The keyword research tool most marketers open first cannot see Uzbek, the state language of a country with 33 million internet users by DataReportal’s Digital 2026 count. Google Ads, whose Keyword Planner is the usual starting point, publishes the list of languages an advertiser can target. It has 51 entries. Russian is on it, and Uzbek is not. So the keyword list has to be built by hand, from what the search box itself suggests, in two scripts and on two search engines. I did that this week for one commercial query, rus tili kursi, which is Uzbek for “Russian language course”, and the six steps below are the method, with what each step turned up.
1. Type The Seed The Way People Actually Type It
Open Google with the interface set to Uzbek and the region set to Uzbekistan, type the seed one word at a time, and write down every suggestion. The first surprise is the apostrophe. Official Uzbek Latin writes oʻ and gʻ with a dedicated character, the modifier letter turned comma, which almost no keyboard carries, so people type a straight apostrophe, a curly quote, or nothing at all.
Google’s autocomplete treats those as different strings. On 23 September, “rus tilini o’rganish” (learning Russian) typed with a straight apostrophe returned suggestions about books, PDFs and films. The same phrase with the official ʻ returned “tez va oson” (fast and easy) and “0 dan” (from zero).
A third spelling, with a space where the apostrophe belongs, turned up inside the suggestions themselves: “rus tilini noldan o rganish”.
This step goes wrong in the spreadsheet. A tool that deduplicates will merge the variants, or keep one and drop the rest, so keep every spelling as its own row until you know which ones carry traffic.
2. Run It Again In Cyrillic, Twice
Uzbek’s current Latin alphabet dates from 1995, and a 2021 government plan set 1 January 2023 as the date to finish the switch. Cyrillic is still in daily use alongside it, which means two more passes: Uzbek written in Cyrillic, and plain Russian. The Uzbek Cyrillic seed “рус тили” came back with “рус тилини урганиш”, spelled with у where the dictionary wants ў, the same keyboard shortcut as the missing apostrophe. The Russian phrase for “Russian for Uzbeks” pulled a different intent altogether: PDF, download, and “разговорный”, conversational.
The mistake here is filing Uzbek Cyrillic under Russian because the script looks the same. It is the same language as your Latin list and belongs next to it.
3. Check The Engine That Holds a Quarter Of The Market
Yandex is not a rounding error in Uzbekistan. StatCounter’s figures for August 2026 put Google at 73.9 percent of searches and Yandex at 24.0 percent, and over the previous twelve months Yandex moved between 20.4 and 28.8 percent. Its suggestions are not a copy of Google’s, either. Yandex’s public suggestion service returned nothing for “rus tili kursi”, and for the shorter “rus tili” it offered a list that included school material: tests, a grade 8 textbook, noun gender. None of those three appeared in Google’s top ten for the same prefix. Skip this step and you are building for three searchers in four without ever seeing what the fourth one typed.
4. Sort The Suggestions By What The Searcher Wants
A raw suggestion list for this topic is loud with free intent, and a course page will not rank for someone hunting a PDF. Here is how the suggestions from steps 1 to 3 split.
| Intent | Suggestions that showed it | Page that fits |
|---|---|---|
| Buy a course | rus tili kursi toshkent, rus tili kursi samarqand, online rus tili kursi, rus tili kurslari | Course page or city landing page |
| Free material | rus tilini o’rganish uchun kitoblar pdf, bepul dastur (free app), kinolar (films), скачать | Resource article |
| Lesson series | rus tilini 0 dan o rganish 1 dars, 2 dars, 3 dars | Numbered lesson series |
| Exams and admission | rus tili milliy sertifikat, rus tili kirish ballari 2026 | Exam guide |
Only the first row is a buyer. The other three are where a course brand earns the trust that makes the first row convert, so a content plan for this market needs all four. It goes wrong when the commercial row gets the whole budget because it looks like money, and the lesson-series row, the one with numbered demand written into the query itself, gets nothing.
5. Estimate Volume Without a Planner That Can See Uzbek
Without Keyword Planner, volume becomes relative. Google Trends compares whatever strings you give it, in any script, so set the region to Uzbekistan and put the step 1 variants against each other, then the winners against their Cyrillic twins. What you learn is which spelling belongs in the title, and that is the decision that matters.
Treat any third-party volume figure for an Uzbek string as a guess until Search Console confirms it. The Queries report is where real Uzbek search data first appears, and it only arrives after you publish.
6. Give Each Language Its Own Page And Tell Google They Are Twins
Once the clusters exist, each language needs a page written for its own query, and the pages need hreflang so Google serves the right one. Prep.uz, an online Russian course for Uzbek speakers that pairs live mentors with a video platform, is a clean public example. Its blog lives at /blog in Uzbek and /ru/blog in Russian, and every article declares its uz and ru versions, plus an x-default, in the page head.
The titles are the part worth studying. Its Uzbek beginner’s guide is titled “Rus tilini noldan o’rganish”, in the learner’s own words.
The Russian twin is not a translation of that line. It is “Русский с нуля: с чего начать и в каком порядке”, built on “русский с нуля”, a phrase Google suggests to Russian-language searchers in Uzbekistan. Two pages, one topic, two queries.
The common failure is a machine-translated twin. The hreflang is valid, the page is indexed, and it targets a phrase nobody types.
When The Right Page Still Does Not Rank
Sometimes Search Console shows Latin-script Uzbek queries landing on the Russian page, or the reverse. Check the hreflang block on both pages first. Google’s own documentation on localized versions is blunt about it: if two pages do not both point to each other, the tags are ignored. If the tags are reciprocal and the wrong page still wins, the title is the problem. Rewrite it with the exact spelling from step 1 that Trends favoured, apostrophe and all, and give it a few weeks before you judge the result.
