Keyword Clusterer — Group Keywords Into Topic Clusters

Overlapping keywords split your ranking power

A keyword list is full of different phrasings of the same question: "how to remove coffee stains", "remove coffee stains from carpet", "coffee stain removal tips", "best method to clean coffee stains".

Each one looks like it could be a page. But the person asking any of them wants the same thing — how to get the stain out — and one thorough article satisfies all four. Build four thin pages instead and they work against each other: your internal links spread across four URLs rather than concentrating on one, and a search engine facing four near-identical pages of yours has to pick, often choosing the weakest or none at all.

The rule worth remembering

If two keywords would be satisfied by the same page, they belong on the same page. Not because the words look alike, but because the need behind them is identical. "Coffee stain removal" and "how to remove coffee stains" are one page. "Coffee stain removal" and "best coffee maker" share a word and nothing else.

How this tool groups

It takes the most substantive word in each phrase — discarding articles, prepositions, question words and intent modifiers, along with anything of two characters or fewer — and groups keywords that land on the same root.

The filtered-out list includes words like "a", "the", "how", "why", "best" and "free", which is deliberate: they describe the framing rather than the topic, so leaving them in would group every "best…" query together regardless of subject.

Where word-shape matching succeeds and fails

It works because related words often share a stem — "removal", "remover" and "remove" cluster naturally, and that covers a good proportion of real keyword lists.

It fails on synonyms with different roots. "Stain removal" and "getting marks out of fabric" want the same page and share no vocabulary, so the tool will keep them apart. It can also occasionally join unrelated keywords that happen to share a long word. This is grouping by vocabulary rather than by meaning, and knowing that tells you exactly where to check.

The human step

Treat each cluster as a candidate page and ask two questions:

1. Would one page genuinely satisfy every keyword in this cluster? If not, split it.
2. Do two clusters actually want the same page? If so, merge them — this is where the synonym limitation shows up.

Both corrections are trivial before you write and expensive afterwards, once you have published two pages and have links pointing at each.

Choosing the primary keyword

Within a cluster, one phrasing earns the title and the URL while the rest appear naturally in the body and headings. You do not need to force every variant in — a search engine matching a page to a query does not require the exact string, and stuffing variants reads badly to the person who actually arrives.

Once the clusters are settled, the intent classifier is worth running over each to confirm the whole cluster shares one intent. A cluster mixing informational and transactional keywords usually needs splitting, because no single page format serves both.