The structure, in one paragraph
One pillar page covers the topic broadly and links out to every supporting page. Each supporting page — a spoke — covers one narrow slice properly, and links back up to the pillar. That's the whole model. Hub and spoke, borrowed from airline route maps, which is a better metaphor than it sounds: the hub exists so the spokes connect.
What the structure buys you is threefold. Each page gets to match one query type precisely instead of hedging. Internal links concentrate rather than scatter. And a search engine reading the cluster sees a site that covers the subject rather than a site that once wrote about it — which is the practical route to topical authority.
The planning step everyone skips
Most cluster plans are a list of twelve titles. Titles aren't a plan, because two titles can describe the same search. "Dental implant cost" and "Are dental implants worth the money" look like different articles and are frequently the same query with different words on it.
So write the intent, not the title. One sentence per page, describing the person typing and what they need to leave with. If two sentences describe the same person in the same moment, you have eleven pages, not twelve. Do this before a single brief goes out and you've prevented the expensive version of the problem.
Here's what a real one looks like — a Bangalore dental clinic building out implants.
| Page | The intent it owns |
|---|---|
| Dental implants in Bangalore (pillar) | Broad commercial — someone choosing where to get it done |
| What is a dental implant | Informational — first-time researcher, no vocabulary yet |
| Dental implant cost in Bangalore | Commercial investigation — comparing prices before enquiring |
| Single tooth implant | Product-specific — one missing tooth, wants that case explained |
| Full mouth implants and All-on-4 | Product-specific — complete restoration, much higher value |
| Implants vs bridges | Comparison — deciding between two treatments a dentist offered |
| Implants vs dentures | Comparison — older patient, entirely different objections |
| The implant procedure, step by step | Reassurance — anxious patient wanting the sequence and the timeline |
| Recovery and aftercare | Post-decision — already booked, searching at 11pm |
| Am I a candidate for implants? | Qualification — bone density, diabetes, smoking, age |
| Implant cost, insurance and EMI in India | Financial objection — can afford it monthly, not upfront |
| Implant brands and materials | Technical comparison — informed buyer checking what you use |
The linking pattern
Links are what convert twelve files into a cluster. The pattern is simple enough to hand to a writer as a rule.
- Every spoke links up to the pillar, once, with a consistent anchor. That's the vote that says which page is the hub.
- The pillar links down to every spoke, from the section that summarises it, using the spoke's own subject as the anchor — "implant recovery timeline", not "dental implants".
- Spokes link sideways only where a reader would actually want it. Cost to EMI, yes. Cost to recovery, no.
- Never link two spokes that compete for the same query family. If you find yourself wanting to, that's the two-page test failing late.
- Fix the anchors when you consolidate. Redirecting a merged page and leaving forty internal links pointing at the old URL wastes the consolidation you just paid for.
Size, order, and why this stops cannibalisation
Six to fifteen pages is the workable range. Fewer than six and the pillar has nothing to route to. More than twenty and it's usually two subjects sharing a folder — split it, and give each half its own hub.
On order, we go against the common advice. Most plans build the pillar first, which produces a hub page linking to nothing for six weeks. Build two or three of the clearly commercial spokes first — the cost page, the main comparison, the highest-value treatment — then the pillar, then the informational spokes. The commercial pages pay for the rest of the cluster while it's being built, and the pillar launches into a cluster that already exists.
The cannibalisation argument is the real reason to do any of this. When two of your pages target one query, Google has to choose. It alternates between them, neither settles, your internal links split across both, and any external link you earn lands on whichever one the writer happened to find. One intent per page removes the choice entirely — see keyword cannibalisation for what the alternative costs, and our post-mortem for what unwinding it looks like.
The cluster also fails quietly if nobody owns it after launch. Six months in, someone writes a new post that overlaps with page four, because the plan lives in a spreadsheet nobody opens. Put the intent sentence in the CMS next to the page, where the next writer will actually see it.