International SEO
Get the domain structure right and international SEO becomes ordinary SEO
Almost every international search problem traces back to one early decision: country domains, subfolders or subdomains. Get that right, add hreflang that validates, localise rather than translate, and the rest is the same work you already do — done once per market.
The decision everything hangs on
Country domains, subfolders or subdomains
This choice is expensive to reverse and is usually made by whoever registered a domain first. It deserves a proper decision.
| Dimension | ccTLD (example.de) | Subfolder (example.com/de/) | Subdomain (de.example.com) |
|---|---|---|---|
| Geotargeting signal | Strongest and automatic. The domain itself says which country it serves. | Clear once configured, and reinforced by content, currency, language and linking. | Workable but the weakest of the three, and often ambiguous to both engines and users. |
| Authority | Each domain starts from zero and must earn its own links and history. | Every market shares the authority of one domain. New markets launch from a strong base. | Treated as related to the parent but not identical. Consolidation is less certain. |
| Local trust with buyers | Highest. In several markets buyers actively prefer a local domain. | Good, provided the page shows local currency, contact details and terms. | Neutral. Rarely gains or loses trust on its own. |
| Cost and maintenance | Highest. Separate registrations, hosting, certificates, deployments and often separate teams. | Lowest. One codebase, one deployment, one set of templates. | Middling. One brand, but often separate infrastructure in practice. |
| Best when | A market is large, legally separate, or local trust is a genuine buying factor. | You are expanding into several markets and want each to benefit from the others. | A platform constraint forces separation, or a market runs on a different stack. |
| Where it hurts | Ten small markets, each with a weak domain nobody links to. | Regional teams who wanted autonomy and now share a release process. | Ambiguity: neither the authority of a subfolder nor the clarity of a country domain. |
The second decision
Translation is a language job. Localisation is a demand job.
A translated site says the same things in a different language. That is fine for a manual and inadequate for a commercial page, because the words people search are not a translation of the words you use.
The pattern repeats in every market. A category has a common name in one country and a completely different common name in the country next door, both speaking the same language. A benefit that leads in one market is regulated language in another. A payment method that is unremarkable at home is the single biggest trust factor somewhere else. None of this reaches a translator, because a translator was given a source text rather than a market brief.
Localisation starts from local search behaviour instead. We research demand in the target market first, in the target language, and write a brief that says what the page has to cover there. Sometimes that produces a page close to the original. Often it produces a different page with a different emphasis and a different order, which is the point.
The same logic applies to markets that share your language. A page written for a United Kingdom audience that references local regulation, local pricing and local providers is genuinely a different page from its United States equivalent. If you cannot list what would differ, that is a signal you may not need a separate version at all.
How a market launch runs
From market selection to a version that holds its own
Prioritise markets on evidence
Search demand in the local language, competitive density, the cost of serving the market, and whether the product is legally and practically sellable there. Sales ambition is an input, not the decision. Sequencing matters more than most teams expect.
You get: Ranked market list with the reasoning for each position
Decide the structure
Country domain, subfolder or subdomain, decided against the market list rather than one country at a time. We include the migration cost of moving from where you are today, which is often the deciding factor.
You get: Structure recommendation with a migration and risk assessment
Specify hreflang properly
Correct language and region codes, return tags in both directions, self-referencing entries, an x-default for unmatched audiences, and a decision on whether it belongs in the markup, the headers or the sitemaps. Then a validation routine that runs continuously rather than once.
You get: hreflang specification and an ongoing validation report
Localise from local demand
Keyword and question research in the target market, in the target language, feeding briefs that say what the page must cover there. Translation happens inside that brief rather than instead of it.
You get: Per-market content briefs and localised commercial pages
Handle the local engine reality
Where a market is not dominated by Google, the technical requirements and the content expectations differ. We assess whether the share justifies separate work, and say when it does not.
You get: Per-market engine assessment and any additional requirements
Measure each market separately
Separate baselines, separate expectations and separate reporting. A market entered last quarter should not be averaged against a market you have led for a decade, and doing so hides both the wins and the problems.
You get: Market-level reporting with independent baselines
What we find
Four faults behind most international underperformance
- hreflang is implemented and returning errors.
- The most common causes are missing return tags, region codes used where language codes belong, entries pointing at redirecting or non-canonical URLs, and templates that generate the tag set from a stale list. An implementation with errors on a large share of URLs is treated as unreliable, which means the effort produced nothing.
- The English variants are cannibalising each other.
- United Kingdom, United States, Australian and Canadian versions with near-identical content and no meaningful local difference. Search engines pick one, often the wrong one for the market, and the internal linking usually reinforces the mistake. This is an editorial problem wearing a technical costume.
- Visitors are redirected by IP address.
- It seems helpful and it is quietly severe. Crawlers arrive from a limited set of locations, so most of your market versions may never be crawled at all. A suggestion banner achieves the same user outcome without removing the other versions from the index.
- Regional teams each built their own site.
- Five country domains, five agencies, five taxonomies and no links between any of them. Each one is a small weak site rather than part of a strong one. Consolidating is a real project with real politics, and it is usually the highest-value thing available to a global brand in this position.
Scope
What an international engagement covers
Scope depends on whether you are entering your second market or repairing an estate of fifteen.
- Market demand research in local languages, not translated keyword lists
- Domain structure recommendation including the cost of migrating from the current setup
- hreflang specification, implementation support and continuous validation
- Search Console property and targeting configuration per market
- English-variant differentiation or consolidation plan
- Localisation briefs built from local search behaviour and local regulation
- Currency, payment, contact and trust signal requirements per market
- Local link and coverage acquisition in the target market
- Assessment of non-Google engines where market share justifies it
- Per-market reporting with independent baselines and targets
Questions
What teams ask before expanding
Should we use country domains, subfolders or subdomains?
For most businesses, subfolders on one strong domain. Every market benefits from the authority the whole domain has built, there is one site to maintain, and adding a market is a directory rather than a project.
Country domains make sense when local trust genuinely matters to buyers, when a market is large enough to justify running a separate site, or when legal and regulatory separation requires it. The cost is that each domain starts from nothing and has to earn its own authority. Subdomains sit awkwardly between the two and are usually chosen for platform reasons rather than search ones.
What does hreflang actually do?
It tells a search engine that several pages are alternatives of one another for different languages or regions, so the appropriate one can be shown to the appropriate audience. It is a swap instruction, not a ranking signal.
That distinction matters. If a page is not ranking in a market, hreflang will not fix it. If the wrong country version is ranking in a market, hreflang usually will. Getting the return tags right in both directions is the part most implementations fail on.
We have UK, US and Australian sites in English. Is that a problem?
It is the most common international problem we see. Three near-identical English sites with no clear market signals compete with each other, and search engines have to guess which one belongs in which country.
The fix is partly technical — hreflang with correct region codes, an x-default, and correct targeting — and partly editorial. Each version needs genuine local difference: local pricing and currency, local terminology, local contact details, local examples and local regulatory context. If there is nothing genuinely different to say, the honest answer may be that you need one English site rather than three.
Should we redirect visitors based on their location?
No, not automatically. Crawlers usually visit from a small number of locations, so an automatic redirect can mean most of your market versions are never seen or indexed at all. It also frustrates travellers and anyone who deliberately wants a different version.
Suggest instead of redirect: a dismissible banner offering the local version, with the choice remembered. Every version stays crawlable and the visitor keeps control.
Is Google the only engine we need to think about?
In most Western markets, effectively yes, with Bing worth attention where desktop and corporate use is high. In several important markets it is not: local engines hold significant share in China, Russia, South Korea, the Czech Republic and Japan among others, and each has its own indexing behaviour and expectations.
We will tell you when a market genuinely needs separate work and when it does not. Adding a second engine to the plan is a real cost, and it is only worth it when the share justifies it.
Do we need to translate the whole site?
Rarely all of it, and translation alone is usually not enough anyway. The commercial pages, the pages that answer buying questions and anything with legal or regulatory weight need to exist locally. Deep archive content often does not.
What matters more is localisation: using the terms people in that market actually search, prices in their currency, examples they recognise, and regulatory detail that applies to them. A translated page uses your terminology in their language. A localised page uses theirs.
Settle the structure question before anything gets built
If you are about to launch a market, the domain decision is the cheapest thing to get right now and the most expensive to change later. Bring us the plan and we will tell you what each option costs you over the next five markets.
If we don't deliver the work we agreed to deliver for reasons within our control, you don't pay for the undelivered work. Read our guarantee
Related
Where to go next
- the technical groundworkhreflang and geotargeting are implementation problems first.
- programmes across many stakeholdersRegional autonomy is usually the hardest part of a global rollout.
- our United Kingdom market work
- search in the United Arab EmiratesA market where language and audience rarely line up neatly.
- redesign and replatform projectsThe moment structure decisions are cheapest to make.
- catalogue workInternational stores compound both problems at once.
Last updated · Reviewed by Zubair Afzal