First, see the results the way the target market sees them
Searching from your own office proves little, because results depend on location and language settings. Use a browser profile set to the target language, a search with the market's country parameter, or a colleague in that country, and record which URL appears for the main queries. Then check Search Console (GSCGoogle's search performance and indexing tool.): filter the performance report by country and look at which pages earn impressions there. A UK page collecting most of its impressions in the United States confirms the problem is systematic, not a one-off result.
Now inspect both URLs, the one that should rank and the one that does. URL inspection shows whether each is indexed and which canonical Google selected. If the intended page reports the other locale as Google's chosen canonical, Google has merged the two and is serving one for both markets. If both are indexed under their own URLs, the issue is which one Google prefers for the market, and hreflang and supporting signals are where to look. Note the language and country codes each page declares; small coding mistakes are common and decisive.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Both versions indexed, the wrong one shown; no return-link warnings resolved | View hreflang on both pages; check codes and reciprocity | hreflang missing, one-way or with wrong codes | Add complete, reciprocal sets with valid language and region codes |
| The intended page reports the other locale as Google's canonical | URL inspection of both versions; canonical tags in the source | Canonical tags pointing across locales | Make each locale canonical to itself and connect them with hreflang |
| One version is never crawled or indexed at all | Fetch each locale without cookies from different regions; check redirects | Automatic redirects by IP or browser language hiding versions from Google | Replace forced redirects with a suggestion banner and a language switcher |
| Two same-language editions with almost identical copy | Diff the pages; compare prices, currency, contact details | Near-identical versions Google folds together | Localise each edition meaningfully, or merge them into one |
| Links, menus and sitemaps mostly reference one edition | Crawl inlinks by locale; sitemap contents per edition | Internal links and sitemaps favouring one version | Link each edition within itself and list it in its own sitemap |
Fix, verify and prevent each cause
hreflang that is absent, broken or one-sided
Google's documentation on localized versions explains that each version should list itself and every alternate, and that the annotations must be confirmed from both sides; a page that points to an alternate which does not point back is ignored for that pair. Codes must be valid language codes, optionally with a region, and the region alone is not enough.
Fix: generate a complete hreflangAnnotation for language/region versions. set for every page from one source, with self-references, return links and an x-default where a fallback exists.
Verify: spot-check pairs by hand and confirm that both versions reference each other with the right codes.
Prevent: keep code values in configuration rather than templates, and test sets with the methods in the hreflang guide.
Canonicals that cross borders
A Canonical TagTells search engines the preferred URL. tells Google which URL represents a group of duplicates. If the Australian page declares the UK page as canonical, Google is being asked to drop the Australian URL, and hreflang between them then has nothing to work with. Google's guide to consolidating duplicates covers how canonical signals are combined.
Fix: make every localised page canonical to itself, and use hreflang alone to express the relationship between locales.
Verify: URL inspection of each version shows its own URL as both declared and selected canonical.
Prevent: review canonical logic whenever a new edition is added, using the canonicalisation rules in crawling and indexing.
Visitors and crawlers redirected by location
Sending people to an edition based on IP address or browser language seems helpful, but Googlebot generally crawls from a limited set of locations and may never be allowed to reach the others. Google's guidance for multi-regional sites advises against automatic redirection based on perceived language.
Fix: let every URL load regardless of location, and offer a banner suggesting the visitor's likely edition along with a permanent language switcher.
Verify: request each edition from different locations and without cookies; every one loads directly, and each starts appearing in its market.
Prevent: write the no-forced-redirect rule into the international standards set out in the international SEO guide.
Editions too similar to tell apart
Two English editions with the same copy, the same prices and the same contact details give Google little reason to keep both, and it may treat one as a duplicate of the other. hreflang helps Google pick between alternates, but it works best when the alternates are genuinely localised.
Fix: localise what differs in reality: currency, prices, delivery terms, spelling, phone numbers, addresses, local proof. If nothing differs, consider one edition serving both markets.
Verify: both editions are indexed under their own canonical and earn impressions in their own country.
Prevent: define a localisation minimum for every new edition, the point the hreflang guide makes about what annotations will not fix.
Internal signals leaning one way
If menus, footers, in-content links and the XML sitemap all point at the primary edition, the secondary editions look orphaned.
Fix: keep navigation within each edition so that a German page links to German pages, add the language switcher on every page, and give each edition a sitemap that lists only its own URLs (with hreflang included there if that is how you declare it).
Verify: a crawl shows healthy inlink counts for each edition's key pages.
Prevent: design the folder or domain structure so that editions are self-contained, as the site architecture guide recommends.
Prevention: locale signals that agree
Google combines several signals to decide which version to show, and the wrong version wins when those signals disagree. Agree one source of truth for the list of editions, their codes and their URL patterns, and generate hreflang, canonicals, sitemaps and the language switcher from it. Ban forced redirects. Set a localisation standard before launching an edition, and treat a new edition as a migration with its own checklist. Monitor impressions by country for each edition monthly, and investigate when a page starts gaining impressions in a market it was not written for. When editions start to fold into each other without any cross-locale tag, the sibling page on duplicates without a user-selected canonical helps separate URL variants from genuine editions.
An online course provider with UK and US editions
A course provider runs /uk/ and /us/ editions in English. American searchers see the UK page for 70% of the priority queries. Inspection shows the US pages declaring the UK pages as canonical, left over from a template copied when the US edition launched in 2024, and hreflang present only on the UK side. Prices on the US pages are also still in pounds. The team makes each page self-canonical, adds reciprocal en-GB and en-US annotations with an x-default, and switches US pricing to dollars. Eight weeks later the US pages hold the main positions for 90% of those queries in the United States.
Common mistakes
Using a region code without a language code. Redirecting visitors by IP and assuming Googlebot will see every edition. Setting a canonical from every edition to the main site. Expecting hreflang to rescue editions that are copies of each other.
