Language-specific keyword research.

Keyword research is done separately for every language, because the term searched for the same service in different markets may not be the direct translation.

This is the most frequent and most expensive error on a multilingual site: translating the Turkish keyword list and counting it as the target list. The translation can be grammatically correct while nobody in that market searches for the term.

The reason is simple: search terms form culturally. Which word settles in an industry depends on that market’s own history and on the content written in that language — not on its counterpart in another language.

Why translation alone is not enough

It produces three separate problems, and all three stay invisible because the translated page looks right.

The first is a term difference: a service’s official name and its market name can diverge, and translation usually produces the official one. The user searches for the market term.

The second is a volume difference: both terms may be used in that language, but one is searched many times more than the other. Translation does not know which one is dominant.

The third is an intent difference: the same word can carry different search intent in two languages. A query seeking information in one language can be a commercial query in the other — and the page’s type should change accordingly.

How it is done

1. Search in that language — on the target market’s search engine, in that language, from that region if possible.
2. Look at the competitors — which terms do the sites ranking in that market use?
3. Ask a native speaker — the cheapest and most reliable verification. “How would you search for this?”
4. Read the autocomplete — the search box’s suggestions come from real queries.
5. Use the search console — if you already get impressions in that language, the data is yours.
6. Verify the translation, do not source from it — translation is a starting point, not a result.

The third item is the most skipped and the highest-yielding. Asking a native speaker “how would you search for this service” gives an answer faster and more accurate than most tools.

The fifth item is a head start: if the site is already live in that language, the search console says which queries earn impressions — and that list comes from fact, not guesswork.

Slugs and titles derive from here

The research’s output is not just a list; it also determines the address structure and the page titles. Whichever term is searched in that language, the slug and title should carry it.

On this site the slugs are localized separately in three languages, and each carries that language’s term. The decision raises the translation load but puts that market’s own word into the address bar.

And a consistency rule: title, slug and page content should use the same term. If the three diverge, the signal blurs for user and engine alike.

Three frequent errors

There are recurring errors in language-specific research, and all three come from the same root: carrying the source language’s assumptions into the target language.

The first is the volume assumption: assuming that the counterpart of a much-searched Turkish term will also be much searched in German. Market sizes, competition levels and search habits differ — a word that is the main term in one language can stay niche in the other.

The second is the structure assumption: something searched with two words in Turkish can be searched with a single compound word in German. Word-for-word translation never produces that compound, and the page misses the real term.

The third is the competition assumption: building a strategy without knowing how competitive a term is in the target market. A term that ranks easily in the source language can be crowded with established authorities in the target market.

1. Measure the volume in that language — do not guess from the translation.
2. Search for compound words — single-word counterparts, especially in German.
3. Check the competition — who ranks for that term?
4. Verify the intent — the same word can carry different intent in two languages.
5. Ask a native speaker — the cheapest and most reliable verification.

The last item bears repeating because it is the most skipped and highest-yielding step. Asking a native speaker “how would you search for this service” gives an answer faster and more accurate than most tools — and its cost is one message.

When to do the research

The timing is often wrong: the research is done AFTER the translation is finished, and at that point the text is already written. The correct order is the reverse — the research comes before the translation, and the translation is written to the terms found.

That order also creates a cost difference: research done afterwards requires the texts to be rewritten. Research done beforehand produces only a term list, and the translator writes to it.

1. Research first — extract the target market’s real terms.
2. Prepare a term glossary — tell the translator which words to use.
3. Then translate — let the text be written to the terms found.
4. Set slugs and titles by the term — let all three carry the same word.
5. Measure after release — the search console shows the real queries.
6. Update the list — terms change over time.

The fifth item builds a feedback loop: after release, which queries earn impressions becomes visible, and that data feeds the next round’s research. The first round starts with estimation, the second with measurement.

And a caution: on a freshly published site that data does not exist. This site’s search console currently shows a total of one impression — meaning measurement has not begun yet, and the term decisions rest on research.

A frame to close: this research is not a one-time job. Terms change over time, new words settle in and the competition shifts. A light scan repeated once a year suffices to keep the list current — and when that scan is fed with the past year’s impression data, it has moved from estimation to measurement.

A final scope note: the research is not limited to keywords. What questions people in that market ask, what worries they voice and what comparisons they make should be researched too — because the content plan really comes from there.

The word list does not say what to write; the question list does. When the two are gathered together, both the terms and the topics have been set by the target market.