Free keyword research in Yandex Wordstat filtered to Uzbekistan
You can build a full keyword map without paid tools. You need Wordstat with the region set correctly, three operators, and discipline while cleaning.
Setting the region comes first
By default Wordstat shows volumes across all regions, so your list fills with queries nobody in Uzbekistan types. Switch to region selection and pick Uzbekistan, or Tashkent specifically when working on the capital.
Compare two numbers: total volume and regional volume. The gap shows how relevant the query is locally. A query with high global and near-zero regional volume is not worth pursuing — you would be competing in someone else's market.
Save the region in settings, or the filter resets on each new query and half your list ends up collected for the wrong region.
Three operators you always need
Quotation marks fix the phrase composition: a quoted query shows volume only for that phrase and its inflections, without tails. Unquoted, the figure includes every query containing those words and is inflated several times over.
An exclamation mark before a word fixes its exact form, which helps separate singular from plural — often decisive for commercial queries.
Square brackets fix word order. Rarely needed, but useful when order changes meaning.
The rule is simple: judge demand on quoted volume, gather ideas on unquoted. Budgeting against unquoted volumes is the single most common source of unrealistic expectations.
Building the list
Start with five to seven base terms from your niche. For each, work through the left column to the end and record everything relevant to your service. The right column surfaces adjacent phrasings you had not considered, and often the cheapest queries come from there.
Assemble a set of commercial modifiers and run every base term against them: price, cost, buy, order, cheap, reviews, near me. Keep informational ones separate: how, what is, comparison.
Track it all in one sheet: query, quoted volume, type (commercial or informational), language, target page. The last column gets filled during clustering, but create it from the start.
Uzbek-language queries
Uzbek data in Wordstat is thinner than Russian: some queries show zero volume purely because the sample is small, not because demand is absent. Do not discard them automatically.
Supplement Wordstat with search autocomplete: type a base term in Uzbek Latin and see what gets suggested. These are live queries that statistics may not reflect.
Account for both scripts. Part of the audience types in Cyrillic and part in Latin, and these are separate rows to collect even though the meaning is identical.
Cleaning the list
Remove three groups: queries naming other cities and countries, queries naming competitors, and queries with a clearly different intent — jobs, courses, free, download — unless you sell those.
Separately remove queries pointing at services you do not offer. The temptation to keep them 'for later' produces a keyword map with nothing to promote against.
After cleaning, the list shrinks by half or two thirds. That is normal: a thousand queries of which eight hundred are junk is worse than three hundred clean ones.
Clustering without paid tools
Group by meaning and by page type: one group, one landing page. Test yourself with the question of whether a single page could answer every query in the group exhaustively. If not, split it.
Check borderline groups manually against live results: search two queries and compare the top five. If the same pages rank for both, they can share a page; if not, they need two.
Move finished clusters into the site structure before commissioning any copy. The reverse order — write first, assign queries later — produces pages that fully answer nothing.