Keyword Clustering Tool

Why a keyword list isn’t a content plan

An export gives you rows. A plan needs pages. The gap between them is that several keywords usually describe one page, and if you treat each row as an article you produce a set of thin pages that compete with each other — a problem you then have to diagnose later as cannibalisation.

Clustering is the step that closes that gap. Group the list, and each group becomes one brief with one primary keyword and a handful of variations the page should also cover.

What this tool does, precisely

It groups keywords by the words they share, weighted so that rare words count for more than common ones. That last part is what stops a list where everything begins with “best” from collapsing into one useless group called “best”. Plurals are folded, so “running shoe” and “running shoes” group; function words like how, to and vs are capped so they can never be the reason two keywords group.

Each group grows from a seed — the highest-volume keyword in it, or the shortest when your list has no volumes — so a group means “these are all variants of this head term”, and the name it’s given is the words every member has in common.

What it can’t do, precisely

This is the part most free clustering tools leave out, so here it is plainly: grouping by shared words is a first pass, not a verdict.Proper clustering asks a different question — which pages actually rank for each keyword, and do those results overlap? Two keywords that return the same ten results are one page no matter how differently they’re worded. That comparison needs a live results lookup for every keyword in your list, which is a paid request per keyword, which is why no tool does it for free.

So there are two failures to watch for, and both are easy to spot by eye:

• Synonyms split.“cheap flights” and “budget airfare” are one page and share no words, so they land in different groups. Nothing lexical can fix that; you can, in about ten seconds of reading.
• Intents merge.“what is project management” and “project management software” share three words and want completely different pages — one explains, one compares. Load the example above and you can watch it happen.

Which is the right way to use the output: as a first draft of a plan that you spend five minutes correcting, rather than a plan you brief straight from.

How to work the slider

There is no correct sensitivity, because it depends on how broad your list is. A list pulled from one topic needs a tighter setting to separate anything; a list spanning a whole site needs a looser one or you get hundreds of pairs.

The practical method is to start at Balanced and move one step at a time until the groups look like pages you would actually commission. If a group contains two obviously different jobs, tighten. If half your list is sitting in the ungrouped list at the bottom, loosen.

That ungrouped list is worth reading rather than skipping. Keywords that match nothing else are often the most interesting rows in an export — a question nobody on your site has answered, phrased in a way nothing else in your list uses.

Where the list should come from

Clustering is step seven of the process in the keyword gap guide, and the six steps before it decide whether any of this is worth doing. Grouping a list you haven’t filtered just gives you tidy groups of keywords you shouldn’t target.

Keyword Gapproduces the list this tool is built for: every keyword a competitor ranks for and you don’t, with both positions, volume, difficulty and an intent label, which you can export and paste straight in above. If you’re starting from a seed term rather than a competitor, Keyword Research is the other end of the same job, and Organic Keywords shows what you already rank for — worth checking before you brief a page that competes with one of your own.