Topic ClustersOwning a subject, not a keyword

From keywords to subjects#

Early SEO treated each keyword as a separate target: one page per phrase. Modern search systems understand topics, entities and the relationships between them. They are more likely to trust a site that covers a subject thoroughly and coherently than one with a scattering of unrelated articles.

Topic clusters are the editorial response. Instead of asking 'which keyword next?', you ask 'which subjects do we want to be known for, and what does complete coverage of each look like?' The cluster is the unit of planning, publishing and measurement.

Choose fewer subjects and cover them completely; breadth without depth reads as noise.

Anatomy of a cluster#

The pillar page answers the broad query and gives a map of the subject. It should be genuinely useful on its own, not a list of links, but it deliberately does not try to answer every sub-question fully. Its job is orientation.

Cluster pages each own one specific intent: a how-to, a comparison, a definition, a checklist. Together they cover the questions people ask on the way to a decision. Each links up to the pillar and across to siblings where a reader would naturally go next.

Choosing subjects you can actually own#

The commonest mistake is picking subjects that are too broad. 'Marketing' is not a cluster; it is a library. 'Email deliverability for D2C brands' might be. A good cluster subject sits where three things overlap: real search demand, genuine expertise in your organisation, and a clear link to what you sell.

Size matters too. A cluster that needs a hundred pages before it is complete will take years and may never reach coherence. Aim for subjects where perhaps eight to twenty cluster pages would cover the main questions. You can always add a neighbouring cluster later.

Mapping the cluster#

Use keyword research to list every question in the subject, then group them by intent. Each group that needs its own answer becomes a cluster page. Groups that can be answered in a paragraph become sections of the pillar or of a related cluster page.

Check each proposed page against existing content. If something already covers that intent, improve it and fold it into the cluster rather than publishing a competitor to it. Clusters built on top of unresolved cannibalisation inherit the problem.

Linking the cluster together#

Links are what turn a set of articles into a cluster. Every cluster page links to the pillar with descriptive anchor text. The pillar links to every cluster page, usually in context within its sections rather than only in a list at the end. Sibling pages link to each other where the reader's next question lives on the sibling.

Then connect the cluster to commerce. Practical and evaluative pages should link to the relevant service, product or demo page where it genuinely helps. A cluster that never links to anything you sell is a library with no exit. See internal linking for placement and anchor text.

Publishing sequence#

The timeline is illustrative; your pace depends on resources. The principle is to complete one cluster to a useful level before starting the next. Five half-built clusters demonstrate less expertise than one complete one.

Quality control for clusters#

Common ways clusters go wrong#

• Thin spokes: cluster pages written to fill a map rather than answer a real question. If a sub-question can be answered in two sentences, it belongs inside another page.
• A pillar that tries to do everything: a ten-thousand-word page that competes with its own cluster pages for the same queries.
• Links only in a box: a 'related articles' list at the bottom is weaker than links placed where the reader's question arises.
• No owner: clusters drift out of date when nobody is responsible for the subject as a whole.
• Starting too many: several clusters at twenty per cent completion demonstrate little expertise in any of them.

Each of these has the same root: treating the cluster as a structure to fill rather than a subject to explain. The best test is to imagine a newcomer to the subject arriving at the pillar. Could they, by following its links, become reasonably competent? If so, the cluster works.

Assign each cluster a named owner, ideally someone with real expertise in the subject, who reviews it at least twice a year. Their job is to spot outdated advice, missing questions and opportunities to add first-hand examples, which keeps the cluster worth citing as the subject evolves.

Measuring a cluster#

Report on the cluster as a whole. Combined impressions and clicks across all its pages, the number of distinct queries it ranks for, and conversions from sessions that touched any cluster page tell you whether you are becoming the reference on the subject. Individual pages will rise and fall; the cluster trend is what matters.

Use the cluster's own data to plan its next pages. Queries in Search Console that the cluster appears for but no page directly answers are the clearest signal of a missing cluster page or section. Over time, the cluster tells you what it needs.

Clusters also matter for AI search. Systems that summarise answers favour sources that cover a subject clearly and consistently, which is exactly what a good cluster provides. Read generative engine optimisation for that angle, and programmatic SEO if your cluster includes many templated pages.

Published by Fabulous.Media, a network of specialist marketing agencies. Updated 9 October 2026. Platform features change often; check current official documentation before acting on platform-specific detail.