How to do keyword research in a second language
Build a keyword list for Spanish or French search when the usual tools report almost no volume for the phrases your customers really use.
The tools report nothing and the customers are real
A home services firm in Texas takes calls from customers who search in Spanish. A clinic in Montreal takes enquiry forms in French. The phrases go into a keyword tool and come back as zeroes. The enquiries arrive in exactly those words. The conclusion drawn is that there is no demand. The tool simply cannot see it.
Volume data for non-English queries is sparse and inconsistently reported. An accented spelling, the same phrase typed without accents and the same phrase spoken aloud may be three records or one, depending on the tool. Some tools fold the query into its English equivalent. Research here has to be built from observation rather than from a volume column.
What this work does not include
Keyword research in a second language excludes translation of your website, and translation is where the real cost sits. A list of Spanish phrases is useless without pages written in Spanish by somebody who writes it natively. Machine translated pages read badly and tend to be treated as low quality by search engines. If you will not fund native writing then stop at the research and apply the findings to your English pages.
It also excludes the decision about which audience you are writing for. Spanish written for a Mexican American buyer in the south west is not the same as Spanish written for a Cuban American buyer in Florida. Serving both properly means more pages. Your own time to listen to calls and read enquiry messages is the main input. Rates are stated before tax. Sales tax on marketing services varies by state.
Who needs this, and who is wasting money on it
This matters where the purchase decision is made by somebody who thinks and reads in the language. Home repair and trades, tax preparation, dental and family practices, driving schools, childcare, insurance sold across a kitchen table. It also matters for any business whose call recordings show customers struggling with English product names.
It is a waste for a business selling to procurement departments or to any buyer who works in English all day. A software company, an export consultancy or a corporate law firm will find that its buyers search in English even when another language is spoken at home. Building second language pages for those buyers produces pages nobody reads, dilutes the site and is money better spent on proof.
How to build the list without volume data
Start with your own records. Read six months of enquiry forms, call notes, your site search box queries and social comments and extract the exact words customers used for the product, the problem and the price question. The mixed forms are the data. A customer asking for the cheapest water heater in a sentence that is half Spanish and half English is showing you the query.
Then expand using autocomplete in the language. Set the browser and search language to the target language, type the phrase with accents and again without them and record every suggestion. Do the same inside YouTube. Then check the People Also Ask results and the related searches at the foot of the page, which are generated from real queries and cost nothing.
Then group what you have by meaning rather than by spelling. One intent will have several spellings and you cannot build a page per variant. Pick the spelling your customers use most in their own messages, write the page around it and include the main variants naturally in the text. Finally rank the groups by proximity to a purchase, since that ordering is the only prioritisation available without volume.
Judging the result
The metric that matters is enquiries in the language, counted per page. Ask whoever answers the phone to record the language of each caller for a month. If the Spanish pages are working then the mix shifts. Pair that with the query list in Search Console filtered to those pages, which shows which spellings bring clicks.
The metric that flatters is the count of keywords you now rank for. Accent and spelling variants multiply that figure quickly and most carry no traffic. A report showing hundreds of new ranking keywords and no change in Spanish speaking callers is reporting spelling permutations.
The common mistake
The expensive error is translating existing English pages word for word. The English page was written around English queries so the translated version targets phrases nobody types. It also carries the English page's structure, which rarely matches how the question is asked in the other language. The result ranks for nothing and reads like a manual.
The second error is building one page that mixes both languages to cover both audiences. It serves neither. Search engines struggle to identify the page's language and the reader sees half a page they can use. Separate pages, separate URLs, one language each.
A caveat on data quality
No tool currently gives reliable volume figures for non-English queries and no figure quoted for the share of searches run in another language should be accepted without a named source and a date. Build from your own customer language, publish a small set of pages and let the query reports tell you what to do next. That is slower than buying a keyword list and it is the only method that rests on evidence.
This guide explains how something works. It is not legal, tax or financial advice, and rules change. Confirm anything regulatory with a qualified professional before you act on it.