Briefing a translator so the page actually ranks

Localisation 11 September 2026·9 min read

A translator's job is to preserve meaning. An SEO's job is to match how people search. Those two goals disagree more often than anyone expects, and the gap between them is where most multilingual content quietly fails.

The failure, precisely

You send a good English page to a good translator. They return a German page that is accurate, fluent and pleasant to read. It ranks for nothing.

Nobody did anything wrong. The translator rendered your English keyword into the most faithful German equivalent — which is frequently not the term German buyers type. Search behaviour is not a translation of search behaviour. It is its own thing, formed by local competitors, local habits and the structure of the language.

This failure is unusually hard to catch because there is no error to find. The page is correct. It simply targets a phrase with no demand behind it.

Eight things a translator cannot know

Not won't — cannot, unless you tell them. None of these are visible in the source text:

  • Which term actually has search volumeTwo correct translations can differ by an order of magnitude in monthly searches. Nothing in the English page indicates which.
  • What the page is competing againstThe local pages currently ranking set the expected depth, structure and tone. A translator has not seen them.
  • Which words must appear in the H1 and titleAnd which must not be varied for style, however repetitive it looks to a careful writer.
  • Which links are structuralInternal links are frequently dropped or re-pointed during translation because they read as incidental.
  • That the meta description has a length budgetAnd that a faithful translation of a 155-character English description is routinely 20 per cent longer in German or Spanish.
  • Which proper nouns and product terms stay in EnglishTranslating a product name or an industry term of art that the market uses in English is a common and costly error.
  • The register the brand uses in that marketFormal or informal address is a positioning decision, and inconsistency across a site reads as carelessness.
  • That schema values are contentStructured data fields get overlooked entirely, so the page ships with English values in its markup.
Two panels side by side. The left, headed 'in the source page', lists four things a translator receives: source text, headings, links as markup, and the meta description. The right, headed 'not in the source page', lists the eight they need: target keyword and variants, search intent, competitor URLs ranking locally, fixed terms for H1 and H2, a do-not-translate list, register, elements that must survive unchanged, and title and meta length budgets.
Everything in the right-hand panel has to be supplied. None of it can be inferred from the page being translated.

The brief that closes the gap

Everything above becomes a line in the brief. This is the structure I use, per page:

SEO localisation brief — required fields

Target keyword in the destination language, with the two or three variants that are acceptable substitutes, taken from local search data rather than from the source text.
Search intent in one sentence: what the person wants when they type this, which may differ from the intent behind the English term.
Two or three competitor URLs currently ranking locally, with a note on what they cover that the source page does not.
Heading structure, with the terms that must appear in H1 and H2s marked as fixed.
Do not translate: product names, industry terms the market uses in English, and any term where the English is the searched form.
Register: formal or informal, stated once, applied everywhere.
Must survive unchanged: internal links and their anchors, calls to action, schema field values, tracking parameters.
Length budgets: title and meta description character limits, stated as limits rather than suggestions.
Local additions: anything this market needs that the source page lacks — local trust signals, market-specific FAQs, regulatory wording.

That last field is the one that turns this from translation into localisation. The best localised pages are not translations at all; they are the source page plus what the local market needs and minus what it does not. A brief that only ever subtracts will produce a page that ranks below the local competition by design.

Who reviews what

Three distinct checks, and conflating them is how errors get through:

  1. Linguistic review — a native speaker

    Does it read as though it was written in this language? Not: is it accurate. Accurate and obviously translated is still a failure.

  2. SEO review — whoever owns the channel

    Are the fixed terms present where they were specified? Did internal links survive? Are title and meta within budget? This is a checklist, and it can be automated most of the way.

  3. Subject review — where the category requires it

    In regulated categories, local claims need local sign-off, because what is permitted differs by market. See healthcare SEO across countries for what that does to a schedule.

Scaling it

Writing this brief by hand takes twenty minutes a page, which is fine for ten pages and impossible for four hundred. The inputs, though, are almost entirely mechanical: keyword and variants from your keyword data, competitor URLs from the SERP, internal link targets from your own site, heading structure from the source page.

That is a workflow, not a writing task. Assembling the brief automatically and leaving a human to review it rather than compile it is one of the five automations I build most often, precisely because the manual version is what stops multi-market content programmes from scaling past the first language.

The short version

If you take one thing: give the translator the keyword, not the English page alone. Almost every other item on this list is a refinement of that single instruction.

Questions

Because translation optimises for meaning and SEO localisation optimises for how people search, and those produce different words. A translator given an English page renders the English keyword faithfully into a phrase that is correct and that nobody types. The page reads well and ranks for nothing — hard to diagnose, because there is no error anywhere.
The target keyword in the destination language with variants, the intent behind it, two or three competitor pages that currently rank, the required heading structure, terms that must not be translated literally, the register, and an explicit list of what must survive unchanged: internal links, schema, CTAs and structured data values.
No, unless they are also SEOs with access to search data, which is rare. Keyword research is a separate job done against local search volume and the local competitive set. Asking a translator to do it is asking them to guess — and the guess will be based on what sounds natural, not on what people search.
Generate the brief rather than writing each by hand. The inputs are mechanical — keyword and variants, intent, competitor URLs, internal link targets, heading structure — so a workflow can assemble them from your keyword data and CMS, leaving a human to review rather than compile.