What is keyword clustering?

Keyword clustering is the process of grouping a large keyword list into sets of terms that share the same search intent, so that each group becomes one page instead of many thin pages competing with each other. It is the step between keyword research and actually writing.

The problem it solves

Keyword research hands you a spreadsheet. A small site might export 400 rows; a serious one exports 40,000. Nothing in that spreadsheet tells you the thing you actually need to know, which is how many pages to write and what each one should cover.

The naive answer — one page per keyword — is where most sites go wrong. If you publish separate pages for “how to cluster keywords”, “keyword clustering guide” and “how do I group keywords for SEO”, you have written three articles that a search engine sees as one topic. They compete with each other for the same result, your internal links are split three ways, and each page is thinner than the single strong page you could have written instead. That failure has a name: keyword cannibalisation.

Clustering fixes the input to that decision. Group the keywords first, and the page count falls out of the data rather than out of guesswork.

Keyword grouping and clustering: the same thing?

The two terms get used interchangeably, and arguing about them is not a good use of your time. There is a real distinction worth knowing, though:

• Grouping usually means sorting by a shared token — everything containing “price”, everything containing “vs”. Fast, and useful for filtering.
• Clustering usually means grouping by relatedness: by meaning, or by how much the actual search results overlap.

The second is what produces a content plan. The first mostly produces tidier spreadsheets.

Keyword clustering methods

There are three methods in common use — matching shared words, comparing search results, and comparing meaning. The first is the one most people reach for, and the one that breaks first. Consider three keywords from a coffee site:

“cold brew ratio” · “how to make cold brew” · “iced coffee recipe”

Word-matching puts the first two together, because they share “cold brew”. It leaves the third somewhere else entirely — it shares no words with either. But anyone searching all three wants roughly the same article, and the search results for them overlap heavily.

This is the central limitation. Language has synonyms, abbreviations, and completely different phrasings for identical intent. “cheap flights berlin” and “budget airline tickets to berlin” share one word. “SERP” and “search results page” share none.

That first method has a name — lexical grouping — and a natural home, which is a spreadsheet. It is a perfectly good keywords grouping method for a few hundred terms you already understand, and the formulas for doing it in Excel are here. Beyond that, two approaches handle the problem properly:

SERP-overlap clustering

Search each keyword, and group keywords whose top results substantially overlap. This is the most direct evidence available, since it is Google's own opinion about which queries are the same. The catch is cost: it needs a live search result for every keyword, which is why tools that do it well are expensive.

Semantic clustering with embeddings

Convert each keyword into a vector — a list of numbers positioning it in a space where distance means dissimilarity of meaning — then group the ones that sit close together. A model trained on a lot of text learns that “SERP” and “search results page” belong together, without ever being told.

This is what Graph My Keywords does, using a compact sentence-embedding model that runs inside your browser. It costs nothing to run, needs no API keys, and never uploads your keyword list. It does not see live search results, so it can occasionally group two things Google treats separately — which is why the output is a draft plan you review, not an oracle.

The benefits of keyword clustering

Four of them, in the order you tend to notice them:

• You find out how many pages to write. A 3,000-row export is not 3,000 articles, and it is usually not 300 either. Clustering turns an unreadable list into a countable one.
• You stop competing with yourself. Pages built from clusters cannot cannibalise each other, because no two of them target the same intent by construction.
• Priorities change. Summed cluster volume routinely promotes a topic of thirty small keywords above an obvious head term — and the small-keyword topic is far easier to rank for.
• Briefs write themselves. The keywords inside a cluster are the outline of the page. Handing a writer a cluster is a better brief than handing them a keyword.

From clusters to pages

A cluster is not yet a plan. Three things turn it into one:

1. A page title. Take the most representative keyword in the cluster, not the highest-volume one — they are often different, and the representative one describes the article better.
2. The search intent. A cluster of “best X”, “X vs Y”, “X alternatives” is commercial: the reader is choosing. A cluster of “what is X”, “how does X work” is informational: the reader is learning. The same subject needs a completely different article depending on which it is.
3. Total search volume. Sum the volume across the cluster, not the head term alone. A topic of thirty small keywords routinely outweighs one obvious head term, and it is far easier to rank for.

That last point is the one people miss most often. Clustering does not just tell you what to write — it changes which topics look worth writing at all.

A keyword clustering example

The most useful example we have is our own. We ran this site's Search Console export through this site's tool, and one cluster came back holding these terms:

keyword clustering · keyword grouping · what is keyword clustering · how to do keyword clustering · keyword clustering methods · benefits of keyword clustering · keyword clustering examples · keyword grouper excel · keyword grouping excel · free keyword grouper · keyword grouper software · group keywords · cluster keyword · pre-clustered keywords

Semantically that is one tight cluster, and the grouping is correct: every term is about the same subject. But read it for intent and it is plainly two piles.

Written as one page it would half-serve both and win neither. This is the single most common thing a cluster hides, and no tool will catch it for you — the split is a judgement about people, not about words.

How big should a cluster be?

There is no correct number, but there are two failure modes.

Too coarse and you get one enormous cluster containing genuinely different intents — exactly the case above, where one correct cluster of twenty-odd terms was two pages wearing a single label.

Too fine and you get forty clusters of two keywords each, which is just your original spreadsheet with extra steps.

A reasonable working rule: anything above roughly fifteen keywords deserves a manual look to see whether it hides two intents. Anything below three is usually noise to fold into a neighbour.

Common mistakes

• Treating clusters as final. They are a first draft produced by a machine that has never seen your market. Review them.
• Ignoring intent inside a cluster. A group that is 60% informational and 40% commercial is two pages, however tightly it clusters.
• Sorting only by head-term volume. The cluster total is the number that matters.
• Clustering a dirty list. Branded terms, duplicates and typos distort the groups. Strip them first.
• Writing every cluster. Most keyword lists contain topics you have no business competing for. Clustering tells you what exists, not what you should attempt.

Frequently asked questions

How many keywords should one page target?

As many as share one intent. A single well-written page routinely ranks for dozens of variants of the same question, and trying to split them across pages actively hurts. The limit is intent, not count.

Do I need a paid tool?

No. Paid tools mostly buy you SERP-overlap data and scale. For a list of a few thousand keywords, semantic clustering in the browser gets you a usable plan for free — this one does.

Can I cluster keywords in Excel or Google Sheets?

You can, and for a few hundred keywords it is genuinely the right tool — here are the formulas, including the tag-matching one most people are looking for. What a spreadsheet cannot do is spot that two phrasings with no words in common mean the same thing, which is where most of the value is.

How often should I redo it?

When your keyword data materially changes — a new Search Console export after a few months, or entering a new area. The clusters themselves are stable; your coverage of them is what moves.

Next