Keyword Clustering
Agent

How the agent groups your list

Every keyword is trimmed, deduplicated case-insensitively and capped at the first 300 entries before anything else happens. Spreadsheet uploads and Google Sheets read the first column and skip a header row automatically; a private sheet prompts you to connect Google Sheets rather than failing silently. The list then goes to a model that groups by shared intent and semantic theme, targeting between 3 and 12 clusters. The result is validated against your original list: only verbatim input keywords survive, each is assigned exactly once, and anything the model drops is swept into a catch-all cluster so nothing disappears.

What each cluster gives you

A cluster is more than a bucket of keywords. Each one comes back with a short name, a one-line theme explaining what unites the terms, an intent label, a keyword count and a concrete pillar-page suggestion you can hand to a writer. The optional seed topic field nudges grouping toward your niche when a list spans several product areas. Every run exports flat, one row per keyword, so clusters drop straight into a content calendar or a sheet you already maintain.

• ›Intent labels: informational, commercial, transactional, navigational, branded
• ›A theme sentence per cluster, not just a group name
• ›One pillar-page suggestion per cluster
• ›CSV or Excel export with cluster, intent, pillar and keyword columns

When to run a clustering pass

The obvious moment is straight after an export from a keyword tool, when you have a few hundred terms and no structure. It is also the fastest way to sanity-check an existing site architecture: cluster the terms you already rank for and see whether your URL structure matches the grouping. Agencies use it at the start of a content audit, in-house teams before a quarterly plan. If the AI layer is unavailable the run still completes using deterministic token-based grouping, and the summary says so.