How to Group Cannabis Keywords Into Content Clusters

Cluster by SERP overlap, not word similarity

The classic mistake is grouping keywords because they share words. "edibles dosage" and "edibles recipes" look like siblings on a spreadsheet, but their search results share almost nothing — Google treats them as different needs. Meanwhile "weed delivery" and "cannabis delivery" read as different terms and return nearly identical results. The test is always the SERP. Same results, same page. Different results, different pages. Every good clustering decision reduces to that comparison.

Run the overlap check

For a small list, do it manually: search both terms in an incognito window and count shared URLs in the top ten. Four or more shared results is a strong signal one page can rank for both. For bigger lists, let a tool do the counting — Ahrefs groups terms by Parent Topic in Keywords Explorer, and dedicated clustering tools such as Keyword Insights run pairwise SERP comparison across thousands of terms. Manual for fifty keywords, tooling for five hundred.

Name the pillar

Each cluster gets one broad page that targets the head term and defines the topic — the pillar. For an edibles cluster that is a comprehensive edibles guide. The pillar aims at the hardest term in the cluster, so it is usually the last page to rank and the one that benefits most from every supporting page published around it.

Map the supporting pages

Everything in the cluster that failed the overlap test with the pillar becomes its own supporting page. Dosage, storage, how long effects last, making them at home, traveling with them: each a distinct SERP, each a page. Write the list down with one target URL per group and you have converted a keyword pile into a build order.

Link the hub together

Every supporting page links up to the pillar with descriptive anchor text, and the pillar links down to each supporting page from the section that mentions its subtopic. Add sideways links where topics touch, like dosage referencing how long effects last. This mesh is what tells Google the pages are one body of expertise rather than eight strangers, and it is the part most sites skip.

A worked example: the edibles cluster

For a retail client I clustered 38 edibles-related terms. SERP overlap collapsed them to nine pages: one pillar ("edibles" guide) and eight supporting posts, including dosage, onset time, storage, and flying with edibles. Eleven phrasings of the onset question all shared a SERP, so they became one post with the variants as H2s. Six months after the internal linking went in, the pillar itself entered the top ten for the head term — carried there by the supporting pages, not by any change to the pillar.

Frequently asked questions

How much SERP overlap means one page?

A common working threshold is four or more shared URLs in the top ten. Above that, Google treats the phrasings as the same intent and one page can rank for both. Below two shared results, plan separate pages.

Can I cluster without paid tools?

Yes, just slowly. Search each pair of candidate keywords in an incognito window and compare the top ten by hand. It is practical up to a hundred or so terms, which covers most single-location dispensary sites.

How big should a cluster be?

One pillar plus roughly four to ten supporting pages is workable. Fewer and the hub is too thin to signal authority; many more and it likely contains two topics that should be split into separate clusters.