Keyword Clustering Tool

Keyword clusters built from demand, not invention

Before anything is grouped, every candidate is priced and checked. Keywords with no search demand go. So do keywords carrying a year, which dates the page, and anything with “near me”, which is local intent. Keywords already on your roadmap, or a variant of a page on it, are removed, and so are keywords that are not your territory, such as a competitor’s branded product queries. A same-intent check then compares the survivors with each other and with your roadmap. The seed itself is always kept, because it is your wording.

How the clustering works: AI names it, code decides it

The model’s only job is to name what each keyword is about. Deterministic code does the grouping, so two different subjects can’t be merged by a creative model, and country and language are hard boundaries. A breadth setting is applied on top of the result, which keeps it predictable:

• Tight: every distinct subject keeps its own cluster. Most clusters, each one precise.
• Balanced: a subject folds into a broader one already on the roadmap when it carries all of its words. The default.
• Broad: as Balanced, then any cluster still holding one page joins the subject it shares the most words with. Fewest clusters.

A keyword clustering example: under Balanced, “best anti fatigue mat for standing desk” carries every word of “standing desk mat”, so it becomes an angle on that subject instead of a cluster of its own. When you re-cluster the whole roadmap through the SEO Copilot, you can ask for fewer, broader clusters or tighter ones for that run, preview every move, then apply or discard it. Undo restores the previous grouping.

From clustering keywords to pages you can build

Each surviving keyword gets the top ten results pulled to measure DR and traffic, then one of nine page types, from Guide Blog and VS Blog to Collection, Product and Tool/Calculator pages, chosen with a prompt grounded in your brand context. It is written to the roadmap builder under its seed with volume, difficulty, intent and a Priority Score. Duplicates are checked again against the roadmap at the moment of writing, the same rule the keyword cannibalization checker applies everywhere else.

In our own workspace, 777 pages sit under 665 seed keywords, and one seed, “topical authority”, produced 16 of them. Every page traces back to a search people make. You can see the same clusters as territory on the topical authority map.

Grow keyword clusters without overlap

Every seed group on the roadmap ends with an Expand the cluster row. It mines related queries and adds only those that don’t cannibalize anything already there, up to 50 pages per run; run it again for the rest. The + beside any root keyword expands from that one page. If nothing qualifies, you are told why and no run starts. Roadmap expansion opportunities go further: new clusters, each required to name the existing seed it builds on, with demand measured when available and never guessed.

A keyword grouping tool for the list you already have

Import List takes keywords, URLs or a CSV up to 2 MB and 1,000 rows. Before anything is applied it sorts the rows into high risk of cannibalization, already covered on the roadmap, ready to import, and not recommended by house rules or for having no search volume. Values you supplied stay authoritative; Vee3.ai fills in only what is missing.