How to Brief a Localized Article: Redo the Keyword Research in Each Market
By Jascha Kaykas-Wolff, former CMO of Mozilla and BitTorrent ·
To brief a localized article, localize the brief first: have a native-speaking editor redo the keyword research and read the top results in the target market, then hand the writer that brief alongside the English article. For pages that need to rank, transcreation from a localized brief is the best default. Machine translation with a native post-editor is fine for help-center content that only needs to be correct.
A US HR software company ranks well for “employee onboarding checklist” and decides to launch in Germany. The 1,600-word article goes to a translation agency, comes back two weeks later as clean German with “Checkliste für die Einarbeitung neuer Mitarbeiter” in the H1, and gets published under /de/. Six months on it has almost no impressions. The German HR managers it was written for search “Onboarding Checkliste,” mostly in English loanwords, and the pages they click on are full of things the American article never mentions: the works council, the probation period, the paperwork for the employment contract that German law wants in writing. The translation was accurate. It was accurate about the wrong country.
That is the most common way localization fails, and the translator is almost never at fault.
Our position, and it makes localization slower and more expensive, is that any localized page you expect to rank needs its own brief, researched in that market by someone who reads the language natively. Translating the finished article carries every English decision across the border, including the keyword. Search behavior does not travel that way. Spanish searchers in Madrid look for an “ordenador,” while searchers in Mexico City look for a “computadora,” and one Spanish translation will miss one of them.
What changes in a localized brief
Most of a normal content brief survives the trip. The audience carries over, and so does the argument you want the page to make. What has to be rebuilt is everything that came from looking at search results: the keyword, the questions people ask, the length and structure of the pages that rank, and the examples a reader will recognize. Those came from an English results page, and the German or Japanese results page for the same idea often looks different enough to change the outline.
Five ways to localize an article, compared
Teams usually pick the method by budget and then discover what it costs in traffic. Pick it by what the page has to do.
| Method | Best for | What the brief must add | Relative cost | Key trade-off |
|---|---|---|---|---|
| Machine translation, published as is | Nothing a customer reads. Internal drafts, gist checks. | Nothing, which is the problem | Lowest | Fast and nearly free. Reads like a translation, keeps the English keyword choices, and a native reader can tell inside two sentences. |
| Machine translation plus a native post-editor | Help-center articles, release notes, pages that do not need to rank | A glossary of product terms and the formal or informal address rule | Low | Accurate and cheap. Still built on the English outline, so it answers questions the local searcher may not be asking. |
| Human translation of the finished article | Legal pages, regulated claims, anything where the wording has already been approved | The approved source text and a do-not-change list | Medium | Faithful to the source. Faithful to the source’s keyword too, which may be a phrase nobody in that market types. |
| Transcreation from a localized brief | Most marketing articles that need to rank in a second market | In-market keyword, local SERP notes, local examples to swap in | Medium to high | Keeps the argument, rebuilds the evidence. Needs an editor who can say “this example means nothing here” and be listened to. |
| Native writer drafts from a localized brief | Pillar pages in a market you plan to win, topics with no English equivalent | Everything a normal brief has, researched in that market | Highest | The only option that can disagree with the English page. Costs about the same as writing it twice, because you are. |
Where each method fits (and where it does not)
Machine translation, published as is. Do not. DeepL and Google Translate are good enough now that the output is grammatical, and grammatical is the trap, because nobody on the team can see what is wrong with it.
Machine translation plus a native post-editor. The right call for support content. A customer who searches your help center in French already knows your product name and wants the answer to a specific error message, so the English structure is fine. Give the post-editor a glossary, or your product terms will come back translated three different ways in one article.
Human translation. Use it when the words have been through legal review and changing them would mean another round. Terms of service, or a claim about a security certification. Leave the SEO out of it.
Transcreation. Where most marketing teams should land. The editor keeps your argument and replaces the parts that only work in English: the keyword, the examples, the statistics drawn from US surveys, the jokes. The brief tells them which parts those are, so they are not guessing.
Native writer from scratch. Worth it for a handful of pages in a market you are serious about, and for topics where the local results page is about something else entirely. The German onboarding page is a decent example. A writer starting from the German results would have built the outline around the works council and the probation period, which the English article could never have handed them.
The honest concession, since we sell briefs: if you are testing whether a market is worth entering at all, post-edited machine translation of your five best pages is a reasonable experiment, and a localized brief for each one would be money spent before you know the answer. Brief properly once the test says yes.
Research the keyword in the market
Start from the English brief, then throw out its keyword. Ask the native editor what a person in that market would actually type, and check the answer in a keyword tool set to that country's database, the way our guide to keyword research for content teams describes. Expect English loanwords in B2B software. German and Dutch borrow heavily (“Onboarding” and “Recruiting” are ordinary German words in an HR department), and the formal native term often has a fraction of the searches.
Then read the top 10 results in that market, with search set to the country and language and a browser that is not logged in to your own account. Our SERP analysis guide applies unchanged once you are looking at the right page of results.
Here is where our own tool falls short. The ContentBrief.io brief generator is free and builds the English brief in about 30 seconds, with the headings and word counts of the top 10 results and the People Also Ask questions. Run your English keyword through it first and keep that brief as the source document. But the generator does not let you choose a country or a language for the search, so for a German or Japanese brief you need a results page from that market, read by someone who speaks the language. A brief built from American results and then translated is the onboarding page at the top of this article.
Fields a localized brief needs
- Target market, written as language plus country. Write “Spanish, Mexico” rather than “Spanish.” That pair is also the value your developers will put in the hreflang tag (es-MX), so write it the way they will.
- The in-market keyword, with who chose it. Name the native editor who confirmed it. If nobody did, the field is empty.
- What the local results page covers that the English one does not. One or two lines. This is the field that would have saved the German onboarding page.
- Address form. German “Sie” or “du,” French “vous” or “tu,” Japanese politeness level. Pick once per market and record it in the editorial style guide, because a site that switches between them reads like three different companies.
- Examples to replace. List every US-specific reference in the source article: 401(k), the IRS, Thanksgiving, prices in dollars, a survey of American marketers. For each one, say whether to swap it, cut it, or keep it with a note.
- Number and date formats. Currency causes the most trouble. Germany writes 1.200,50 € where the US writes $1,200.50, and a translated price table that keeps the American punctuation looks wrong to every reader in Berlin.
- Product glossary. Which interface terms stay in English because the product UI does, and which are translated. Pull this from whatever your localization team already uses for the app.
- Local claims that need review. Anything about law, tax, data protection or employment. GDPR is one regulation, but each EU country adds its own rules on top, and an English article written for Ohio is the wrong source for them.
- Internal links in the target language. Link to the localized versions of your other pages, or leave the link out. Sending a French reader to an English page mid-paragraph is the fastest way to remind them the page was translated.
Who should review the draft
A native speaker who lives in the market, or did recently, and who has never seen the English version.
Before it goes live
Add three checks to your pre-publish QA for localized pages: the hreflang tags point both ways (the English page names the German one and the German page names the English one), the meta title and description were rewritten in the target language instead of left in English, and every number with a decimal or currency symbol follows the local format. If the writer is a freelancer working into their native language, our guide to briefing a freelance writer covers the rest of the handoff. And if you are using a model for the first pass, the content brief for AI writing guide explains which fields to give it, which matters more here, since a model translating from English will keep the English structure unless the brief tells it what to change.