SEO for International Sites: URL Structure, Hreflang and Local Signals

How to decide between country domains, subdomains and subfolders, write hreflang that validates, and give each market the signals it needs.

Published by AI Knows Us · Updated 12 October 2026

SEO for an international site comes down to three decisions. Choose one URL structure for all markets, country domains, subdomains or subfolders, and never mix them. Tell search engines which version serves which audience using hreflang annotations that point both ways and include a self reference. Then give each market genuine local signals: content written in the language, prices in the local currency, a local address or phone number where you have one, and internal links between pages in the same market. Everything else in international SEO is detail hanging off those three.

The most common failure is not technical. It is launching twelve translated copies of the same pages, each one competing for the same handful of head terms, with nobody in any of those markets searching in the words the translation used. The structure below is ordered the way the work should actually be done.

Step one: decide what you are targeting

Language and country are different targets and they need different setups.

  • Language targeting suits one Spanish site serving readers in Spain, Mexico and Argentina. Use es with no region code.
  • Country targeting suits different prices, stock, shipping, legal terms or support hours per country. Use the language plus region form, such as en-gb and en-au.
  • Both applies where a single country has two served languages, for example fr-ca alongside en-ca.

Write this down before you choose a URL structure, because the answer decides the structure.

Step two: pick one URL structure

StructureExampleStrongest whenCost
Country domain (ccTLD)example.deSeparate legal entity, strong local brand, one clear country per siteEach domain earns authority on its own, and you manage many domains
Subfolderexample.com/de/Most businesses, especially those with one main brandLowest, everything sits on one domain's authority
Subdomainde.example.comSeparate hosting or separate platform per marketAuthority is shared less readily than with subfolders
Parametersexample.com?loc=deNever for public market pagesHard to target, hard to link, easy to get wrong

Subfolders are the default recommendation for a reason: one domain accumulating references is almost always easier to rank than six domains each starting from nothing. Choose country domains when the business is genuinely run as separate country operations, or where buyers in the market expect a local domain.

Step three: hreflang that validates

Hreflang tells search engines that several URLs are the same page for different audiences. Three rules make it work:

  • Every page in the set lists every other page in the set, including itself.
  • The annotations are reciprocal. If the German page points to the French page, the French page points back.
  • Use full absolute URLs, and use the canonical version of each URL.
<link rel="alternate" hreflang="en" href="https://example.com/page/" />
<link rel="alternate" hreflang="de" href="https://example.com/de/seite/" />
<link rel="alternate" hreflang="fr-ca" href="https://example.com/fr-ca/page/" />
<link rel="alternate" hreflang="x-default" href="https://example.com/page/" />

Use x-default for the page that serves everybody you have not built a version for. Language codes follow ISO 639-1 and region codes follow ISO 3166-1 Alpha 2, so the United Kingdom is gb, not uk. That single mistake is the most frequent error in hreflang implementations. You can place the annotations in the HTML head, in HTTP headers for files such as PDFs, or in the XML sitemap, which is the easiest option when a developer deployment is slow.

Step four: let people land where they searched

Serve the version that matches the query and offer a change, rather than forcing a redirect based on IP address. An automatic redirect sends a buyer who deliberately searched for your United States pricing to a local site they did not ask for, and it stops search engine crawlers from reaching most of your site, because they largely crawl from one region. A small banner saying "You are on the United Kingdom site, switch to Germany" does the job and keeps every version reachable.

Step five: research each market in its own language

Translating your English keyword list is the step that quietly wastes the most budget. People search with different words, different word order and different intent. Do the research natively in each market, using local search volume data, local autocomplete and competitors who already rank there. Expect the page set for one market to be a different shape from the original, not a mirror of it.

Machine translation is a fine starting draft and a poor finished page. A native editor reviewing and reworking the output costs far less than a full retranslation and is the difference between a page that ranks and a page that reads as foreign. Translate the parts buyers judge you on first: pricing, product detail, support and legal.

Step six: the local signals that still count

  • Local currency and local payment methods on commercial pages.
  • A local address and phone number where one exists, matched to your business profile for that country.
  • Internal links kept inside the market, so the German navigation links to German pages.
  • References and press from sites in that country, which is the strongest market level signal you can earn.
  • Server performance in the region, which content delivery networks handle well.

Set the geographic target in Google Search Console for generic domains where the whole site serves one country, and verify each variant as its own property so you can see performance market by market.

This page covers the architecture and annotations for a multi market site. For translation workflow and tooling choices, work from a separate localisation plan and bring the URL decisions here.

Questions buyers ask about SEO for International Sites

Do I need hreflang if my pages are in different languages?

Yes. Different languages are exactly the case hreflang was built for. It stops the wrong language version being shown to a user and keeps near identical pages from competing with each other.

Is translated content treated as duplicate content?

Translations of the same page into different languages are handled as alternates, not duplicates, when hreflang is in place. The real duplication risk is several country versions in the same language, for example United States and Australian English, which is why reciprocal annotations and self references matter most there.

Should I use a country domain or a subfolder?

Subfolders for most businesses, because all markets build on one domain's authority. Country domains when you operate as separate country businesses, or when local buyers expect a local domain and you have the resource to build authority for each one.

Can I target a region such as Europe or Latin America?

Hreflang works at language and country level, so target the languages and the specific countries that matter instead. One Spanish page with es plus country versions for the two or three markets with different pricing covers Latin America cleanly.

What should I check first when international traffic is flat?

Check indexing per market in Search Console, confirm hreflang is reciprocal and uses valid codes, confirm no IP redirect is blocking crawlers, and confirm the keywords were researched in that language rather than translated. Those four cover most cases.

The market by market answer you are not measuring

There is now a second place where a buyer in Germany or Brazil asks your category question, and it is an assistant rather than a search box. The answer differs by language and by market, because the sources the assistant leans on differ by language and by market. A business can be the obvious recommendation in English and absent from the same question asked in Spanish. That gap shows up in no rank tracker and in no hreflang validator. AI Knows Us runs category questions across ChatGPT, Gemini, Perplexity and Claude and shows which sources produced each answer, which makes the per market picture visible.

See what AI says about you.

The first scan is free, it needs no account, and it asks the questions your buyers ask.

Free. No card. We ask 5 real buyer questions on 2 AI apps.

Already sure you want this? Join the waitlist. We open on 31 October 2026. It costs an email address. No card, and nothing is charged. Or talk to a person if you have a question first. Or read how we fix what it finds.