster pages ranks like a category page, because it is one. It explains the subject well enough that someone who reads nothing else leaves informed, and it links onward for those who want depth.
Choosing cluster topics
Cluster pages are chosen by four tests, applied to each candidate sub-topic in turn.
Demand: people search for it, in enough volume to justify a page, in words the research has recorded. Candidate sub-topics usually come from keyword research grouped into clusters, and a sub-topic with no queries behind it belongs in a section of the pillar rather than on a page of its own.
Distinct intent: the results page for the candidate query (SERPThe page of results a search engine shows.) differs from the pillar's and from every sibling's. If two candidate queries return substantially the same pages, they are one sub-topic and get one cluster page; if a candidate query returns the same pages as the head term, the pillar already owns it. This test is what stops a cluster from cannibalising itself.
Adjacency: the sub-topic sits close enough to what the business sells, or to the step before buying it, that a page on it can lead somewhere. A cluster can carry a few purely informational pages that earn links and trust, but a cluster made entirely of them educates and never converts.
Credibility and capacity: the business can write the page with genuine expertise, and the team can produce and maintain it. A sub-topic that passes the first three tests and fails this one is deferred, not forced.
The cluster framework sits beneath two broader decisions. The content strategy framework decides whether the subject is a theme the brand should own at all, and the content pillars and calendar guide covers how an editorial pillar is scheduled; the cluster is the search-specific shape one of those pillars takes.
Internal linking rules
The links are what turn a set of pages into a cluster, and they follow rules rather than instinct.
- The pillar links to every cluster page, from the section that introduces that part, with a sentence of context rather than a bare list. A cluster page the pillar does not link to is not in the cluster.
- Every cluster page links to the pillar, with anchor text that names the subject, usually in the opening section where the reader is being placed in context.
- Sibling pages link to each other where the content meets, and only there. Forced links between unrelated siblings add nothing for the reader and dilute the signal for the related ones.
- Anchor text describes the destination. The starter guide's advice on descriptive anchors applies with particular force inside a cluster, where the anchors are the clearest statement of how the pages relate.
- Every page has a next step. The pillar and the cluster pages closest to a purchase link to the product, service or contact page the subject leads to, so the cluster has somewhere to send people.
- No page in the cluster is reachable only through the cluster. Navigation, hub pages or older content should link to the pillar at least, so that the whole cluster does not depend on one path.
The internal linking guide covers the mechanics: anchor choice, link placement, and the audit that finds pages the rules have missed.
Worked cluster plan
A cluster plan for a commercial coffee machine supplier
A supplier of commercial coffee machines plans a cluster around the subject "commercial coffee machines". The pillar is a buyer's guide owning the head term, which the research shows at roughly 3,600 searches a month. Eight cluster pages pass the four tests: bean-to-cup versus traditional espresso (1,900 a month), machines for small cafés (880), machines for offices (1,300), leasing versus buying (720), maintenance and servicing (590), water filtration (480), how many group heads a café needs (320) and running costs (410). Two candidates fail: "espresso machine history" has demand but no adjacency, and "best coffee beans for offices" returns the same results page as an existing page in a different cluster.
The next-step page is the product range, linked from the pillar and from the leasing and small-café pages. Build order: the pillar in month 1, then 3 cluster pages a month for 3 months, closest-to-purchase first. Targets at 6 months: the pillar in the top 5 for the head term, cluster clicks rising from 300 to 2,400 a month, and 40 quote requests a month where a cluster page appears in the journey, against 9 today.
The plan fits on a page and already contains the three decisions that matter most: which candidates were rejected and why, where the cluster leads, and the order in which it is built. A plan that lists pages without those decisions is a list of titles.
Measuring a cluster
A cluster is measured as a unit, with one headline KPIA metric tied directly to an objective. and a few diagnostics, because single pages within it fluctuate for reasons that have nothing to do with the structure. The headline is the outcome the cluster was built for: enquiries, quote requests, trials or purchases where a cluster page appears in the journey. Beneath it sit the diagnostics. Coverage: the share of the planned cluster pages that are published and linked according to the rules. Position: the pillar's ranking for the head term and each cluster page's ranking for its own query. Volume: total impressions and clicks across the cluster, and the share of those clicks each page takes, which shows whether the pillar is drawing people in or the cluster pages are doing all the work. Link completeness: a periodic check that every cluster page still links to the pillar and its siblings and that the pillar links to every page, since site changes break links quietly.
The reading that matters is the trend of the unit against the plan, not the position of any one page in any one week. The SEO strategy framework places cluster measurement within the wider set of search metrics, where it sits beside technical health and authority.
Limits
The framework assumes the subject has search demand and the business has something to say about it, and it cannot test either assumption. A cluster planned from a product list rather than from research will be structurally perfect and empty.
It can manufacture thin pages. The pressure to fill a cluster produces pages for sub-topics that did not pass the intent test, and those pages compete with each other and with the pillar. The distinct-intent test is the guard, applied honestly.
Pillars bloat. As cluster pages accumulate, the pillar is extended to introduce each one until it tries to be every page at once; it needs editing back to breadth as the cluster grows.
Results pages change. Two sub-topics the engine once served separately can merge into one, or one can split, and a cluster planned against last year's results pages drifts out of alignment with this year's.
The model is a content structure, not a technical one. URL hierarchy, navigation and site architecture can match it or not, and a cluster spread across a blog and a resources section works as long as the links hold, but the mismatch confuses teams that expect the folder to be the cluster.
Finally, what the model owes to the engine is partly undocumented. Google describes links, anchors and site organisation; it does not describe clusters, and claims about how much a cluster lifts rankings are practitioner observation rather than documented behaviour.
Related frameworks
| Framework | Reach for it when |
|---|---|
| Content strategy | The question is whether the subject is a theme worth owning at all |
| SEO strategy | The cluster needs to sit within technical, authority and measurement work |
| See-Think-Do-Care | Cluster pages need assigning to audience intent stages so the next step fits each |
| Measurement plan | The cluster's headline KPI needs event definitions and an owner |
Common mistakes
- A pillar that is a list. Links without explanation rank as a category page and pass little to the cluster.
- Cluster pages chosen from the product catalogue. Sub-topics people search for rarely match the way a business names its products.
- Skipping the intent test. Two pages for one results page cannibalise each other inside the cluster.
- Links left to chance. Pages published without the pillar and sibling links are an archive, not a cluster.
- No next step. A cluster that educates and leads nowhere earns trust the business never collects.
- Measuring single pages. Judging the cluster by one page's weekly position hides whether the unit is working.
- Letting the pillar absorb its cluster. A pillar extended with every new part becomes too deep for the broad query and competes with its own pages.
