Multilingual SEO

Multilingual SEO services because you cannot translate a keyword list

Translation carries your meaning across and leaves the search demand behind. Germans looking for a mobile phone do not search Mobiltelefon, they search Handy. A perfectly accurate translation of your best page can target a term nobody types.

Keyword research done natively in each language, by people who search in it, before a single page is translated.

The basics

What are multilingual SEO services?

Multilingual SEO services make a site rank in more than one language: researching how people actually search in each language rather than translating your existing terms, producing content that reads as though it was written in that language, and annotating the versions so search engines serve the right one without treating the others as duplicates.

Multilingual SEO is about what language. Serving Canada in English and French is multilingual work inside one market. Serving the United States and Australia in the same language is international SEO, which is about where. Businesses that need both usually do one first, and doing them at once is how projects stall.

The core problem

Translation preserves meaning and destroys search demand

This is the failure that quietly wastes most multilingual budgets, and it is invisible to everyone involved. The translation is accurate. A native speaker reviews it and confirms it reads correctly. The page goes live and ranks for nothing, because the term it targets is not the term people use when they are looking to buy.

German speakers searching for a mobile phone overwhelmingly use Handy rather than the literally correct Mobiltelefon. French speakers looking for cheap flights search vol pas cher rather than the grammatically faithful vols bon marche. Spanish speakers renting a car search alquiler rather than contratacion. In each case the translation is not wrong. It is simply not what anybody types.

The rule this produces: keyword research must be done natively in each target language, from scratch, before translation begins. Not translated afterwards, and not run through a tool on your English list. The research determines what the page is about, which means it has to come first or the page is built around the wrong idea.

Occasionally the research finds that the concept itself differs. Categories that make sense in one market are split differently in another, and a page structure that works in English can need reorganising rather than rewording. That is the point at which a translation project becomes an SEO project, and it is better discovered at the research stage than after forty pages exist.

Quality levels

Machine translation, edited translation, and writing

There are three ways to produce content in another language and they cost and perform very differently. Choosing deliberately is more important than choosing the expensive one.

Raw machine translation

Cheap, instant, and risky at scale. Published unreviewed across hundreds of pages it produces thin content that can be treated as spam, and it will faithfully translate keywords into terms nobody searches.

Machine plus human editing

The sensible default for most content. A native editor corrects register, fixes terminology and swaps in the terms people actually use. Cost-effective and good enough for the large body of a site.

Written in the language

For your commercial pages, where persuasion matters and the argument may need to differ. Objections are not universal, and a page written natively can address the ones that exist in that market.

Search engines do not object to machine translation as such. What causes problems is publishing large volumes of unreviewed output, which tends to be low quality and adds little. The practical approach is to write your money pages natively, edit machine output for everything else, and never publish a page nobody who speaks the language has read.

Process

How our multilingual SEO services run

Research natively first

Keyword research carried out in each target language by someone who searches in it, producing the terms buyers actually use rather than the terms your English list translates into.

Map before translating

Which pages are needed in which languages, where the structure differs, and which pages are not worth producing at all. Not every page deserves a version in every language.

Produce at the right level

Money pages written natively, supporting content machine-translated and edited by a native speaker, with terminology held consistent through a glossary.

Annotate and validate

Hreflang between language versions, built symmetrically and checked, so search engines serve the right version and treat the others as alternates rather than duplicates.

The order is the important part. Research before mapping, mapping before translation, translation before annotation. Reversing any two of those produces pages that are correct and pointless.

Deliverables

What the language work includes

Native keyword research

Done separately per language, including the terms that differ conceptually rather than only linguistically, with search demand attached to each.

Terminology glossary

Your product and category names fixed per language so twelve pages do not use nine different words for the same thing, which is the most common cause of diluted rankings.

Content mapping

A page-by-page plan of what gets a version in which language, at which quality level, with the pages not worth producing named explicitly.

Hreflang for languages

Language and where necessary language-region annotation, implemented symmetrically with return tags and an x-default, then validated rather than assumed.

Translated and native copy

Commercial pages written in the language, supporting content edited to read naturally, with titles and descriptions written per language rather than translated.

Language switching

A visible switcher that is crawlable, does not force a choice by IP, and keeps the visitor on the same page rather than dropping them on a translated homepage.

Duplicate resolution

Shared-language variants and partially translated pages handled through correct annotation and indexing rules rather than left to be guessed at.

Per-language reporting

Traffic and conversion split by language, because an underperforming language is invisible inside a combined total.

A maintenance plan

How updates propagate. Every language version you add is a version that goes stale, and unmaintained translations are worse than not having them.

Honest answer

When adding a language is a mistake

Nobody can answer in that language

Ranking in a language your team cannot reply in produces enquiries you have to turn away. Sales and support capability should exist before the content does.

You cannot maintain it

Every language multiplies ongoing work. Prices, terms and products change, and translated pages that quietly go stale damage trust more than a single well-maintained language would.

Your audience already searches in English

In several industries and several countries, professional buyers search in English regardless of their first language. Worth checking with the data before assuming a translation is needed.

The English version is not working yet

Translating a page that does not convert produces the same page in more languages. Fix the argument first, then multiply it.

A quick sanity check on any existing translated page: run it through our readability checker. Machine output frequently produces longer, more complex sentences than the original, which is one of the more reliable signs a page was translated rather than written.

Investment

How much do multilingual SEO services cost?

Priced per language and by how much of the site each language needs, because the research and the glossary are fixed costs per language while translation scales with page count.

Research and setup, per language

$1,800 to $4,000. Native keyword research, terminology glossary, content mapping and hreflang implementation. Done once per language before any pages are produced.

Content production

$120 to $400 per page. The range reflects the difference between edited machine output for supporting pages and natively written commercial pages.

Ongoing per language

$800 to $2,500 a month. New content, updates propagated, and monitoring. Languages go stale faster than anyone expects, and stale is worse than absent.

The most common budgeting error is funding the translation and not the research, which produces accurate pages targeting terms nobody searches. If the budget only stretches to one of the two, the research is the half that determines whether any of it works.

Common questions about multilingual SEO

Multilingual is about language, international is about country. Serving Canada in English and French is multilingual work within one market. Serving the United States and Australia in English is international work in one language. Many businesses eventually need both, and they are usually best done one at a time because each has its own structure and its own research.

No, and this is the mistake that wastes the most money. Translation gives you the accurate word, which is frequently not the searched word. German speakers search Handy rather than Mobiltelefon for a mobile phone. Research has to be done natively in each language, from scratch, before any pages are written, because it determines what the page should be about.

Not inherently. What causes problems is publishing large volumes of unreviewed output, which tends to be low quality and adds little value, and can be treated accordingly. The workable approach is native writing for commercial pages, machine translation with native editing for the rest, and nothing published that a speaker of the language has not read.

Content in genuinely different languages is not duplicate content. The risk sits in shared-language variants, partially translated pages that fall back to English, and pages available under several URLs. Correct hreflang annotation between language versions, plus sensible canonical and indexing rules, resolves it.

It can suggest, but it should not force. Automatic redirection based on browser language or IP prevents crawlers from reaching your other versions and traps anyone whose location does not match their language. Offer a visible, crawlable switcher, suggest a version with a dismissible prompt, and keep the visitor on the same page when they switch rather than sending them to the homepage.

Rarely, and trying to is how these projects stall. Start with the pages that earn: your main commercial pages, the top few informational pages, and anything legally required. A well-maintained twenty-page version in a language outperforms a two-hundred-page version nobody updates, and it costs a fraction to keep current.

By deciding the process before you add the language. Someone has to own it, changes to the source page need to trigger a review of every version, and a glossary keeps terminology consistent as different people work on it. Without that, versions drift, prices go stale, and you end up with pages that actively mislead the market you were trying to reach.

Find out whether your translations target anything

Send us a translated page and we will check what it currently targets against what people in that language actually search for, and tell you whether the gap is a wording fix or a research problem. It is usually a research problem. No obligation, no sales pitch.

Get my free language review
  • Native term check
  • Hreflang validated
  • Honest scope