Multilingual Keyword Research: How to Build a Keyword List in Any Language
How to do multilingual keyword research in a language you don’t speak: build a native keyword list, compare search volume by country and language, and plan for both Google and AI search.
Multilingual Keyword Research: How to Build a Keyword List in Any Language
Multilingual keyword research means building a keyword list in each target language from the words local buyers actually type, not from translations of your English list. Doing it in a language you do not speak was always hard, and today it is harder and more valuable at the same time.
The multilingual opportunity is bigger than ever, because English is still only about half of web content and most non-English markets are underserved. At the same time, the game has shifted, because LLMs like ChatGPT, Claude, Gemini, and Perplexity have rewritten what “keywords” even means.
The shift shows up in every market, from Spanish e-commerce stores fighting native competitors to German B2B launches that need to appear in both Google and AI Overviews. Here is how to approach multilingual keyword research, with the data, the pitfalls, a step-by-step method for any language, and the tools that actually work.
The Multilingual Search Opportunity Is Bigger Than You Think
For businesses operating primarily in English, expanding into markets where English is not the first language remains one of the strongest plays in digital marketing. The reason is simple arithmetic.
English is used by 49.5% of the websites whose content language W3Techs can identify, followed by Spanish at 6.0% and German at 5.9%.Source: W3Techs, "Usage Statistics and Market Share of Content Languages for Websites", October 2026
The top eight languages look like this.
| Rank | Language | Share Of Websites |
|---|---|---|
| 1 | English | 49.5% |
| 2 | Spanish | 6.0% |
| 3 | German | 5.9% |
| 4 | Japanese | 4.9% |
| 5 | French | 4.5% |
| 6 | Portuguese | 4.1% |
| 7 | Russian | 3.4% |
| 8 | Italian | 2.8% |
Source: W3Techs, content language usage survey, checked October 2026. W3Techs updates this survey continuously, so exact percentages shift slightly from month to month; the ranking is the more durable signal.
English dominates web content, but every other language in the table is a market where buyers search in their own words. Fewer pages usually compete for each query there, even when absolute search volume is lower than in English, which is why a country-specific content strategy pays off.
How Direct Translation Quietly Breaks Your Keyword List
The most common mistake in foreign-language keyword research is building a keyword list from direct translations of your English terms. People do not search the way they write. Search queries skew toward slang, compound nouns, colloquialisms, and the single word most people actually say out loud in a given country.
Take the keyword “best washing machines” as a working example. Run through Google Translate for the French market, the term comes out as meilleures machines à laver. Enter it into Google.fr with your browser language set to French (the guide to changing Chrome language helps), and the top-ranking pages actually optimize for meilleurs lave-linges.
Lave-linge is the compound noun French natives use in daily life, and it is the version that native French SEO writers target. The same pattern repeats across most European languages, though the specific mismatch varies.
| Target Language | Google Translate Output | Native-Preferred Term | Common Issue |
|---|---|---|---|
| French | meilleures machines à laver | meilleurs lave-linges | Compound noun preferred in daily speech |
| German | beste Waschmaschinen (plural) | beste Waschmaschine (singular) | Singular form dominates commercial intent searches |
| Spanish | mejores lavadoras | mejores lavadoras | Rare case where direct translation matches |
| Italian | migliori lavatrici | migliori lavatrici | Rare case where direct translation matches |
| Dutch | beste wasmachines | beste wasmachine | Singular preferred, similar to German |
Illustrative examples of the pattern. Search results change, so always verify a term in the target market before committing to it.
The practical move is to validate every translated keyword in a local SERP before trusting it, ideally with a native speaker reviewing the output. Translation, localization, and SEO belong in the same room for exactly this reason.
Comparing Search Volume by Language and Country Without Getting Fooled
If you do not have a deep feel for the language and culture you are researching, you lean more heavily on tools to assess demand. Keyword tools report an estimated monthly search volume for any given term, and the more advanced ones let you filter volume by country and by language, which matters in multilingual markets like Belgium, Switzerland, or Canada.
The catch is that tool data is never exact. Third-party tools do not see Google’s full search logs, so the reported figures are estimates built from sources such as Google’s advertising data, clickstream panels, and statistical modeling.
The good news is that the error is usually consistent across countries and languages, which makes the tools reliable for comparison even when the absolute numbers are fuzzy. If you know the English version of a keyword converts at a certain volume, you can use the proportional ratio to estimate the opportunity in French, German, or Spanish.
Google Keyword Planner, free with a Google Ads account, lets you set both a location and a language for every lookup. Set both every time, because Google search volume by language and by country are different numbers in places like Montreal, Brussels or Zurich.
The deeper point is that lower volume in another language often still beats higher volume in English, because competition is thinner and conversion intent tends to be stronger when the page is genuinely localized.
Source: Statista, Most Common Languages On The Internet. The chart shows the January 2024 edition; the latest edition (October 2025) keeps the same top three of English, Spanish, and German.
Long-Tail Keywords and the LLM Shift That Changed Everything
Here is the part most older SEO articles miss, because it only came fully into focus in 2025 and 2026. Users are not searching the way they did five years ago. The rise of ChatGPT, Claude, Gemini, and Perplexity has trained people to type full, natural-language questions instead of fragmented keywords, and the shift has fed back into Google as well.
A Nectiv analysis of more than 8,500 prompts found that ChatGPT ran a web search in 31% of them. On average it ran 2.17 searches per prompt, each 5.48 words long, roughly 60% longer than the 3.4-word U.S. average for Google.Source: Search Engine Land, "ChatGPT Performs A Search In 31 Percent Of Prompts" (Nectiv study), October 2025
Those individual LLM queries keep getting longer.
Peec AI analyzed more than 20 million ChatGPT query fan-outs between October 2025 and January 2026. The average rose from about 6 words in October to about 12 words by January 2026, with a peak of about 16 words around calendar week 49.Source: Peec AI, "ChatGPT fan-outs have doubled in length in 4 months", February 2026
Google’s AI features lean the same way. Ahrefs found that AI Overviews trigger overwhelmingly on informational, longer queries.
99.2% of the keywords that trigger an AI Overview are informational in intent, and the typical AI Overview keyword is 4 words long versus 2 words for keywords without one.Source: Ahrefs, "I Analyzed 300K Keywords. Here’s What I Learned About AI Overviews", October 2024
Query length also grows as buyers get closer to a decision.
Prompts get longer as buyers move closer to a purchase, rising from 11.5 words for informational searches to 18.2 words for branded searches.Source: Position Digital, "100+ AI SEO Statistics and Insights for 2026" (Position Digital data, September 2026)
The practical consequence for multilingual keyword research is huge. You are no longer researching two or three-word fragments, you are researching entire questions your target audience might ask in their own language, often in full sentences with context about budget, use case, and constraints.
Here is what the same commercial intent looks like in traditional search versus in AI search, side by side.
| Dimension | Traditional Google Query | LLM / AI Search Query |
|---|---|---|
| Typical length | About 3 words | About 5 words per ChatGPT search; fan-outs of 12 words or more |
| Format | Fragmented keywords | Natural-language question |
| English example | best washing machine | what is the best quiet washing machine for a small apartment under 500 euros |
| French example | meilleur lave-linge | quel est le meilleur lave-linge silencieux pour un petit appartement à moins de 500 euros |
| German example | beste Waschmaschine | welche leise Waschmaschine ist die beste für eine kleine Wohnung unter 500 Euro |
| Intent clarity | Medium | High (full context in the query itself) |
| Content that tends to win | Short, heavily on-page-optimized pages | Complete, well-structured, citation-friendly content |
Side-by-side query patterns. Long-tail in the target language is now a commercial opportunity as well as an AI-visibility opportunity.
For foreign-language markets, long-tail keyword research is not a side activity anymore. It is the core of the strategy, because LLMs tend to pull their answers from content that matches the shape of the question, and that question is almost always long, conversational, and written in the user’s own language.
Our piece on machine versus professional translation explores why that shift reshapes every multilingual content plan.
NLP and Why Search Engines Now Understand What You Mean
One quiet reason the LLM shift works is that search engines themselves got much better at natural language processing years before the current AI boom. Google’s BERT update in 2019 and the MUM model that followed it taught search to read context, not just words.
Google’s newer AI Overviews and AI Mode go further, processing the full meaning of a query across languages, identifying entities, and weighing semantic relevance far more than exact-match keywords.
For your keyword research, the implication is straightforward. You do not need twenty near-duplicate pages each targeting a small keyword variation, you need one thorough page that covers the topic and its natural sub-questions in the target language, in a structure search engines can parse. Semantic depth beats keyword stuffing, which is the core argument in our piece on multilingual websites.
How to Build a Keyword List in Any Language
The same six steps work whether the target language is Spanish, Japanese or Polish, and whether or not you speak it.
- Brainstorm native seeds. Start from the top-ranking local pages, Google autocomplete set to the right country and language, and local marketplaces or forums, rather than from your English list.
- Translate and verify in the SERP. Translate your English terms as a cross-check, then search each candidate in the local Google and keep the version the ranking pages actually use.
- Add long-tail AI queries. Write out the full questions a buyer in that market would type into ChatGPT or Google, in their own language and with their own constraints.
- Native speaker review. A native linguist removes false friends, awkward literal forms and regional mismatches before any content is written.
- Publish and track in Google Search Console. Filter the performance report by country and look at the queries each page already earns impressions for.
- Refine from real data. Move budget and content toward the terms that convert, and drop the ones that only looked good in a tool.
The steps compound, so skipping one makes the later ones underperform. Brainstorming in the native language first keeps the logic clean, translating early exposes the translation traps, and AI long-tail expansion covers LLM query patterns.
Native review then removes the last mile of error, and Google Search Console data finishes the job once the pages are live. For the site-level side, such as hreflang and URL structure, our guide to multi-language SEO for websites picks up where keyword research stops.
Example: Spanish Keyword Research for Spain and Mexico
Spanish keyword research starts with picking the country, because everyday words differ. A buyer in Spain searches for an ordenador or a móvil, while a buyer in Mexico searches for a computadora or a celular.
Build one keyword list per country rather than one list for "Spanish," and check each term in Google.es and Google.com.mx separately. Our Spanish SEO services page covers how those lists turn into localized pages.
Example: Keyword Research in Canada
Canada is a bilingual market, so keyword research there runs twice, once in English and once in Canadian French. Set the location to Canada and switch the language in each lookup, then check the French terms in a Quebec context.
Canadian French has its own search habits. A shopper in Quebec writes courriel for email and magasiner for shopping, where a reader in France would often write mail and faire du shopping.
Work with a Native Translator to Verify Your Keywords
High-quality keyword research in an unfamiliar language is possible without a native speaker, but it is slow, error-prone, and easier to get wrong than to get right. Mistakes that a local linguist spots in three seconds can sit in an SEO strategy deck for months before anyone notices why the traffic plateaued.
Running your final shortlist past a professional translator catches the colloquial misses, the cultural false friends, and the regional variants that change keyword intent between Mexico City and Madrid, or between Paris and Brussels. Native translators also tend to surface long-tail AI queries that a non-speaker would never guess, because they know how people in that market phrase a buying decision or a technical question.
76% of online shoppers prefer to buy products with information in their native language, and 40% will never buy from websites in other languages.Source: CSA Research, "Consumers Prefer their Own Language" (press release on the "Can’t Read, Won’t Buy: B2C" survey of 8,709 consumers in 29 countries), July 2020
A local term that nobody types costs as much as a mistranslated one. Whether a page gets found depends on search demand as much as on translation accuracy.
Ahrefs’ study of roughly 14 billion pages found that 96.55% of pages in its index get zero traffic from Google, and it names missing search demand as one of the main reasons.Source: Ahrefs, Search Traffic Study
Keep the translator involved through content creation, not just keyword validation. A writer who understands the keyword strategy and the audience in one head produces a better final page than one who handles either piece in isolation, which is the argument we make in our human-in-the-loop approach to AI-assisted translation.
Use Google Search Console and LLM Visibility Tools for Real Data
Once a site is live and has been crawled, you stop guessing. Google Search Console reports the queries that are landing users on each page, with country and device breakdowns that third-party tools cannot match. The data comes straight from Google, so it puts you on the same footing as someone working in their native language, at least for traditional search.
For AI search, the tooling is newer but catching up fast. AI visibility platforms like Profound and Peec AI, and the AI visibility tools inside Semrush (the AI Visibility Toolkit) and Ahrefs (Brand Radar), track whether your brand is mentioned or cited in answers from ChatGPT, Perplexity, Gemini, and Google’s AI features. Coverage differs by tool, so check which platforms and languages each one supports.
The mindset shift is the same in both cases. Stop trying to perfect keyword research before a page exists, and start treating the first version as a probe.
Aim to rank in the right ballpark, then let real Google Search Console data and LLM visibility reports reveal the better long-tail keywords you never would have guessed.
Where BeTranslated Fits into Your Multilingual Keyword Strategy
The agencies and in-house teams we work with tend to have the SEO muscle already. What they often need is a partner who can carry the keyword strategy into French, Spanish, German, Dutch, Italian, or Portuguese through our language translation services, without losing the cultural nuance or the long-tail intent.
Our team combines native linguists with hands-on international SEO experience, which is why we can deliver French SEO, Spanish SEO, German SEO, and Dutch SEO work that reads native and ranks natively.
For website projects, our WordPress translation and WPML workflows keep hreflang, canonicals, and language variants tidy enough that the keywords you research actually translate into rankings. When the work extends into paid channels and campaigns, our marketing translation and advertising translation teams carry the same keyword discipline through to ad copy and landing pages.
The wider strategy playbook sits in our brand localization strategy guide.
Ready to Research Keywords That Actually Convert in Your Target Language?
Whether you need a single landing page researched and translated, a full multilingual site built from scratch, or keyword validation support for an in-house SEO team, we can help. Tell us the target market and the core intent, and we will come back with a clear scope, a realistic timeline, and a quote.
Request a free, no-obligation translation and SEO quote, call our Valencia office at +34 962 02 22 22, or write to hello@betranslated.com.
Frequently Asked Questions about Multilingual Keyword Research
Can I Do Keyword Research in a Language I Don’t Speak at All?
Yes, up to a point. You can brainstorm in English, translate carefully, verify each term in the local SERP, and lean on keyword tools for relative volume data, but the final step almost always needs a native speaker to catch colloquial mismatches and long-tail variations.
How Has LLM Search Changed Keyword Research in 2026?
Users now type longer, more natural questions into ChatGPT, Perplexity, and Google’s AI features, which means longer, question-style phrases matter far more than two or three-word fragments.
Your keyword research needs to include full question phrases in the target language, not just short fragments.
Can AI Translation Replace a Native Translator for Keyword Work?
AI translation is useful for the first pass, but it still misses the colloquial forms and compound nouns that native speakers actually search. A practical workflow is AI-assisted first drafts with a native linguist validating keywords, anchor text, and content before anything goes live.
Are Keyword Tools Accurate for Non-English Languages?
Tool data is never exact in any language, including English, but it tends to be consistent across markets. Use the numbers for relative comparison rather than absolute volume, and confirm each top keyword in a live local SERP before committing content to it.
How Long Before Multilingual Keyword Research Shows a Return?
Newly targeted foreign-language keywords often take three to six months to bring meaningful organic traffic, assuming the pages are well-localized and earn a few relevant backlinks. Competitive markets and brand-new domains can take longer, especially when the site has to build regional trust from scratch.
What Is the Best Multilingual Keyword Tool?
Google Keyword Planner is the free starting point, because it lets you set a location and a language for each lookup. Ahrefs and Semrush add country-level keyword databases, difficulty scores and competitor data, and Google Search Console shows the real queries once pages are live.
No tool replaces a native check. Use the tools to compare markets and a native linguist to choose the words.
Should I Target Short-Tail or Long-Tail Keywords in a Foreign Language?
Long-tail, almost always, especially when entering a new market. Short-tail keywords are dominated by established native competitors, while long-tail queries carry higher intent, face less competition, and align with how LLM-era users actually phrase their questions.
