SEO

Keyword clustering and page mapping

Use when a large keyword export needs turning into a finite list of pages to build.

keyword-clustering-plan.md

Download .md

You are clustering keywords into pages. One cluster equals one page.

Keyword export (keyword, volume, difficulty if available): {{KEYWORD_EXPORT}}
Existing site pages: {{EXISTING_PAGES}}
Site type and constraints: {{SITE_CONTEXT}}

Output a table: Cluster name | Primary keyword | Supporting keywords | Combined volume | Intent | Map to existing URL or create new | Page type | Priority (1-3).

Rules for clustering:
- Group by shared intent and by whether one page could satisfy all of them, not by string similarity.
- If two keywords would need genuinely different pages to satisfy them, keep them apart even if they share words.
- Any keyword that does not fit a cluster goes in a final "Unassigned" list with a one line reason.

Priority rule: 1 = high combined volume and no existing page competing; 2 = worth doing but overlaps existing content; 3 = low volume or hard given {{SITE_CONTEXT}}.

Constraints:
- Do not report volume figures that are not in the export. Sum only the numbers supplied.
- Do not create more than {{MAX_CLUSTERS}} clusters. If the list needs more, say which keywords were left out.
- Prefer mapping to an existing URL over creating a new page whenever intent matches.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

{{KEYWORD_EXPORT}}
{{EXISTING_PAGES}}
{{SITE_CONTEXT}}
{{MAX_CLUSTERS}}

Getting a better result

Include your existing pages, otherwise it will propose building things you already have.
Set a max cluster count that matches your real publishing capacity for the quarter.
Check the Unassigned list - it often contains the most interesting queries.

Questions about this prompt

When do I need clustering rather than a keyword list?

When the export is bigger than the number of pages you could ever build. Clustering turns thousands of queries into a finite list of pages, which is the actual decision: not which keywords matter but how many pages you are going to make.

Why include my existing pages?

Otherwise it proposes building things you already have. Mapping clusters against the current site is what turns the output into a plan rather than a wish list, and it usually reveals that several clusters are already covered by a page that just needs improving.

How should I set the maximum cluster count?

To your real publishing capacity for the quarter, not your ambition. A plan for sixty pages when you can produce twelve is a plan to abandon the plan, and the first twelve will be chosen by whoever is least busy rather than by value.

Is the unassigned list worth reading?

It is often the most interesting part. Queries that do not fit any cluster are either noise or a genuine gap nobody in your market has covered, and the second kind is where the easiest rankings tend to be.