Query clustering by meaning alone often gets contradicted by the actual SERP: two queries look like synonyms but rank completely different pages, and a single page trying to serve both ends up losing ground on each. This skill checks every disputed cluster boundary against real search-result overlap instead of semantic guesswork.
How the query clustering works
It collects core and long-tail queries for a topic, drafts clusters by meaning, then verifies disputed boundaries by comparing the actual Yandex SERPs for neighboring clusters' lead queries — merging clusters when most top URLs overlap, keeping them separate when they don't, even for queries that sound alike. It separately flags the pairs where semantic intuition and the actual SERP disagree the most, since that's where manual search query clustering tends to get it wrong.
What you get
A chat report: clusters with name, page type, queries with frequency, and total demand per cluster; a separate section listing which pairs were merged or split against first semantic impressions and on what SERP overlap; clusters ranked by demand-to-competition ratio so it's clear where to start.