What is keyword clustering?

Keyword clustering is the process of grouping keywords into clusters with a shared search intent (SERP intent), so that each cluster targets exactly one destination page. Goal: maximize the semantic reach of a page without cannibalization and without building separate URLs for synonyms.

Instead of writing 50 separate pages for 50 queries, clusters let you optimize 10-15 pages, each ranking for 10-30 related queries.

Why does it matter?

Elimination of cannibalization — when 2 pages rank for the same thing, Google lowers both. Clusters prevent this
Higher content efficiency — one page = many keywords = more clicks
Better topical mapping — clusters are the foundation of topical mapping
Scalability — instead of 500 micro-pages, you leave 50 strong pillars
Better structure planning — clusters show how many URLs you actually need

How does SERP-based clustering work?

The most reliable method doesn't use word2vec similarity, but real Google SERPs:

Fetch top 10 URLs for each query from the list
Compare overlap — if 2 queries have ≥3 common URLs in top 10, they belong to the same cluster
Build a graph — queries are vertices, common URLs are edges; clusters are connected components
Pick the pilot keyword — query with highest search volume in the cluster

Why SERP overlap is best: Google knows best which queries have the same intent — it shows the same top 10. If it shows different URLs, intents differ and require separate pages.

Types of search intent

Each cluster has one of 4 base intents:

1. Informational

"what is [X]", "how does [X] work", "[X] vs [Y]"
Clusters → blog, glossary, guides
Conversion: low, but builds authority + remarketing

2. Navigational

"[brand] login", "[product] pricing"
Clusters → brand pages
Conversion: medium, user already knows the brand

3. Commercial Investigation

"best [X]", "[X] review", "[X] alternatives", "[X] for small companies"
Clusters → comparison pages, listings, lists
Conversion: high, user close to decision

4. Transactional

"buy [X]", "[X] price", "order [X]"
Clusters → product pages, landing pages
Conversion: highest

How to do clustering — step by step

Step 1: collect the query list

From Google Search Console (already ranking)
From Ahrefs/Semrush (potential)
From Google Autocomplete + AlsoAsked
From competition (Ahrefs site explorer)
From keyword research

Typically 500-3000 queries at the start.

Step 2: collect SERPs

For each query → SerpAPI/DataForSEO API → top 10 URLs
Cost: ~5 USD per 1000 queries

Step 3: build overlap matrix

1000 queries × 1000 queries = 1M pairs
For each pair: number of common URLs in top 10

Step 4: apply threshold

≥3 common URLs → same cluster
Connected components in graph = final clusters

Step 5: manual validation

Check top 10 clusters — is the macro-intent really uniform?
Drop clusters with mixed intent

Step 6: assign URLs

One cluster → one existing page (update) or new (write)
Pilot keyword in title/H1, others in H2/H3/body

Tools

MarketingOS — cli has cannibalization audit + striking distance per cluster
Keyword Insights — best dedicated tool ($)
Ahrefs Parent Topic — automatic clusters with SERP overlap
Surfer Keyword Clusterer — built into Surfer
Custom Python + DataForSEO API — cheapest, most flexible

Common mistakes

Clustering by semantic similarity — "SEO" and "search engine optimization" are semantically similar but may have different SERPs
Clusters with mixed intent — "CRM" (informational) and "CRM for 50-person companies" (commercial) → separate clusters
Threshold too low — ≥1 common URL gives too loose clusters; ≥3 is the sweet spot
Ignoring SERP features — featured snippet, video carousel, "people also ask" change intent
No refresh — SERPs change every 3-6 months; clusters need refreshing

Keyword clustering and B2B SEO

In B2B, clusters are usually smaller (5-15 queries vs 50+ in e-commerce), but yield higher ROI:

Each query has low volume but high intent
The cluster ties them into 1 pillar page of 1500-3000 words
The page ranks for the entire long-tail of intent simultaneously
For ABM — clusters per industry/use case

Related terms

Keyword research — input to clustering
SERP — source of overlap signal
Topical mapping — higher structure level
Topical authority — effect of good clustering
Long-tail keywords — most queries in cluster
Cannibalization — what you eliminate through clustering

Keyword clustering — grouping keywords for SEO