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

A keyword grouper for Excel and Sheets lists

Most keyword lists live in a spreadsheet, so the grouper reads them where they are. Upload an Excel or CSV export from any keyword tool, or paste a Google Sheet link, and the first column becomes the input. The grouped result exports back to CSV or Excel with one row per keyword, so you can sort, filter and pivot it in the same workbook you started from instead of grouping keywords by hand with formulas.

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.