How-To Guides
How to Build a Topic Cluster With AI (and What Google Actually Indexed)
I run two different sites on the same hosting, but when I checked my Search Console in November 2026, the results were quite different. One was this site, ArticleDojo, whose older Dojo Vault articles were broad advice on topics thousands of other sites already cover and linked mostly to a general library page rather than to each other by topic. Only 2 of those 21 were indexed in November 2026.
The other site, a self-improvement audio app I run called SelfRecode, had a different relationship with the search index. That site uses a specific content structure built around a topic cluster model to organize its library of ideas. I had published 28 pages in total, including various concept articles and hub pages. When I checked the index status for those pages, 25 of them were already showing up in Google. Most had been crawled on November 3 or 4.
The Gap Between Discovery and Indexing
You might think that if a search engine finds your page, it will put it in the database. That is not how it works. When you see a status like "Crawled, currently not indexed" in your reports, it means the bot read the text but decided there was not enough value to include it in the public results. It is a judgment call made by the engine. A page can be technically fine and still not be indexed, and maybe weak connections to related pages are part of the reason.
This is where the idea of a topic cluster comes into play. It is about how those pages talk to each other through links. If you have a single post that exists in a vacuum, Google might see it as a random piece of content. But if that post is part of a network, the context changes. Links can help a search engine see how one idea connects to another.
I cannot be certain that my linking structure was the only reason for the difference in indexing between my two sites. The topics and the writing quality were different too, so it is not a controlled test. However, the difference in how many pages were accepted suggests that structure plays a role.
How a Hub and Spoke System Works
The system relies on two parts: the hub and the spoke. A hub is a broad, high-level page that covers an entire category or theme. It does not try to answer every tiny question, but instead provides a high-level overview and directs people toward more specific information. It acts like a map for a certain subject.
A spoke is a much narrower piece of writing that focuses on one specific question or single concept within that broader category. You do not want a spoke to be too broad, or it becomes its own hub. If you try to cover too much in one article, you lose the benefit of having many specialized pages. Instead, each spoke should be a deep dive into one small corner of the main topic.
Links are the glue in this system. A spoke page must link back to its parent hub so that the connection is clear. You also have to connect your spokes to each other where it makes sense. This creates a network of information that makes the relationship between pages clear to readers and search engines.
The Way I Built My Library
For my project with SelfRecode, the structure was very strict. I had a library index page that acted as a starting point for everything. From there, I had six topic hubs like Habits, Identity and Mindset. Those hubs linked down to the individual concept articles that made up those topics.
Each concept page covered one specific, named idea and was written from a brief of checked facts, then reviewed before publishing. I made sure that each page linked back up to its topic hub and also back to the main library index. I also included sideways links between the concept pages, connecting them to three other related ideas within the same category. This created a network of connections that prevented any single page from being an isolated piece of content.
The publishing process was not a single event. I did not dump everything onto the site at once. Instead, I published in batches. The first pages went live in late September, more followed in October, and a batch of 17 went live on October 31, 2026.
Steps to Build Your Own Cluster
You can try this yourself without a complex system if you have the time to do the manual work. First, you must pick a topic that is deep enough to support many different angles. If your topic is too thin, you will run out of things to say after two pages. You need a subject that has many different sub-questions people ask about it.
Once you have the topic, write your hub page first. This should be an overview that explains what the category is about and why it matters. Do not worry about being too specific here, as that is what the other pages are for. You need to be careful about how you handle the research and editing process, as quality is just as important as structure. Before you publish each page, check every claim against your facts, and link to How to Edit AI-Generated Articles: 35 Real Fixes From 5 Drafts.
Do not write the spokes yet. Instead, make a list of specific questions that each spoke will answer. Each question must be narrow enough to be covered in a single article. Once you have that list, write the articles one by one. Avoid writing generic advice that could apply to any topic. Focus on being specific and useful for a very narrow audience.
Connecting the Dots
The most important part of this entire process is the linking strategy. You must ensure that every single spoke page has a clear path back to its hub. If you forget this, the chain breaks and you lose the benefit of the cluster. You should avoid having links that only move in one direction without any way for a user or a bot to find their way back to the main topic.
You also need to implement lateral links. This means if you are writing about a specific habit, and it relates to a specific mindset concept, you should link between them. This keeps readers on your site and shows the search engine that your knowledge is interconnected. It can be time-consuming to do this manually for every single post, especially if you are scaling up.
If you have hundreds of ideas, the manual work becomes a lot. While some people prefer to build every link by hand, there are more efficient ways to handle the structure. Our tool, the ArticleDojo Content Cluster Builder, is included on every paid plan starting at $29 a month. You enter a niche and optionally your domain, then the AI maps 2 to 10 pillars with a hub guide each and plans the spokes. A 2x2 test results in 6 articles, the default 5x5 is 30, and you can go up to 110. Every spoke links back to its hub and related spokes cross-link, which is the same up, down, and sideways linking described above. Setting your domain pre-fills hub URLs and a Find Positions button inserts links into the text, and you can add more spokes later. You can export a ZIP with a WordPress XML file or publish straight to WordPress as drafts, scheduled posts, or live posts.
Avoiding Common Mistakes
One mistake people make is trying to write one giant article instead of many small ones. It might seem easier to just put everything in one place, but that makes it harder for a specific query to find your page. You want to be the answer to a very specific question, not just a general source of information.
Another mistake is being too broad with your topics. If a page feels like it could belong to any site on the internet, Google might decide it is not valuable enough to index. You need to provide specific details that only your site provides. This does not mean you have to invent new facts, but it means you should apply those facts to specific problems.
You must also be careful about how much of your site is just filler. If you have a cluster that does not provide real value, the structure will only take you so far. The goal is to build a library that actually helps people, not just a collection of links. If you are weighing tools for this, the ArticleDojo Alternatives: Who Should Use Something Else page compares several.
Will you prioritize the structure of your site or just keep posting one-off articles?
How this article was made: drafted by a local AI model from a fact-checked brief, then reviewed and edited by Andy C. Prices and features were checked against each tool's own site on the dates stated. ArticleDojo is Andy's product.