The fastest way to waste an international SEO budget is to translate your keyword list. Translation produces correct phrases; search demand lives in the phrases people actually type, and those are different things in every market.
Keywords do not translate. They regenerate. Here are the basics of doing it properly.
Translate the seeds, regenerate the list
The full pipeline: translate your core seeds, verify them with a native eye, then run them through autocomplete set to the target country and let the market phrase everything else itself.
Step 1: seeds, with a native check
AI translation is a fine seed factory, it is a language task, so ask for several ways a local would say each seed, not just the textbook one. Then have a native speaker glance at the list.
The glance matters because dictionaries lag streets. A language's real product vocabulary includes slang, borrowed words and regional picks no translator defaults to, and natives spot the term nobody uses in seconds.
Step 2: regenerate through local autocomplete
Set the country selector on the generator to the target market and run the native seeds. The suggestions come back localized: the market's own phrasings, in the market's own frequency order.
This step is where the trap-avoidance pays: you will watch the "correct" translation return three suggestions while the street term returns forty. The list writes itself in the vocabulary that wins.
Step 3: markets, not languages
One language is many markets, and the vocabularies diverge exactly where money lives: product names, service terms, buying words.
| Language | The divergence | Consequence |
|---|---|---|
| Spanish | Spain vs Mexico vs Argentina vocabularies | Generate per country, not once |
| Portuguese | Brazil and Portugal name things differently | Two lists, two validations |
| English | US, UK, Australia term and spelling swaps | Even "easy mode" splits |
| French | France vs Canada usage | Same drill |
The operational rule: a keyword list belongs to a country, not a language. Run the country selector per market and keep the lists separate in your master sheet, one market column deciding which pages each row feeds.
Step 4: validate with local numbers
Volumes are per-country too, so validate each market's list against that market's data: the free volume tools all take a country setting, and the same phrase can be a giant in one market and a ghost in the next.
While validating, watch the borrowed-English effect: many markets type English terms for tech, business and gaming topics, sometimes outsearching the native word. Check both forms and let the local numbers pick; the answer differs by niche and country, which is why assuming loses.
Step 5: content follows the same rule
Regenerated keywords deserve regenerated content: pages written for the market, not machine-translated from your originals. That is a bigger topic than this one, but the keyword principle extends cleanly, and search engines increasingly judge whether content reads native.
Start small the same way you would in any new niche: one market, its long-tail rungs first, and let real rankings fund the next market.
The one-line takeaway: translate ten seeds, verify them natively, then regenerate the list through that country's autocomplete and validate with that country's volumes. Per market, every time, because keywords are typed, not translated.