Is the list uploaded?

No. Clustering runs in your browser. The file never leaves the device.

Do you estimate volume or CPC?

No. Those need a paid data source. This tool only groups the phrases you provide.

How is a cluster named?

An existing keyword in the group, usually the head term the others extend. We do not append words to invent a title.

Why did two similar phrases stay apart?

Raise the match by choosing Low (broader), or drag one keyword into the other cluster. Containment is weighted so long-tail extensions of a head term still tend to join it.

What happens over 5,000 rows?

Extra rows are dropped and the notice says so. Cluster the rest in a second file if you need them.

Does clustering stem words?

Yes, lightly, so closely related forms can share a cluster. Keyword Density does not stem; those tables are exact tokens.

Keyword Clustering

Loading tool…

About Keyword Clustering

Upload a CSV or TXT (15MB max) or paste a list, then click Cluster keywords. One keyword per line; comma and tab lists are accepted on paste. Multi-column CSV lets you pick the keyword column. Up to 5,000 keywords; extra rows are dropped and a notice says so. Nothing is uploaded.

Keywords group by shared tokens. Short head terms seed clusters; long-tail phrases join them when overlap is high enough. Tune similarity: Low (0.35) makes fewer, broader clusters; Balanced (0.5) is the default; High (0.7) makes more, narrower ones. You can also drag the slider from 0.30 to 0.90. Drag, merge, split, and rename edits are discarded if you cluster again. Download CSV or copy the grouped list. Extra columns from the upload come through on re-export.

How to use Keyword Clustering

1Paste or upload keywords. Upload is the default. Switch to Text for a list. One per line, or a CSV/TXT file.
2Click Cluster keywords. Keywords group by shared tokens. Progress shows on lists over 500 rows.
3Tune similarity if you need to. Low, Balanced, or High. Changing similarity re-runs clustering and discards drag and rename edits.
4Export. Download CSV or copy the grouped list. Original extra CSV columns come through on re-export.

How clusters are named and scored

Similarity is token overlap (part shared-set, part containment) after light stemming, so scores and scoring can join. The cluster title is an existing keyword in the group, usually the shortest head that others extend. We do not invent a new phrase. Rows that look non-topical (navigational junk) go to Excluded as non-topical, which is separate from clustering.

Minimum cluster size can hide singletons, or collect them into Ungrouped. Export uses the clusters themselves, not that on-screen hide. Search volume and CPC are not estimated. If those columns were in the file you uploaded, they pass through in the CSV.

Limits and privacy

Clustering runs in your browser. Files and pasted lists are not sent to a server. No account. Closing the page drops the list. Cap is 5,000 keywords and 15MB for an upload.

FAQs

Is the list uploaded?: No. Clustering runs in your browser. The file never leaves the device.
Do you estimate volume or CPC?: No. Those need a paid data source. This tool only groups the phrases you provide.
How is a cluster named?: An existing keyword in the group, usually the head term the others extend. We do not append words to invent a title.
Why did two similar phrases stay apart?: Raise the match by choosing Low (broader), or drag one keyword into the other cluster. Containment is weighted so long-tail extensions of a head term still tend to join it.
What happens over 5,000 rows?: Extra rows are dropped and the notice says so. Cluster the rest in a second file if you need them.
Does clustering stem words?: Yes, lightly, so closely related forms can share a cluster. Keyword Density does not stem; those tables are exact tokens.