Keyword Clustering Tool

How to use the Keyword Clustering Tool

1. Export or copy your keyword list from your research tool, spreadsheet or Search Console and paste it with one keyword per line.
2. If you have a metric for each keyword, such as monthly searches from your keyword tool, keep it after a comma or tab. Clusters will then show a total for that metric.
3. Choose the similarity rule and threshold. Start with the defaults; raise the threshold for tighter, more specific groups or lower it for broader topics.
4. Add any words that appear in most keywords but do not define the topic, such as your brand name or a year, to Words to ignore.
5. Select Cluster keywords, review the groups, and download them as CSV to plan pages in your spreadsheet.

Example: a running-shoe keyword list

Select Load example list to cluster twelve running-related keywords, each with a search-volume figure. With the default settings (Jaccard, 0.40) the result is:

The broad “running shoes” cluster suggests one strong category page. If you raise the threshold to 0.70, each modified phrase — “best running shoes”, “trail running shoes” and so on — becomes a cluster of its own, with “waterproof trail running shoes” grouped under “trail running shoes”. Neither answer is automatically right: the threshold controls how finely you split topics, and checking the live search results tells you whether searchers expect separate pages.

Why group keywords before writing

A keyword list from any research tool is usually long and repetitive: dozens of variations of the same idea alongside genuinely different topics. Writing one page per keyword leads to thin, overlapping pages that compete with each other. Writing one page for everything leads to a page that serves nobody well.

Clustering sits between those extremes. It groups phrases that share their core words, so each cluster becomes a candidate page, and the keywords within it become the phrases and subtopics that page should cover.

The clustering rule, step by step

The tool uses lexical similarity — shared words — rather than any hidden or paid data source. Each step is deterministic, so the same list and settings always produce the same clusters.

1. Normalise. Each keyword is lower-cased, split into words and stripped of common English stop words (“for”, “the”, “how” and so on) and any words you chose to ignore.
2. Simplify plurals. Regular English plurals are reduced to their singular form, so “shoes” and “shoe” count as the same word. Irregular plurals are left alone.
3. Order. Keywords are sorted from fewest to most significant words, so short head terms are considered first and become the natural centre of a cluster.
4. Assign. Each keyword joins the existing cluster whose head term it is most similar to, provided the similarity reaches your threshold. Otherwise it starts a new cluster.

Similarity is measured in one of two ways. Jaccard divides the number of shared words by the number of distinct words across both phrases; “running shoes” and “best running shoes” share 2 of 3 words, a score of 0.67. Overlap divides shared words by the size of the shorter phrase; the same pair scores 1.0. Overlap produces larger, broader clusters because any longer phrase containing a head term joins it.

Who it helps

Content strategists use it to turn a research export into a list of pages. Bloggers use it to avoid writing three articles that target the same question. Agencies use it to present a client with a structured topic map, and online shops use it to decide which product-category pages to create.

Limitations to keep in mind

• Clustering is based on shared words, so synonyms with no words in common — “sofa” and “couch”, for example — are not grouped together. Merge such clusters manually.
• The stop-word list and plural handling are designed for English. Other languages can still be clustered, but every word is treated as significant.
• The tool does not provide search volumes, difficulty scores or ranking data. Any metric shown is one you supplied.
• Search engines may treat two clusters as one intent (or one cluster as two). Check the current results for a few keywords before committing to a page plan.

How your data is handled

Your keyword list is processed entirely in your browser and is not uploaded, stored or shared. Many people treat their keyword research as confidential; with this tool it never leaves your device.

Frequently asked questions