Keyword clustering is the step between keyword research and writing. You have a spreadsheet with dozens (or thousands) of keywords, and you need to decide which ones belong on the same page and which ones need a page of their own.
Get it wrong in one direction and you write five articles that compete with each other (keyword cannibalization). Get it wrong in the other and you cram three different questions into one page that answers none of them well.
In this guide I show the method I use, with a real example: 14 keywords about keyword research that I clustered using live Google results in September 2026. Some of the results surprised me, and they are a good lesson in why you should not cluster by gut feeling.
What is keyword clustering?
Keyword clustering is the process of grouping keywords that can be answered by the same page. Each group (cluster) becomes one URL: one main keyword plus the variations and related questions that page should also rank for.
It is not the same as a topic cluster. A topic cluster is a group of pages: a broad hub page plus several supporting pages that link to each other, which is how you build topical authority. Keyword clustering happens one level down. It decides what each of those pages targets.
Why keywords that look similar may need different pages
The obvious way to cluster is by meaning: “keyword clustering” and “keyword grouping” mean the same thing, so they go on the same page. The problem is that Google does not always agree with you, and Google is the one deciding which page ranks.
That is why the most reliable method compares the search results themselves. If Google shows many of the same URLs for two keywords, it considers them the same need, and one page can rank for both. If the results are different, it sees two different needs, even when the words look alike.
How to cluster keywords with SERP overlap
Step 1: Start with a clean keyword list
Pull your keywords from your research tool and remove the obvious junk: misspellings you will never target, brand names you cannot rank for, and keywords with a different intent altogether. If you need a refresher on this part, see my niche keyword research guide.
Step 2: Get the top 10 results for every keyword
For each keyword, record the URLs ranking in the top 10 for the country you target. You can do this by hand for a small list (search in a private window, copy the URLs) or with a SERP API or clustering tool for bigger ones. Use the full URL, not just the domain: a big site can rank with two different articles for two keywords, which does not mean those keywords share a page.
Step 3: Count shared URLs for each pair
Compare the lists two by two and count how many URLs they have in common. Most clustering tools use a threshold of 3 or 4 shared URLs out of 10. I use 3: if two keywords share three or more ranking URLs, they go on the same page.
Step 4: Name each cluster and pick the main keyword
Within each cluster, the keyword with the highest volume (or the clearest intent) becomes the page’s main keyword, used in the title and H1. The others become subheadings, FAQ questions or natural variations in the text.
Step 5: Map clusters to existing pages
Before writing anything, check whether a page on your site already targets each cluster. If it does, the cluster tells you what to add to that page, not what new page to create. This step alone prevents most cannibalization.
A real example: 14 keywords, 9 pages
To show how this works in practice, I took 14 keywords I was considering for this blog, all related to keyword research. I pulled the top 10 results for each one on Google US in September 2026, kept the organic web pages (videos and other features out, which left 7 to 9 URLs per keyword) and counted the shared URLs for every pair.
Here is what the overlap looked like for the most interesting pairs:
| Keyword pair | Shared top-10 URLs | Decision |
|---|---|---|
| long tail keywords + what are long tail keywords | 6 | Same page |
| what are long tail keywords + long tail keywords examples | 4 | Same page |
| long tail keywords + long tail keywords examples | 3 | Same page |
| search intent + types of search intent | 6 | Same page |
| low competition keywords + how to find low competition keywords | 5 | Same page |
| keyword difficulty + what is keyword difficulty | 4 | Same page |
| keyword clustering + keyword cluster tool | 2 | Separate pages |
| keyword clustering + keyword grouping | 0 | Separate pages |
| topic clusters + pillar page | 0 | Separate pages |
| long tail keywords + low competition keywords | 0 | Separate pages |
And these are the clusters that came out of it, with US monthly search volume for each keyword:
| Cluster | Keywords | US searches / month | Cluster total |
|---|---|---|---|
| Long-tail keywords | long tail keywords what are long tail keywords long tail keywords examples | 1,000 720 390 | 2,110 |
| Low competition keywords | low competition keywords how to find low competition keywords | 140 140 | 280 |
| Keyword difficulty | keyword difficulty what is keyword difficulty | 320 140 | 460 |
| Search intent | search intent types of search intent | 390 70 | 460 |
| Keyword clustering (guide) | keyword clustering | 480 | 480 |
| Keyword clustering tools | keyword cluster tool | 720 | 720 |
| Keyword grouping | keyword grouping | 70 | 70 |
| Topic clusters | topic clusters | 260 | 260 |
| Pillar pages | pillar page | 320 | 320 |
What the example shows
- Question and definition keywords usually cluster together. “Long tail keywords” and “what are long tail keywords” share 6 ranking URLs. One solid guide can target both, and the “examples” variation fits in the same page.
- Synonyms do not always cluster. “Keyword clustering” and “keyword grouping” share zero URLs. The “grouping” results lean towards PPC ad groups and grouping tools, while “clustering” results are SEO guides. Writing one page for both would miss one of the two audiences.
- Modifiers like “tool” change the page type. “Keyword cluster tool” shares only 2 URLs with “keyword clustering”: people searching it want a list of tools, not a how-to. That is a different page.
- Related concepts are not the same page either. “Topic clusters” and “pillar page” are discussed together in almost every article about content strategy, yet they share no ranking URLs. The same sites (Semrush, HubSpot) rank for both, but with different articles.
- 14 keywords became 9 pages, not 14 and not 3. That is typical. The number of pages you need depends on how Google splits the intent, not on how many keywords you found.
Keyword clustering by hand vs with tools
For a small niche site, doing it by hand is perfectly fine. Fifteen to thirty keywords take an hour with a spreadsheet and a private browser window, and you learn a lot about the search results while you do it.
Beyond a hundred keywords, a tool saves real time. Keyword Insights, SE Ranking, Semrush’s Keyword Strategy Builder and the Keysearch cluster tool all group keywords by SERP overlap. Whatever you use, check two things in the settings: the country (clusters change by market) and the threshold (3 is a good default; higher gives you more, smaller clusters).
Be careful with “AI” or semantic clustering that groups keywords by meaning alone. As the “keyword grouping” example shows, meaning and search results do not always agree.
Common keyword clustering mistakes
- Clustering by domain instead of URL. Big sites rank for everything; only the exact URLs tell you whether it is the same page.
- Using another country’s results. A cluster that works in the US may split differently in the UK or Australia.
- One page per keyword. This is how blogs end up with five near-identical articles and none on page one.
- Never re-checking. Search results change. For your most important clusters, check the overlap again every six to twelve months.
- Ignoring pages you already have. The best page for a new cluster is often an existing post that just needs updating.
Keyword clustering FAQ
How many keywords should be in one cluster?
There is no fixed number. Some clusters have one keyword, others have dozens of long-tail variations. What matters is that every keyword in the cluster can be answered well by the same page.
What threshold should I use for SERP overlap?
Three or four shared URLs in the top 10 is the most common choice. Lower thresholds make bigger, looser clusters; higher thresholds make smaller, stricter ones. If you are unsure, start with 3 and look at the borderline pairs by hand.
Is keyword clustering the same as topic clustering?
No. Keyword clustering groups keywords into pages. Topic clustering groups pages into a hub-and-spoke structure around a subject. You usually do keyword clustering first, then organize the resulting pages into topic clusters.
Can I cluster keywords for free?
Yes. Search each keyword in a private window, copy the top 10 URLs into a spreadsheet and count the matches. It is slow for large lists but completely reliable, because you are looking at the same results your readers see.
Once your clusters are ready, the next step is turning them into pages that deserve to rank. My guides to niche research, programmatic SEO (and when not to use it) and the niche blogging hub pick up from there.