Keyword research in 2026 — the cluster-first method
Stop researching individual keywords. Research clusters of intent, then build the entity registry that satisfies them.
Step 1 — collect raw queries
Pull seed keywords from Google Search Console, GA4 site search, Ahrefs/SEMrush volumes, and competitor SERPs. Aim for 8,000–20,000 raw queries before clustering.
Step 2 — cluster by SERP overlap
Two queries belong to the same cluster if their top-10 SERPs overlap by ≥30%. This is purely empirical, not semantic — Google has already done the work of grouping intent.
Step 3 — classify intent and assign surfaces
Each cluster gets one of four intents and one surface: informational → guide page; navigational → brand page; commercial → comparison page; transactional → service page.
Common questions
What tool clusters keywords by SERP overlap?
Keyword Insights, ClusterAI, and Surfer all support this. SEMrush's Keyword Magic Tool does it natively. Manual clustering at this scale is not viable.