A topic cluster is a set of pages on one subject where each page answers a distinct question and the pages link to each other. The bottom line: build clusters around the questions customers ask, give each page its own angle and evidence, link them deliberately, and measure share of answers for the cluster's questions, because the benefit for AI citations is plausible and unproven.
See also: answer engine optimization, explained step by step.
What a cluster contains
| Page type | Job | Example for a commercial roofing contractor |
|---|---|---|
| Pillar | Explains the whole subject and links to every cluster page | Flat roof coating: what it is, when to use it, cost, and lifespan |
| Definition pages | Answer what is questions | What is a fluid applied roof coating |
| Decision pages | Compare options | Silicone versus acrylic coating for a low slope warehouse roof |
| Cost pages | State price ranges and drivers | What roof coating costs per square foot, with the factors that move it |
| Process pages | Explain how the work is done | How a roof coating job is scheduled, prepped, and cured |
| Proof pages | Show real projects | Case page with materials, square footage, and county |
Steps
- Choose one subject. One you must be known for.
- Collect 20 to 40 real questions. From sales calls, support email, Search Console, and the follow up questions AI suggests.
- Group them by intent. Define, compare, price, how to, and proof.
- Assign one question to one page. Two pages that answer the same question compete with each other.
- Write each page answer first, with its own evidence. Numbers, sources, and dates that differ from the other pages.
- Link deliberately. The pillar links to each page, each page links back and to two or three related pages, with descriptive anchor text.
- Publish the pillar and the highest value pages first. Fill the rest over weeks.
- Measure. Run the cluster's questions before publishing and monthly after, and compare shares.
What research and documentation support
- Google's AI features expand queries into related searches, so a cluster that covers sub questions gives more retrieval chances. This is a mechanism, not a measured effect of clusters.
- Google warns against scaled content created mainly to manipulate rankings or AI answers, so each page must have a reason to exist.
- Links are how crawlers discover pages, and Google's guidance lists internal links among the fundamentals.
Common mistakes
- Near duplicate pages that differ by one keyword.
- A pillar with no links to the cluster.
- Filling a cluster with thin posts to hit a page count.
- Never measuring whether the cluster changed your share of answers.
Questions people ask
What is a topic cluster?
A set of pages on one subject where each page answers a distinct question and the pages link to each other, usually around a pillar page.
How many pages should a cluster have?
There is no evidenced number. Cover the real questions in the subject, with each page earning its place.
Do topic clusters improve AI citations?
No controlled study shows it. The mechanism is plausible because AI features expand questions into related searches, so measure the result.
How should cluster pages link to each other?
The pillar links to each page, and each page links back to the pillar and to two or three related pages with descriptive anchor text.