Translating a keyword list is the single most common way international SEO programmes waste a quarter. The output looks like a keyword list, it has volumes attached, and it is measuring the wrong terms.
The method below replaces that with something repeatable: propose many candidates, measure them where they matter, cluster by intent, then map one intent to one page.
Step 1: candidate generation
Start from intents, not strings. For each source keyword, write down what the searcher wants, then ask a native speaker or a language model to produce every plausible way that market phrases it. You want breadth here, including the terms you suspect are wrong.
- The formal or official term.
- The English loanword, which often wins in tech, finance and marketing.
- The colloquial or trade-slang term.
- Compound and separated variants in German, Dutch and the Nordic languages.
- Common misspellings and abbreviations, if they carry real volume.
Step 2: measure in-country
Pull search volume, cost per click and competition for every candidate, scoped to the target country and language. Global volume hides exactly the difference you are trying to detect, and cost per click is a useful proxy for commercial intent when volumes are close.
Smart GSEO Translator runs this pass automatically on every job, so the report shows you the source term, each candidate and its measured demand side by side.
Step 3: cluster by intent
Group candidates that would be satisfied by the same page. If three phrasings all want a definition, they are one cluster. If one of them wants a price comparison, it is a different page.
The output of this step is a table with one row per intent: the winning term, the supporting terms, the questions people ask around it, and the page that will own it.
Step 4: map to pages and check the gaps
Now compare the intent table to your existing site structure in that language. Two things usually appear: intents with no page, which is your content backlog, and pages with no intent, which are translations you did not need.
Prioritise by expected value, not volume. A 300-search term with high commercial intent and weak competitors beats a 20,000-search term owned by three marketplaces.
Step 5: validate with real query data
Sixty days after publishing, open Search Console, filter by country and read the queries the page actually receives. This is the only unbiased feedback loop you get, and it routinely surfaces phrasings no tool suggested.
Feed the surprises back into the glossary so the next batch of pages starts from a better prior.
Frequently asked questions
Can I just translate my English keyword list?
No. Translation preserves meaning but not demand. A literal equivalent frequently has a fraction of the volume of the phrasing that market actually uses, so the page ends up optimised for a term nobody searches.
Which country setting should I use for volume data?
The country you are targeting, combined with the language. Global volume averages away regional differences, and Spanish, Portuguese, French and Arabic all differ enough between countries to change the winning term.
How many keywords should one page target?
One primary intent plus its supporting phrases and questions. Trying to make a single page win two distinct intents usually loses both.
How often should multilingual keyword research be refreshed?
Refresh the winning terms every six months, and always after a product or category rename. Check Search Console query data 60 days after any new page goes live.
Written by Jasper Verelst
SEO consultant with 10+ years in search content and 3 years writing for generative engines. Creator of GSEO Editor and Smart GSEO Translator. About Jasper · LinkedIn · Full profile
Keep reading
Take it one level further in GSEO Editor
Your translation is market-ready. GSEO Editor, our sister tool from the same team, takes the finished page and optimizes it deeper for Google rankings and citations in AI answers: entity coverage, internal linking, and full-page scoring.