International SEO: Markets Beyond Hreflang - Ighenatt Blog

Hreflang is a routing annotation, not an international growth plan. It can help a search engine understand that two URLs are alternatives for different languages or regions. It cannot tell you whether people in the Netherlands search for the category in English or Dutch, whether your UK offer works in Ireland, whether a price includes the right tax, or whether a visitor can complete checkout from the country you appear to target.

International SEO starts with markets, not tags. Choose where the product can genuinely compete, learn how people there describe the problem, decide which combinations of language and region deserve their own experience, and build a publishing system that can keep those experiences accurate. Hreflang comes later, when there are real alternatives to connect.

Google makes the distinction explicit: a multilingual site offers content in more than one language, while a multi-regional site targets users in different countries or regions. A site can be either or both (Google Search Central). Language answers “in what language?”. Market answers “for whom, under which commercial conditions?”.

If you need the syntax, reciprocity rules and implementation methods, use our technical guide to hreflang. This article deals with the decisions that determine whether those annotations point to meaningful alternatives in the first place.

Start with a market hypothesis, not a translation queue

“We receive visits from France” is not yet a case for a French site. Those visits might come from branded queries, expatriates, researchers, bots, or people who cannot buy the product. Equally, a market with little current traffic may be attractive because the site has never supplied a page that matches local demand.

Write a short market hypothesis before commissioning content. It should name the target customer, problem, product availability, supported language, commercial constraint and evidence that would justify investment. For example: “German-speaking operations managers in Germany can purchase and onboard without a local sales entity; non-branded demand exists for the problem category; a German content and demo journey could produce qualified opportunities.” Every clause can be tested.

Use four evidence groups:

1. Demand: query themes, seasonality, search-result formats and the language used for the problem and solution.
2. Commercial fit: product availability, margin, payment, delivery, support hours, sales coverage and realistic conversion paths.
3. Competitive conditions: the pages that satisfy the query now, their offer type, local proof, information depth and brand strength.
4. Operational feasibility: translators or local editors, subject expertise, legal review, engineering capacity and an owner after launch.

Do not turn a keyword volume export into a revenue forecast. Search tools estimate different datasets and can hide low-volume specialist terms. Use them to compare vocabulary and relative opportunity, then check the actual results and speak to people who handle sales, support or customers in the market. The output is a decision about whether to test a market, not a claim that a country will deliver a fixed traffic number.

A useful first release is narrow but complete: one market, one language, one coherent product journey and a small group of high-intent pages. Launching twelve languages with a translated homepage and no local pricing, support or product documentation creates apparent coverage while leaving the customer journey broken.

Research the words people use in that market

Translation answers how to express a source sentence. International keyword research asks which concepts people search for and what they mean locally. Those are related jobs, but they are not interchangeable.

Build a terminology set from the market itself. Begin with the problem, category, alternative solutions, jobs, product features, industries and purchase qualifiers. Review autocomplete and search results, but also use sales calls, support tickets, local product interfaces, professional associations and competitor navigation where available. Record the preferred term, acceptable variants, rejected meanings and evidence.

One language can carry different market vocabularies. “Holiday” and “vacation”, “solicitor” and “attorney”, or “mobile” and “cell phone” are familiar examples, but the consequential differences are often specialist: a regulated service name, a procurement term, an educational stage or the label on a form. Conversely, a market may routinely search for an English product category while expecting explanations and support in another language. Do not replace that mixed reality with a blanket rule.

Map each query to a local intent before mapping it to a source URL. Ask:

• Is the searcher learning, comparing, buying, troubleshooting or navigating?
• Do the current results favour a guide, service page, directory, calculator, product page or official source?
• Does the phrase identify a category in this market or a different concept?
• Are location words part of normal demand, or would adding them create awkward copy?
• Can your business fulfil the promise implied by the query?

This work often changes the page set. A source-market comparison page may need to become an eligibility page elsewhere. A broad guide may split because local buyers use two distinct category terms. A feature page may not deserve a local version because the feature is unavailable there. These are content-model decisions, not translation defects.

Decide whether the locale needs a language or a region

A locale is more than a language label. Unicode’s locale model associates an identifier with preferences for dates, numbers, currencies, measurement units, sorting and other conventions, and it allows language identifiers to include region and script subtags (Unicode CLDR). That does not mean every possible locale deserves an indexable SEO page. It means software and editorial teams must know which conventions they intend to support.

Use the smallest segmentation that preserves a truthful, usable experience:

Maintenance settles the question. If the team cannot explain who owns en-GB and en-IE, which fields may differ, and how releases reach both, it probably cannot support two regional editions safely. Start with en or the best-supported market and add a region when evidence shows that shared content is causing a real problem.

Regional segmentation also has a downside. Google notes that country targeting can help in the target country at the expense of results in other locales or languages, and describes geotargeting as inexact (Google Search Central). Treat a country-specific architecture as a trade-off, not a universal upgrade.

Choose a URL architecture the organisation can operate

Google recommends distinct URLs for language versions instead of changing the page solely through cookies or browser settings. Its international guidance describes country-code domains, subdomains and subdirectories as workable structures, with different costs and separation properties; URL parameters are not recommended for this segmentation (Google Search Central).

The SEO debate often stops at which structure is “best”. In practice, the operating model decides. Choose the structure that gives every supported locale a stable, crawlable URL and can be governed without divergent platforms.

• Country-code domains give users and search engines a clear country signal, but require separate domain operations and can only target one country each.
• Subdomains allow infrastructure or team separation, but may make the language-versus-country meaning less obvious to users.
• Subdirectories keep deployment and authority on one host and are usually simpler to maintain, though separating sites is harder.

Keep slugs local when it helps comprehension, but manage them in a locale registry rather than deriving them with string replacement. A page identity should connect the English, Spanish and Catalan slugs even when none is a literal translation. That same identity can drive alternate annotations, language-selector links, sitemap membership and parity reports.

Do not let canonical tags collapse genuine market variants onto a global source page. If two pages are distinct, indexable alternatives, each normally needs to be canonical in its own right and participate in the alternate set. If they are not distinct enough to maintain or index, reconsider why both exist rather than using contradictory signals to preserve the appearance of localisation.

Let people choose, and keep their choice

A visitor may be travelling, using a corporate network, living abroad, reading a second language or purchasing for another country. Location and language preference are clues, not identity.

Google advises sites to avoid automatic language redirection and to provide links that let people select another version. It also warns against IP-based content adaptation because location inference is unreliable and Google may not crawl all variants; Googlebot generally crawls from the United States and does not send an Accept-Language header (Google Search Central). Its guidance on locale-adaptive pages warns that Google might not crawl, index or rank every variant served from one URL, and recommends separate locale URLs with hreflang (Google Search Central).

Use a selector that distinguishes language from country when both matter. “English” is a language choice; “United Kingdom” can change availability, price and delivery. Labels should be readable in their own language, and every option should be a normal link to a stable URL. On mobile, do not bury the selector behind a fragile script or remove it from the footer.

A non-blocking suggestion can be helpful: “You are viewing the Ireland site. Go to United Kingdom?” The visitor must be able to dismiss it, remain on the current page and change later. Store an explicit choice where appropriate, but never make a cookie the only way to reach content. Preserve deep-page context where an equivalent exists; otherwise explain that the requested page is unavailable instead of silently sending everyone to a homepage.

Test the awkward cases: unsupported countries, crawlers without cookies, disabled JavaScript, VPNs, a shared device, a product available in one market only and a visitor switching midway through checkout. A selector is part of the product, not a decorative flag menu.

Define content parity and controlled adaptation

Two alternate pages need a shared identity, not identical sentences. The aim is semantic parity for the core entity and offer, with deliberate adaptation where the market changes what a reader needs.

Create a field-level contract:

Prices, taxes, cancellation terms, warranties, privacy wording and regulated claims are not copy flourishes. Route them to the appropriate business or legal owner for the target jurisdiction. An SEO editor should confirm that the page exposes the approved information clearly, not improvise legal equivalence from a source-market template.

Localise components around the article too: navigation, breadcrumbs, filters, form errors, confirmation emails, downloadable files, image alt text and customer-support paths. Google says it determines page language from visible content and recommends using one language for page content and navigation rather than placing side-by-side translations on a page (Google Search Central). A translated body inside an English checkout is not a finished locale.

Sometimes parity should break. A market may need a dedicated explanation of shipping zones, a local integration, a different industry example or an unavailable-feature notice. Record the reason and owner. Untracked divergence is dangerous; documented adaptation is the purpose of localisation.

Build local discovery and authority into the release

Publishing a locale creates an isolated island unless the rest of the site can reach it. Add the version to relevant navigation, locale hubs, category pages and contextual internal links. Use anchors a local reader would understand. The language selector connects equivalents; it does not replace normal information architecture.

External authority is also market-specific. Google lists local addresses and phone numbers, local language and currency, links from other local sites and Business Profile signals among the signals it may use to understand intended audience (Google Search Central). That is not a recipe to fabricate offices or buy country links. It is a reminder that a credible market presence produces coherent local evidence.

Start with relationships the business already has: distributors, professional bodies, technology partners, local events, universities, trade publications and customers who have explicitly approved a reference. Give them something worth citing, such as original documentation, a useful tool, market data with a transparent method, or a genuinely local resource. Translate-only outreach tends to reveal the same weakness as translate-only content: no reason for that market to care.

Keep business listings and local service pages factual. An address, service area or local team must exist. If support is remote, say so. The objective is to make the operating model legible, not to simulate localness for a ranking effect.

Measure by market without confusing country and language

The measurement unit should match the decision. A global traffic total cannot tell you whether a German launch works, and a country report alone cannot tell you whether visitors preferred German or English.

Build a scorecard with three joined views:

1. Search demand and visibility: queries, pages, country, device, impressions, clicks and average position.
2. On-site behaviour: landing-page locale, observed country, selector changes, engagement with the intended journey and failures such as unavailable checkout.
3. Business outcome: qualified lead, trial, sale, revenue, margin, activation or another outcome the market hypothesis named.

The Search Console Search Analytics API supports grouping by dimensions including country, device, page and query, subject to its data limits (Google Search Console API). The GA4 Data API schema includes dimensions such as country, language, landingPagePlusQueryString and pageLocation (Google Analytics). These fields make market analysis possible, but they do not remove ambiguity. Browser language is not a declaration of buying market; geolocation can be wrong; consent and thresholding may limit analytics; and Search Console aggregates search data rather than reporting individual people.

Annotate launches and material changes. Compare the locale with its own pre-launch baseline, the source locale and a control group of pages where possible. Look beyond session growth to query mix and qualified outcomes. A regional page can attract fewer visits but produce a better-supported customer journey. It can also cannibalise another edition because the architecture or alternate set is wrong. Both require diagnosis, not a celebratory traffic chart.

Set review thresholds in advance: no impressions after a reasonable crawl period, a high rate of immediate locale switching, leads from unsupported countries, recurring queries in an unserved language, or conversion loss at a market-specific step. Thresholds should trigger investigation; they are not universal SEO benchmarks.

Govern the locale as a product surface

International sites decay quietly. A source page changes, one translation stays old, a price component falls back to dollars, a regional URL redirects to the global homepage, and hreflang still claims the pages are equivalents. Preventing that drift is a governance job.

Assign four responsibilities even if one person holds several:

• Market owner: decides whether the offer and journey remain valid locally.
• Editorial owner: maintains terminology, source meaning and local usefulness.
• Technical owner: maintains URL identity, rendering, canonicals, alternates, sitemaps and selector behaviour.
• Measurement owner: maintains the scorecard and investigates anomalies.

Store page relationships in structured data, not spreadsheets that publishing cannot read. Each record should contain the page identity, locale, market, slug, status, source version, owner, last review, allowed adaptations and alternate URLs. Generate technical annotations from that registry. If a translation is missing or unpublished, do not advertise it as an alternate.

Use a release gate that checks:

1. The URL returns the intended status and is indexable when it should be.
2. Canonical and alternate relationships resolve to published equivalents and return links exist.
3. Title, description, heading, navigation and body use the intended language.
4. Product facts, prices, units, availability and approved legal components match the market.
5. Internal links and the selector lead to equivalent or explicitly explained destinations.
6. Structured data describes the visible local page rather than copied source-market values.
7. Analytics records locale and page identity without exposing personal data.
8. A native or market-competent reviewer has checked the complete journey on mobile and desktop.

Automate what is deterministic, then keep human review for meaning. A crawler can find missing return links and mixed-language titles. It cannot decide whether a Catalan procurement term sounds credible, whether a UK case study answers a Spanish buyer’s objection, or whether a local call to action promises support the company does not provide.

A durable order of operations

International SEO becomes manageable when the sequence mirrors the real dependencies:

1. Qualify the market. Confirm demand, commercial fit and operational ownership.
2. Model language and region. Choose the minimum set of locales that users and the business genuinely need.
3. Research local intent. Build terminology and page mappings from market evidence.
4. Design the system. Set stable URLs, page identities, selector behaviour and adaptation rules.
5. Publish a complete journey. Include navigation, product facts, proof, forms, support and required market information.
6. Connect alternatives. Add canonical, hreflang and sitemap relationships from the locale registry.
7. Validate and measure. Test the rendered experience and monitor search plus business outcomes by market.
8. Maintain parity deliberately. Propagate source changes, review approved divergence and retire unsupported locales cleanly.

Frequently asked questions about international SEO

Is hreflang enough for international SEO?

No. Hreflang describes relationships between alternate language or regional URLs. It does not research local demand, translate an offer, set prices or units, create local links, satisfy market-specific obligations, or measure commercial outcomes. Implement it after deciding which real market versions should exist.

Does every same-language country need a separate page?

Only when the user need or business operation changes materially enough to justify a maintained page. Different products, availability, terminology, currency, delivery, regulation or conversion steps can support a regional variant. If those things are substantially the same, one language page may be clearer and easier to keep current.

Should a site redirect by IP or browser language?

Do not make automatic redirection the only path. Google advises against IP-based adaptation and recommends distinct URLs plus links that let users choose another language or region. A suggestion banner can help, but preserve the chosen version and keep every supported locale reachable.

Should translated pages have identical content?

They should preserve the same core promise and factual truth when they represent the same offer, but they do not need sentence-level identity. Search terminology, examples, proof, units, prices, fulfilment details and calls to action may need local adaptation. Record those fields explicitly so adaptation does not become accidental contradiction.

How should international SEO performance be measured?

Use a market scorecard that combines queries and landing pages with country, device and business outcomes. Search Console can group search performance by dimensions such as country and page; analytics can report country and page location. Interpret both carefully: observed country is useful evidence, not proof of language preference or market eligibility.

The next step is not to translate the whole site. Select one candidate market and audit a single high-intent journey from search result to completed action. List every point where terminology, offer, availability, format, evidence or ownership changes. That inventory will tell you whether you need a language edition, a regional edition, or more operational work before either one is honest.