Back to the blog

A site’s semantic core: how to gather it with Wordstat and not drown in keywords

Step by step, we break down how to gather a semantic core with Yandex Wordstat: query types, clustering, negative keywords and typical beginner mistakes.

May 12, 2026
11 min read
67 views
MOLOTILO

MOLOTILO DIGITAL

A site’s semantic core: how to gather it with Wordstat and not drown in keywords

What a semantic core is in plain words

A semantic core is the list of queries people use to search for your products and services in search engines. Not an abstract “keyword list for show”, but effectively a map of demand: what people ask, how they phrase it, what matters to them — price, timelines, reviews or “near home”.

Without a core, site promotion turns into guesswork. With a core you know exactly which pages to create, what to put in headings and which articles the blog needs.

Types of search queries

By frequency

  • High-frequency (HF) — “buy a sofa”. Thousands of impressions a month, fierce competition.
  • Mid-frequency (MF) — “buy a corner sofa cheap”. More specific already.
  • Low-frequency (LF) — “buy a corner sofa with a storage box in Tver”. Few impressions, but the person knows what they want and is almost ready to buy.

A common-sense rule: for a young site to climb into HF is like opening a stall across from a hypermarket. Start with LF and MF — the traffic there is cheaper and conversion is higher.

By intent

  • Commercial — “buy”, “price”, “order”, “with delivery”. They lead to cards and services.
  • Informational — “how to choose”, “which is better”, “DIY”. They lead to the blog.
  • Navigational — searching for a specific brand or site.

Mixing intents on one page is a classic mistake. A service page with a wall of “how to choose” in the middle ranks for neither commercial nor informational queries.

How to work with Yandex Wordstat

Wordstat (wordstat.yandex.ru) is Yandex’s free query-statistics service. The main techniques:

  1. A base query. Type “apartment renovation” — you see everything containing these words, plus the right-hand “similar queries” column. The right column is gold: synonyms and adjacent topics live there that you wouldn’t think up yourself.
  2. Quotes. «"apartment renovation"» shows the frequency of exactly this phrase without tails. The difference between 100,000 and 2,000 in quotes means the main demand is in refined queries.
  3. The «!» operator. Fixes the word form: «!apartment !renovation».
  4. Region. Always set your city — nationwide frequency tells you nothing if you work in Samara.
  5. Query history. The seasonality tab: air conditioners are searched in May–July, Christmas trees in December. Plan content in advance.

A step-by-step plan for gathering a core

  1. Write out markers — the base names of services and products, as you call them and as clients call them (often different words: you say “oral sanitation”, the client says “fix all my teeth at once”).
  2. Run the markers through Wordstat, gather the expansions and the right column.
  3. Peek at competitors. Open the top-5 sites for your queries and look at the structure: what pages, sections and articles they have.
  4. Collect search suggestions — what Yandex and Google offer as you type. These are people’s live phrasings.
  5. Clean out the junk: negative keywords “free”, “DIY”, “used”, “jobs”, other cities and brands you don’t carry.
  6. Cluster — group queries by meaning. One cluster = one page. “Buy a sofa” and “sofas cheap” are one page, while “how to reupholster a sofa” is a blog article.

LSI words: what to “dress” pages with

LSI (latent semantic indexing) — words accompanying the topic. For “apartment renovation” these are “estimate”, “rough finishing”, “screed”, “turnkey”, “foreman”. By them search engines understand the text is written by an expert, not by a copywriter for 100 rubles. Where to get them: the right column of Wordstat, word highlighting in the results, the “people also search” block at the bottom of the search page.

Typical mistakes

  • Gather 10,000 keywords and stop — a core without pages does nothing.
  • Make a page for every query — you’ll get a hundred duplicate pages that merge in the results.
  • Ignore region and seasonality.
  • Gather it once and forget — demand changes, a core is worth reviewing every six months.

Takeaway

A semantic core isn’t magic but a careful few days’ work: Wordstat, competitors, suggestions, cleaning, clustering. But afterward every page of the site answers real demand, not the owner’s fantasies. If you don’t have time to figure it out yourself — we gather the core and the site structure as part of our SEO services.

Enjoyed the article?

Subscribe to our blog so you don’t miss new posts