SERP-Overlap Keyword Clustering
Group queries by shared search results, then build one comprehensive page per intent.
- Difficulty
- Advanced
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 98%
Start with every query related to a target subject, then compare which URLs appear across their search results. Queries that repeatedly surface the same pages are treated as one intent cluster because Google has observed that users respond well to the same results. Instead of creating a separate article for every wording, build one page that addresses the cluster's full set of subtopics, examples, comparisons, and follow-up questions. This mechanism converts a noisy keyword list into a smaller map of distinct user intents. It reduces keyword cannibalization and unnecessary production while making each surviving page more comprehensive. The framework also replaces subjective clustering decisions with evidence derived from actual search-result behavior.
Origin
Ethan Smith described Graphite's clustering method through a five-year-plan example on Marketing Against the Grain.
Core principles
- 01Search-result overlap reveals shared user intent.
- 02One intent cluster usually needs one comprehensive page.
- 03Subtopics expose the coverage required to satisfy the intent.
- 04Observed SERPs are more reliable than manual keyword assumptions.
How to run it
- 1
Build the keyword universe
Collect the exact query and its variations, modifiers, comparisons, examples, and follow-up questions. Preserve low-volume terms because their combined demand may be substantial.
Pro tip Include personal, professional, template, example, and comparison variants where relevant.
Watch out Do not discard a cluster because its head term alone has modest search volume.
- 2
Measure SERP overlap
Compare the URLs appearing for each query and calculate which queries repeatedly share the same results. High overlap signals that searchers accept the same content for both queries.
Pro tip Use recurring keyword-URL pairs rather than semantic similarity alone.
Watch out Standard language similarity can group phrases that actually require different pages.
- 3
Separate distinct intents
Split queries when their result sets materially differ, indicating that users expect different destinations or answers. Keep queries together when the same pages consistently satisfy them.
Pro tip Let observed result behavior resolve ambiguous cases such as personal versus business intent.
Watch out Do not force every related phrase into one cluster merely because the wording is similar.
- 4
Map required subtopics
List every substantive theme represented inside the cluster. Treat these themes as coverage requirements for the eventual page.
Pro tip Look for examples, templates, alternatives, and comparisons that recur across query variants.
Watch out A page targeting only the head term can remain incomplete even when its prose is strong.
- 5
Publish one comprehensive asset
Create one page that fulfills the shared intent and covers the cluster's meaningful subtopics. Optimize its structure around the complete need rather than repeating the exact keyword.
Pro tip Use clear sections so visitors can reach the part matching their specific query quickly.
Watch out Creating one page per keyword wastes resources and can make the pages compete with one another.
In the wild
A team finds queries for five-year-plan examples, personal plans, professional plans, templates, and five-versus-ten-year plans. Shared search results indicate that these belong to one intent cluster. The team creates one guide containing examples, templates, personal and professional applications, and a comparison of planning horizons rather than commissioning separate articles.
→ One comprehensive asset can rank across the cluster while avoiding the cost and cannibalization of many narrow pages.
Common mistakes
Making one page per keyword
Closely related queries often express the same intent, so separate pages create duplication and unnecessary expense.
Clustering by wording alone
Semantic similarity does not prove that users want the same result; compare actual SERP overlap.
Ignoring cluster coverage gaps
A page can target the correct intent but still fail because it omits examples, templates, or comparisons present in the query set.
Is it for you?
Best for
It is best for SEO teams planning content across hundreds or thousands of related queries.
Not ideal for
It is not ideal for isolated topics with little search demand or insufficient SERP data.
From the transcript
“we basically built an algorithm to look to see which keywords have the same set of search results and if it has the same set…”
“you make one piece of content for all thousand of those”
“it's not only faster and cheaper it's it's significantly better than a human could do it on their own”
From the episode
Why Google Is NOT Dead & How To Dominate Ai-Driven Search ft. Ethan Smith