International Keyword Research

International keyword research runs on one founding fact this silo keeps citing: demand doesn't translate — it re-originates. Every market searches from its own vocabulary, habits, platforms and season, so the research process rebuilds per market rather than porting your home list through a dictionary. The standard process supplies the machinery; this guide supplies the per-market adaptations, the translation traps, and the prioritisation that decides which markets' research happens at all.

Why translated keyword lists fail

The failure catalogue from real audits: synonym mismatch (the dictionary translation is correct and nobody searches it — the market's query uses the other synonym, the borrowed English word, or the brand-name-as-generic); intent drift (the equivalent phrase carries different intent — commercial in one market, informational in another — verified only by the local SERP read); structure difference (compound-word languages turn phrases into single tokens with their own volumes; question phrasing conventions differ); and volume illusions (the translated term's volume measured while the actual local phrasing's demand sits unmeasured beside it). Translation gives you seeds; research gives you the map.

The per-market process

1. Seed bilingually: your translated terms as starters, expanded immediately by native input — the market's own autocomplete (typed in-locale via location-set sessions), local competitor sites' vocabulary (their nav and titles are pre-validated local phrasing, per the export workflow run on local players), and the native speakers you already retain per the editor-pool doctrine — an hour of "what would you actually type?" outperforms any tool pass.
2. Volumes from tools with local coverage, read with the standing caveats squared — minor-market data thins across every tool (the regional-coverage test applies: sample your market before trusting a database), and Keyword Planner's per-country view is often the most honest floor available.
3. The local SERP read per money term: intent, formats and incumbents in the market's own results (location-set, per the honest-checking rules) — where the winnability surprise usually lives: many non-English SERPs are years less competed, the thin-incumbent signatures from the anomaly files everywhere.
4. Map per market: each market's own keyword map — because taxonomies diverge (the German market may slice the category differently than your home nav does), and the map is where localisation stops being translation, per the re-origination rule.

Prioritising across markets

The portfolio math from the sequencing doctrine, research edition: markets ranked by demand (total mapped volume × commercial fit), winnability (the SERP reads' verdict — the less-competed markets jump queues here), and operational readiness (a market you can't fulfil or support is research wasted) — with your own analytics as the pre-vote (existing foreign traffic and query data showing where demand already found you). Research one market deep per cycle; the ten-market keyword spreadsheet nobody actions is this discipline's version of the empty-folders failure.

Frequently asked questions

Can I do keyword research for a language I don't speak?

The machinery, yes (tools, SERP structure, volumes); the judgment, no — synonym choice, intent reading and phrasing naturalness need the native hour at minimum. The working split: you run the process, a native validates the map — the same collaboration model as tier-two translation.

Do non-Google engines change the research in some markets?

Where a market's search share genuinely splits (the handful of famous cases), the local engine's own data and conventions join the process — a per-market check worth making once. For most markets Google's dominance makes the standard process portable; the localisation is in the language layer, not the engine layer.

Our translated pages rank for nothing. Research or translation problem?

The diagnostic fork: impressions present but clicks absent → phrasing/title mismatch (research findings not applied to the page); impressions absent entirely → targeting the wrong terms (the translated-list failure — remap from local research) or the authority cold-start per market. Usually all three in layers — fixed in that order, with the local link layer as the recurring missing piece (our piece).