Multilingual SEO: How to Rank One Site in Several Languages

Multilingual SEO gets sold as a translation project. Buy the words, tick the languages, wait for the traffic to arrive from somewhere new. That is the most expensive misunderstanding in this corner of search, and I have watched it eat serious budgets on sites that finished the year ranking in exactly one language.

I am Armen, an SEO consultant in Barcelona, and I run the agency ACERO Digital with a team of specialists. A lot of the multilingual projects that reach me are already translated. The pages exist, they read perfectly well, and Google is either serving the wrong version to the wrong reader or quietly ignoring most of them.

Here in Barcelona this is not an edge case. Plenty of businesses I sit with run Catalan, Spanish and English at once because their customers genuinely arrive in all three. So this is the version I would give you across a table, including the one thing almost nobody checks any more, because Google stopped reporting it.

What is multilingual SEO?

Multilingual SEO is the practice of making one website rank in more than one language. It runs from the words each audience actually types through to the technical signals that tell a search engine which version of a page belongs to which reader.

The part people skip is the first half. Translation moves your existing pages into another language. Multilingual SEO asks whether those pages were ever the right ones for that audience, then builds the ones that are.

Is multilingual SEO the same as international SEO?

No. Multilingual SEO targets languages. International SEO targets countries. The distinction sounds pedantic until you are three months into a build and cannot work out why your Spanish page ranks in Spain but never in Colombia.

A language problem looks like this: you have one product, one price, one audience, and some of them read in German. A country problem looks different. Different currency, different shipping, different competitors on the first page, sometimes a different product entirely. Deciding which markets you enter and in what order is a country decision, and it comes with its own budget.

Most projects are a mix. The useful move is to name which one you are solving first, because it changes the architecture you build. If you are unsure which side of that line your business sits on, our international SEO service is built around answering that before anyone writes a line of code.

Do you have to translate your whole website?

Almost never, and translating everything is usually the slowest way to get your first result. A new language version has no links, no history and no trust with Google, so a hundred thin translated pages compete against local sites that have been earning authority for years.

Ship a narrow set of pages instead. The homepage, the one or two service or product pages that actually generate enquiries, and a small number of pages built around what that audience searches. Fifteen strong pages in German will outperform four hundred machine translated ones, and you will know within a few months whether the market is worth deepening.

There is a second reason to go narrow. Every language version you add is something you now have to maintain forever. Prices change, offers change, and the version nobody remembers to update becomes the one Google shows to your best prospect.

How do you do keyword research in another language?

You research in the target language from scratch. You do not translate your English keyword list, because a translated keyword is a guess about what the market calls the thing.

Set your keyword tool to the target country rather than the target language, then pull the terms local competitors rank for and check the search volume there. The words shift more than people expect. Spanish speakers in Spain search for an ordenador while Mexicans search for a computadora, and a page optimised for the wrong one is invisible to half the Spanish speaking web.

Then read the results page for your chosen term in that language. The intent moves too. A query that returns product pages in English can return comparison articles in French, which means the page you translated was the wrong format before a single word was changed.

If your budget only stretches to one native speaker, spend it here rather than on the body copy. Getting the title, the main heading and the opening paragraph anchored to the real local term matters more than polishing paragraph nine.

Which URL structure should you use for each language?

Pick one of the three that Google supports and stay with it. Google’s own documentation on multi regional sites lists country specific domains, subdomains and subdirectories as workable, and marks URL parameters such as site.com?loc=de as not recommended.

For a purely multilingual site the subdirectory wins nearly every time. This site runs that way, with English at the root and Spanish under /es, and there is no version of our business where splitting that into two domains would have been worth the authority we would have thrown away.

One rule sits underneath all of them. Google states plainly that you should not use IP analysis to adapt your content, because its crawlers may then never see the other versions at all. Offer readers a language switcher and let them choose. Do not force a redirect on them based on where their connection appears to be.

How does hreflang work, and where does it break?

Hreflang is the annotation that tells Google which language and region each version of a page is for, so the right one gets served instead of the one with the most links. It is a signal rather than a command, and it works when the whole set of pages agrees with itself.

That agreement is where it falls apart. Four failures cause most of the damage I see in audits.

Use two letter language codes on their own, such as es or de, when you are targeting a language regardless of country. Add a country only when you mean it, as in es-MX. You cannot specify a country by itself, which is the invalid code that trips up most first attempts.

Now the part almost nobody knows. Google removed the International Targeting report from Search Console in September 2022 and never replaced it. Hreflang errors are reported nowhere in Search Console any more, which means a broken implementation makes no noise at all. Nothing turns red. Traffic in one language just stops growing, and six months later somebody notices.

So validation has to be something you schedule. A crawler such as Screaming Frog, Sitebulb or Ahrefs Site Audit will show you missing return tags and broken URLs inside the set, and it takes minutes once it is configured. It is one of the first things our team runs during an SEO audit on any site with more than one language, precisely because there is no warning light for it.

What about one language across several countries?

Split it only when the countries genuinely differ. If your prices, delivery or legal terms change between Spain and Argentina, separate pages with es-ES and es-AR earn their place. If the content is identical, splitting gives you two weaker pages chasing the same searches.

Where you do split, the pages must actually differ. Duplicating a Spanish page across five country codes and changing nothing tells Google you have five near identical pages, and it will pick one and largely ignore the rest. Change the price, the shipping information and the examples, or leave it as one page.

There is a middle path most people miss. Keep one Spanish page, target it as plain es, and add a country version later when you have evidence from Search Console that a specific market converts differently. Let the data buy the extra page.

Does multilingual SEO matter for AI answers?

More than it did, and this is the piece the competing guides have not caught up with. Assistants answer in whatever language the question was asked in, and they pull from sources in that language. A business with one English page is simply not in the pool when somebody asks ChatGPT the same question in Spanish.

The practical effect is that a well built second language version now buys you two things at once. You show up in the traditional results for that language, and you become quotable in AI answers where almost none of your local competitors have bothered to appear. That is a wide open door in Barcelona right now, and it will not stay open.

What I would do first

If you already have translated pages sitting on your site, do not write anything new this week. Crawl what you have and check the hreflang set, because a broken implementation makes every other improvement invisible. It is the cheapest win available and nothing in Search Console is going to tell you about it.

If you have not started yet, resist the urge to translate everything. Pick one language, research its keywords natively, and build ten or fifteen pages you would be happy to defend on their own merits. Prove the market pays you back before you scale the effort.

That is genuinely most of it. The technical layer is a solved problem with a known set of failure modes, and the real difficulty is having the discipline to build fewer, better pages in each language rather than a translated copy of everything you already own. If you want a second pair of eyes on a site that is already running in more than one language, tell me what you are working with and I will tell you honestly whether it needs a rebuild or a fix.