International SEO and localization
New markets without losing traffic.
Entering a new country is not translating your website. It is deciding which version each person sees, what changes in each market, and how all of it moves without breaking what already works. We work both halves of the problem: the architecture that tells Google and the AI models which page belongs to which language, and the content that actually answers the way people search in that country.
Translation is not localization
The word people use to find you in Mexico is not the translation of the word they use in Spain. Sometimes it is a different term, sometimes the same one with a different intent behind it, and sometimes the search that drives your business in one country does not exist in the other.
That is why the work does not start with translation. It starts with keyword research by country and language. That tells us which pages each market needs, which ones stay out, and what has to be rewritten rather than translated. Translating is cheap and returns little. Localizing costs more and is what ranks.
The same goes for everything around the text: prices, payment methods, delivery times, legal requirements, product names. A page can say the right thing and still lose the reader in the second paragraph if the currency is wrong.
How we build an international SEO architecture
This is where international SEO is won or lost, and where we find most sites quietly broken.
Structure first. Country domain, subdomain or subfolder. There is no single right answer: it depends on how many markets you are opening, whether your brand is the same in all of them, and who will maintain it. We decide it with you, with a reason behind it, not out of habit.
Then hreflang. It is the tag that tells the search engine “this page is the French version of that one”. It breaks remarkably easily: references that are not returned, languages declared wrong, versions pointing at pages that no longer exist, no default version declared. We implement it in full, with its default, and we check it every time the site changes.
And the rest of the signals have to agree with it: canonical tags, sitemaps per language, internal linking within each market, and the language switcher. When one contradicts the others, the search engine decides on its own, and it rarely decides in your favour.
Migrations without losing traffic
Opening markets usually comes with a move attached: a new domain, a new URL structure, a new platform, or all three at once. It is the riskiest moment in a website's life, and the one where method shows most.
We have done this at scale: at TECH Universidad we moved over a million URLs to a new architecture across more than eleven languages.
- 01
Inventory. Every current URL with its traffic, its rankings and its links. What is not in the inventory is what gets lost.
- 02
Redirect map. One to one, old URL to new URL. No bulk redirects to the homepage and no three-hop chains.
- 03
Cutover. Run against a checklist, with everyone watching the same screen.
- 04
Watch. The following weeks are part of the migration, not the aftermath: crawling, indexing, rankings and errors, reviewed daily through the critical period.
What changes with AI when you sell in several countries
There is a newer part almost nobody is watching yet: AI models do not answer the same way depending on the language you ask in. The same question, asked in English and in Spanish, can return two different lists of brands, citing different sources.
So AI visibility has to be measured market by market too. We do exactly that: we ask the models in each language that matters to you, check whether you appear, which sources they cite you from, and who you are up against in that answer. From there we decide where the gap is and what content opens it.
This builds on the rest, it does not replace it. Around 60% of the sources AI cites are not in Google's top 10, so competing there is a discipline of its own. But a site with broken architecture does not get fixed by showing up in ChatGPT.
What we actually do for you
- Keyword research by country and language, with real volume per market.
- Architecture decision: domains, subdomains or subfolders, and what each one costs you.
- hreflang, canonicals and per-language sitemaps, implemented and audited.
- Content localization: what gets rewritten, what gets translated, what stays put.
- Migrations with inventory, redirect map and post-launch monitoring.
- AI visibility measured per market and language.
FAQ
- What is hreflang and why does it break so often?
- It is the tag that links the versions of the same page across languages. It breaks because it has to be reciprocal — if A points to B, B must point back to A — and because any change to the site can leave a loose end. So it is not something you set once: it is something you check.
- Country domain, subdomain or subfolder?
- It depends on how many markets you are opening and who maintains them. A subfolder concentrates authority and is the cheapest to run; a country domain gives a strong local signal but makes you build authority from zero in every market. We look at your numbers before recommending anything.
- How do you migrate a large site?
- With an inventory of every URL, a one-to-one redirect map, a checklist on cutover day and daily monitoring in the weeks that follow. It is a project with phases and owners, not an afternoon's work, which is why it gets planned before anything is touched.
- Is translating the content enough?
- To be understood, yes. To rank, no. Search changes from one country to the next, and so do the pages you need. Translation is the last step of localization, not the first.