International and multilingual SEO

Choosing a URL structure you will not regret

There are three ways to organise an international site and the choice is hard to reverse, so it is worth getting right at the start. Country-code domains like example.co.uk send the strongest geographic signal and are unambiguous to users, but each domain builds authority separately, so you are effectively running several sites and earning links for all of them.

Subdirectories like example.com/en-gb/ are the pragmatic default for most businesses. Everything sits on one domain, so authority earned anywhere helps everywhere, and you can add a market by adding a folder. Google Search Console lets you set geographic targeting per directory, which covers most of what a country domain would have given you.

Subdomains like uk.example.com sit between the two and rarely win on either count. They separate authority the way a domain does without giving the clear country signal a country domain gives. We generally recommend subdirectories unless there is a legal, hosting or brand reason that forces something else.

• Country domains: strongest signal, highest cost, authority split per market
• Subdirectories: one authority pool, easy to extend, our usual recommendation
• Subdomains: most of the cost of separation with less of the benefit
• Language folders and country folders are different decisions, so plan for both

How hreflang works and where it usually breaks

Hreflang tells a search engine that two pages are the same content for different audiences, so it can show the right version rather than treating them as duplicates. It takes a language code, optionally followed by a region code, and it must be applied to every page in the set, not just the homepage.

The rule that catches almost everyone is that hreflang must be reciprocal. If your UK page points at your Australian page, the Australian page must point back at the UK page. If it does not, engines ignore the whole annotation. On a large site with automated generation, one template that forgets to include the full set breaks thousands of pages at once.

Two other errors are near universal. Using a country code where a language code belongs, so writing "uk" instead of "en-GB", which is invalid and silently ignored. And pointing hreflang at URLs that redirect, are blocked by robots, or carry a canonical tag pointing somewhere else. Any of those makes the annotation void, because the target is not a page the engine can serve.

• Every page in a set must list every version, including itself
• Return links must be reciprocal or the whole set is discarded
• Language codes follow ISO 639-1, region codes follow ISO 3166-1 Alpha 2
• Targets must be indexable, canonical and not redirected
• An x-default entry catches everyone your set does not cover

Same language, different markets: the UK, US and Australia case

The hardest international setup is often not translation at all. It is three versions of English competing with each other. The pages look almost identical to a search engine, and without correct annotation it will pick one and show it everywhere, which means UK visitors landing on dollar pricing and Australian visitors reading about services you do not offer there.

Correct hreflang solves the selection problem. Localisation solves the conversion problem, and it goes beyond spelling. Currency and tax display, phone number format, date format, address format, the terminology the market actually uses, and the legal or regulatory context all change. A page that says "optimization" to a UK reader and quotes prices in dollars reads as foreign even when the offer is right.

Search behaviour differs too. The same service is searched under different names in different markets, so keyword research has to be repeated per country rather than translated from one list.

Translation is not localisation

Machine translation has become good enough to produce readable text and not good enough to produce text that ranks or converts. It misses the terms people search, because search terms are cultural rather than literal. It produces phrasing that native speakers recognise as translated within a sentence or two, which costs trust on a page whose job is to earn an enquiry.

The approach that works is translation followed by local review, with keyword research done natively in the target language first so the copy is built around terms people actually use. Titles, meta descriptions, headings, alt text, schema and form labels all need the same treatment. Half-localised pages, where the body is translated but the navigation and calls to action are not, are extremely common and undermine the whole exercise.

Google has been explicit that machine-translated content published without review is treated as low quality. That is not a reason to avoid translation tools, but it is a reason not to publish their raw output.

Why automatic redirection by location backfires

A lot of sites detect a visitor location and redirect them to the matching version automatically. It feels helpful and it causes two problems. Search engine crawlers usually crawl from one country, so a forced redirect means the crawler only ever sees one version and the rest never get indexed. You end up with a carefully built market structure that no engine can reach.

The second problem is human. Location detection is often wrong, and people travel, use company networks or use privacy tools. Somebody who deliberately wants the UK site and cannot reach it because they are working abroad will simply leave.

The better pattern is to suggest rather than force. Serve the requested URL, then offer a clear, dismissible prompt to switch. Add a visible country and language selector in a consistent place, and remember the choice.

• Never redirect based on IP address alone
• Offer a suggestion banner that the visitor can dismiss
• Keep a persistent, crawlable language and country selector using real links
• Set geographic targeting per directory in Search Console where relevant

Running and measuring a multi-market site

International reporting has to be split by market or it tells you nothing. Aggregate organic traffic can rise while your second-largest market collapses. We report per country and per language: impressions, clicks, rankings and enquiries, so a problem in one market is visible immediately.

The specific check that catches most hreflang failures is comparing which URL ranks in which country search. If your Australian page is the one appearing in UK results, the annotation is not being honoured and it is usually a reciprocity or canonical conflict. That check is quick and worth repeating after every significant release.

Expansion is easier when the first market is set up properly. Adding a fourth country to a clean subdirectory structure with template-generated hreflang is a content job. Adding it to a structure that was improvised is usually a migration.