Should we use separate country domains, subdomains or subfolders?

Subfolders are usually the right answer, because everything a site has earned stays consolidated on one domain. Separate country domains are the strongest signal of local presence and the most expensive choice, since each one starts from nothing and has to be built up independently. The recommendation is made for your specific case with the trade-offs stated plainly.

Why is the wrong country's page showing to our customers?

Usually because hreflang is incomplete or one-directional. If your English page points to the German version but the German page does not point back, the whole set is disregarded and the search engine falls back to picking whichever version it judges most relevant, which is often not the one you want.

Can we just translate our existing pages?

Translation is the smallest part of it. Currency, sizes, phone formats, shipping information, legal pages and the actual words buyers use for your product all differ by market. Two versions that are word-for-word equivalents also risk competing with each other rather than each serving its own audience.

Is machine translation good enough to launch with?

As a first draft, yes. As published copy, no. Machine-translated pages consistently fail to rank and they damage how the brand reads to a native speaker, which matters more in some markets than others. Native review belongs in the budget from the start, and we say so before the work is quoted rather than after.

How do we know it has been implemented correctly?

Every version is re-crawled after deployment specifically to check the hreflang relationships, since crawlers report reciprocity errors as a dedicated set of findings. Search Console's international reporting is checked alongside it, and each market's performance is then tracked separately rather than blended into one number.

Why does this cost more than domestic SEO?

Because the work multiplies by market and the errors are quiet. A broken structure across five countries does not announce itself — the pages look fine and the wrong ones simply get served. Detecting and correcting that requires checking each version against every other, which is where the time goes.

SEO & Organic Growth

International and multilingual SEO

One site serving several countries and languages without the versions fighting each other — URL structure, reciprocal hreflang, and localisation that goes past translation.

From: $700–$4,000/mo
Typical time: ongoing
Usually bought by: Exporters

The wrong country's page keeps showing

The businesses that need this are already selling across borders, and the symptom is oddly specific every time. A buyer in one country lands on the page written for another, with the wrong currency and shipping terms in front of them. Or a language version that was expensive to produce simply does not rank at all. Or two country versions of the same product are both sitting far lower than the original single page ever did.

What makes this different from ordinary search work is that the failure is silent. There is no error message, nothing looks broken, and the pages load perfectly. The site appears healthy while quietly serving the wrong version to a share of its most valuable visitors. Owners often only discover it when a customer mentions it, by which point it has been happening for months.

Two decisions do most of the damage or most of the good. The first is structural — how the market versions are arranged in the URL, which is difficult and costly to change later. The second is whether the versions genuinely differ or are translations of one another. Getting both right is unremarkable when it works, which is precisely why it gets underestimated.

How the work runs

Decide the structure, and write down why. Subfolder, subdomain or separate country domain, chosen for your actual situation — how many markets, whether local entities exist, how much authority the current domain holds and who will maintain it. The recommendation comes with the trade-offs stated openly, because this is the one decision that is genuinely painful to reverse. Two to three days.
Map every page to every version. Which pages exist in which market, which are missing, and which are duplicates that should not have been created. This inventory is what the hreflang implementation is generated from, and skipping it is how sites end up with tags pointing at pages that do not exist. Two to five days depending on site size.
Implement hreflang properly, including the parts people leave out. Every version referencing every other version, each page referencing itself, and a default fallback for visitors who match no specific market. The references must be reciprocal — a one-way relationship causes the entire set to be disregarded rather than partially applied. Three to five days.
Localise beyond translation. Currency and price formats, units and sizing, phone number formats, shipping and returns information, market-specific legal pages, and the product terminology each market actually uses — which can differ between two countries that share a language. Translated copy goes to a native speaker for review before it is published, never straight from a machine to live. Ongoing, per market.
Validate, then track each market separately. A full re-crawl focused on hreflang relationships to catch missing return references and mismatched codes, Search Console's international reporting checked per version, and performance reported market by market rather than as one blended figure that hides which country is failing. Two days, then monthly.

What moves the price

The range is $700–$4,000/mo and the work is ongoing, because every new market and every catalogue change reopens the validation. Three things set the level.

How many market and language combinations there are. The work does not grow linearly. Every additional version has to reference every existing one, so validation effort grows far faster than the number of markets, and so does the number of ways it can break.

How much genuine localisation is required. Two English-speaking markets sharing a catalogue is a light lift. A market needing a different language, different legal pages, different units, different shipping terms and native review of every page is a substantially larger one.

Whether a structure has to be migrated. Setting this up cleanly on a new build is one job. Moving a site that is already split across country domains onto a consolidated structure, without losing what each version has earned, is a migration with its own redirect planning and its own risk.

What usually goes wrong

Hreflang is implemented in one direction only. The English page lists the other language versions and those pages never point back. The relationship is not treated as partially valid — the set is disregarded, and everything that was supposed to prevent the wrong version being served stops working. This error is common enough that crawlers ship a dedicated report for finding it, which tells you how often it happens.

Machine translation is treated as finished copy. It reads acceptably to someone who does not speak the language, which is exactly the problem. Those pages consistently fail to earn position, and a native reader recognises them immediately, which costs credibility in a market you are trying to enter. Native review is a line in the budget, agreed up front.

The site is split into separate country domains too early. It feels like the serious, committed choice. Each new domain begins with nothing behind it and has to be built up independently, so a business with the resources for one strong site ends up with four weak ones. Consolidating them again afterwards is a migration nobody wants to pay for twice.

See where the site stands before you brief anyone.

If the site is already live, start from what it scores rather than from a guess. The audit reports what it checked, what it found, and — the part most tools leave out — what it could not check, and why.

See the website audit Talk about this work

Questions people ask before they commit.

Should we use separate country domains, subdomains or subfolders?: Subfolders are usually the right answer, because everything a site has earned stays consolidated on one domain. Separate country domains are the strongest signal of local presence and the most expensive choice, since each one starts from nothing and has to be built up independently. The recommendation is made for your specific case with the trade-offs stated plainly.
Why is the wrong country's page showing to our customers?: Usually because hreflang is incomplete or one-directional. If your English page points to the German version but the German page does not point back, the whole set is disregarded and the search engine falls back to picking whichever version it judges most relevant, which is often not the one you want.
Can we just translate our existing pages?: Translation is the smallest part of it. Currency, sizes, phone formats, shipping information, legal pages and the actual words buyers use for your product all differ by market. Two versions that are word-for-word equivalents also risk competing with each other rather than each serving its own audience.
Is machine translation good enough to launch with?: As a first draft, yes. As published copy, no. Machine-translated pages consistently fail to rank and they damage how the brand reads to a native speaker, which matters more in some markets than others. Native review belongs in the budget from the start, and we say so before the work is quoted rather than after.
How do we know it has been implemented correctly?: Every version is re-crawled after deployment specifically to check the hreflang relationships, since crawlers report reciprocity errors as a dedicated set of findings. Search Console's international reporting is checked alongside it, and each market's performance is then tracked separately rather than blended into one number.
Why does this cost more than domestic SEO?: Because the work multiplies by market and the errors are quiet. A broken structure across five countries does not announce itself — the pages look fine and the wrong ones simply get served. Detecting and correcting that requires checking each version against every other, which is where the time goes.

Often taken together.

Start with the brief, not the invoice.

Send what you are trying to achieve. You get a written scope, a price and a date — or an honest “this is not the right service for that”.

Talk about International / multilingual SEO All SEO & Organic Growth services