Topic clusters are a way to organise a website by subject: one pillar page covers a topic broadly and links to every article that answers one specific question in depth, and each of those articles links back to the pillar. To build one, pick a topic big enough for many articles but not sprawling, collect the real questions, give each question one page, write the pillar page as an ordered overview, link by a fixed rule and check after launch that the links exist. The structure guarantees no rankings. But Google uses links to find pages and understand them, and a cluster supplies them by plan instead of by chance.
My own website is built this way, with overview pages such as AI Basics: AI Explained Simply, Understanding and Using Claude Right, Kubernetes and AI SEO for Google and AI Search. I built the structure together with Sebastian, my AI employee for SEO and GEO (AI for SEO and Blogging: How Sebastian Works). The basics every site needs before clusters are in SEO Basics for Small Businesses.
What topic clusters and pillar pages are
Three page types get mixed up:
HubSpot described on its blog how it reorganised this way: away from separate posts for separate long-tail keywords, towards topic areas held together by one overview page. A content strategy is the bigger frame (who you write for, about what, to what end); the cluster is how you arrange the content on the site.
What Google says, and what it doesn't
The term "pillar page" does not appear in Google's SEO Starter Guide or its pages on links, helpful content and spam (as of November 2026). What is there fits the model anyway:
- Links get pages found. According to the SEO Starter Guide, the vast majority of new pages Google finds every day are found through links, and "every page you care about should have a link from at least one other page on your site."
- Anchor text matters. Good anchor text, says Google, is descriptive, reasonably concise and relevant to both pages; "Click here" is its example of bad.
- Folders are optional. Topic directories can help crawling once a site has more than a few thousand URLs; a small site needn't restructure.
- Duplicates confuse. Two pages with the same information aren't spam, says Google, but readers may wonder which is right.
- Volume is a warning sign. Generating many pages with AI without adding value for users may violate Google's spam policy on scaled content abuse. More in Programmatic SEO with AI Agents.
For AI Overviews and AI Mode, Google lists no additional requirements; the SEO fundamentals apply, internal links included. Both may use "query fan-out", related searches across subtopics. That a cluster gets cited more often because of it is a hypothesis, not evidence. What is documented for AI search is in What Is Generative Engine Optimization?, and the practical side in Optimizing Your Website for AI Search.
How to build a topic cluster in six steps
1. Scope the topic
HubSpot's yardstick: topics broad enough for 20 to 30 posts, but not so broad that the pillar page can't go into depth; not "Sales", but "Sales Email Templates". An example for a small business: a painting company builds no cluster on "renovation" but one on "exterior house painting", with articles on cost, paint, weather and scaffolding.
2. Collect the real questions
Four sources:
- Google's suggestions as you type, and the "People also ask" box.
- Search Console: the queries your pages already appear for (Setting Up Google Search Console).
- Your customers: questions from emails and calls, word for word.
- A keyword tool for search volume, which Search Console does not show.
In October 2026 I connected SE Ranking* to Sebastian. It was strong on volume and competitors; for my own positions, Search Console remains the source. The test is in SE Ranking Review: API, MCP and Real Numbers. Upfront: I have a partnership with SE Ranking and earn a commission through my link.
* Affiliate link: if you subscribe or buy credits through this link, I earn a commission. Your price stays the same. I have a partnership with SE Ranking.
3. One question, one page
Put your question list next to your existing pages. If a page already answers a question, expand it instead of writing a second. It happens easily: on my site, Sebastian's check on 19 September 2026 found that of 4,770 search queries, 205 pointed to more than one of my pages. So map it out in a table before writing, here for the painting company:
If a search term shows up twice, you have planned a duplicate. And not every phrasing needs a page: building separate pages for every search variation primarily to influence rankings or AI answers violates Google's scaled content abuse policy.
4. Write the pillar page
A pillar page has five building blocks:
- The first paragraph says what the topic is and who it's for, with the core answer.
- The subtopics in order, as groups or a staged learning path, one sentence and link per article.
- Short answers to basic questions, without duplicating the articles.
- A next step: an offer, a contact page or a newsletter.
- The complete list of articles on the topic, so new ones get a link right away.
5. Fix a linking rule
I call pillar pages hub pages, and my rule is: every hub page links its posts, every post links its hub and two to four neighbours, and almost every post ends at a next step, usually the community, the Kubernetes posts at the book. Put each link in a sentence that works without it, with the target's title or a natural description as anchor text. How to find missing links and pages with no incoming link is in my article on internal linking.
6. Check that the links are live
A linking plan on paper is not a cluster. In September 2026 our analysis found three hubs without posts and four posts without an incoming link. So Sebastian checks the real links after every release, not just the plan. For you: crawl after launch, count incoming links per article, and check in Search Console which page shows for which query. Two pages on one query? Back to step 3. How such a check runs with AI is in How to Run an SEO Audit With AI. Whether your cluster shows up in ChatGPT, Perplexity and AI Overviews is measured separately, see AI visibility tracking.
Pillar page examples: how kevinwelter.com is built
My overview pages differ by reader:
All articles sit flat under /blog/; the overview page and the links decide the topic, not the folder.
It grew in stages. From 27 August to 21 September 2026, Sebastian and I took the site from 10 to 224 posts per language: first a hub-and-spoke structure with three hub pages, then a linking plan, later the AI employees' pages as hubs of their own. Search Console showed 1,060 clicks through 17 September 2026, against 5 in August before the start. Structure and posts grew together, so the structure's share can't be isolated, and it is no promise for another site.
Where Sebastian comes in
Sebastian plans the hub structure and linking plan and writes the articles in German and English, search phrase from real Google suggestions, answer first, checked by a test script and an editor in a separate role. I proofread and approve; nothing goes live without me. No ranking promises, and every figure comes from a saved tool result. More on AI for SEO and Blogging: How Sebastian Works.
Frequently asked questions
How long should a pillar page be?
No set length; Google says it has no preferred word count. It is long enough when the basic questions are answered and each cluster article has a sentence and a link.
How many articles does a topic cluster need?
HubSpot scoped its topics for 20 to 30 posts. A small site can start with the pillar page and the three to five most common questions.
Do I need to put cluster URLs into folders?
No. Per Google, folders can help crawling only beyond several thousand URLs. Restructuring costs redirects; the cluster comes from the links.
Do topic clusters help with ChatGPT and Google's AI Overviews?
Not proven. Google says the normal SEO fundamentals apply, internal links included. Whether you get cited, you measure separately.
Can I have AI write a topic cluster?
Yes, if someone supplies the facts, each page answers its own question and everything is checked before launch. Many AI pages without added value are Google's own example of spam.
Where to go from here
Take your most important topic, write the question table from step 3 and compare it with your pages. Then you know whether the pillar page or merging duplicates comes first. Every article on search engines and AI search is on AI SEO for Google and AI Search. How to add and validate structured data is in my article on schema markup. And what an llms.txt does and doesn't do is in llms.txt: Hype vs. Evidence.
To see each week how I run my AI employees, pick a resource on my free templates page and sign up there for my newsletter "From the Engine Room". If you'd rather go through your website with me, book a strategy session.
Your first AI employee up and running within an hour
Join the community$5/month founding price, first 50 members
