27 checklist article

International SEO Checklist for Global Sites

An international SEO checklist is a repeatable set of decisions and technical checks that keep a multilingual or multi-region site crawlable, correctly targeted and measurable. Most teams treat it as a translation task. It isn't. It's an architecture task first, a content task second, and a reporting task last. Work through the steps below in order, because choices you make in step one (market and domain) constrain everything after it.

One warning before you start: this checklist covers search visibility, not legal or translation compliance. Naming conventions, data residency, consumer law and certified translation are separate workstreams. Get a lawyer and a native-language editor involved where those apply.

1. Decide which markets you are actually targeting

International SEO fails most often at the strategy stage, not the technical stage. A manufacturer with a strong German catalogue and a weak Spanish one should not launch twelve language versions at once. Pick two or three markets where you already have demand signals: existing inquiries, distributor requests, trade show contacts, or impressions in Google Search Console from a country you don't yet serve.

For each market, write down three things: the language, the country or region, and the buyer's search behaviour. A buyer in Mexico searching in Spanish is a different target from a Spanish-speaking buyer in Texas. Same language, different market, potentially different content. If you're building this out for the first time, our international SEO service starts with exactly this scoping conversation, and the international SEO strategy guide walks through the market-selection logic in more detail.

Market decision questions

  • Do we already receive inquiries from this country, even without a localised site?
  • Is the search language the same as the business language in that market?
  • Do we have anyone who can review content in that language, or do we need to hire?
  • What is the realistic support model if inquiries triple from that region?

2. Choose your domain and URL structure

There is no universally correct answer here, only trade-offs. Google Search Central documentation explains that country-code top-level domains (ccTLDs) and subdirectories both work for international targeting, and that Google generally relies on content and hreflang signals rather than the domain alone. That matters because it means you don't have to buy a .de to rank in Germany.

StructureExampleBest forMain drawback
ccTLDexample.deStrong local trust signal, separate hosting and legal entityEach domain builds authority from zero; expensive to maintain
Subdirectoryexample.com/de/Consolidated authority, one CMS, cheapest to runWeaker local signal; needs clean hreflang
Subdomainde.example.comSeparate technical stacks or teamsAuthority is partly split; more setup work
Parameterexample.com?lang=deRarely worth itPoor crawl clarity, hard to manage at scale

For most B2B manufacturers and exporters, subdirectories win on cost and authority consolidation. Pick ccTLDs only when you have a genuine local operation, local staff and a reason to look local. Whichever you choose, keep the pattern consistent across every market. Mixed structures are where hreflang implementations go to die.

FREE Want a first read of your own site against these points? We check crawlability, keyword coverage and competitors. Get a Free SEO Analysis

3. Implement hreflang tags correctly

Hreflang tags for international SEO are annotations that tell search engines which language and regional version of a page to serve to which audience. They are not a ranking factor in themselves. They are a disambiguation signal. Get them wrong and you create duplicate-content confusion; get them right and each market sees the right page in results.

Three rules cover most of the mistakes we see. First, hreflang must be reciprocal: if page A points to page B, page B must point back to page A. Second, every hreflang cluster needs a self-referencing tag. Third, use the correct language and region format, for example en-GB and en-US, or es for generic Spanish. Google Search Central documentation states that hreflang can be implemented in HTML head, HTTP headers or sitemaps, and that all three are valid. Pick one method per site and stick to it.

  1. List every language and regional version of every page you intend to serve.
  2. Assign each one a correct language or language-region code.
  3. Add a self-referencing hreflang tag to each page.
  4. Add reciprocal tags pointing to every other version in the cluster.
  5. Add an x-default tag pointing to the page you want served when no language matches.
  6. Validate the whole cluster with a crawler before launch, not after.
  7. Re-check after every new market launch, because clusters break silently.

If you're auditing an existing site, start with the biggest cluster, usually English, and trace it manually. Automated tools flag missing return tags, but they rarely catch a wrong region code that still resolves.

4. Localise content, don't just translate it

Localisation is where the commercial value sits. A translated page says the same thing in another language. A localised page answers the questions that market actually asks, uses the units, certifications and terminology that market expects, and links to local case examples where you have them. Machine translation with no native review is the fastest way to lose trust with a technical buyer.

Three practical rules. Write the localised page for the local query set, not the English one. Keep product names consistent where they are brand names, and translate them where they are descriptive. And check that your structured data (Schema.org markup for products, organisations and FAQs) is present on every language version, because AI search systems and rich results both read it. Schema.org publishes the vocabulary; your implementation still has to be correct per page.

This is also where AI search visibility gets decided. OpenAI's published help documentation explains that ChatGPT answers either from live web search or from knowledge stored in the model without web access. Only the first is influenceable through published, crawlable content. If your German pages are thin translations with no local substance, they won't be cited in German-language AI answers either.

5. Set up measurement before you launch

Measurement is the step teams skip, and it's the one that tells you whether any of the above worked. Set up a separate view or property per market in Google Analytics, and segment Google Search Console by country and language. Without that split, a strong English performance will hide a dead German one for months.

Track four things per market: impressions and clicks by country, average position for your priority local keywords, inquiries by source, and which pages get cited in AI answers. That last one is newer and less standardised. OpenAI's help pages describe how search-enabled answers surface and cite sources; there is no official citation report, so teams monitor manually. Our working rule is a weekly check of ten to twenty core queries per market in ChatGPT search mode, logged out, with screenshots kept as evidence.

In one RAGSEO client program (client anonymized), a lifting equipment manufacturer saw AI-engine-driven inquiries reach 186, which was 35% of all inquiries, with 62% of those coming from Europe and North America at a 28% higher conversion rate than traditional channels. Before the project, the brand appeared in less than 1% of AI-generated results. That's a GEO program, but it ran on the same foundation: clean architecture, correct language targeting, locally relevant content.

6. Audit quarterly, not once

International sites decay. New markets get bolted on without hreflang updates, a redesign drops a language, a CDN change slows one region. A quarterly audit catches this before rankings do. Re-crawl the full hreflang graph, spot-check ten localised pages per market for accuracy and load speed, review Search Console country data for sudden drops, and confirm that every new page published since the last audit is in the correct cluster.

If the technical side is solid but the content pipeline isn't producing localised pages fast enough, that's usually a resourcing problem rather than a strategy problem. Our SEO services for manufacturers and SEO pricing pages lay out what a monthly content and optimisation cadence looks like, including how many articles and backlinks each plan covers as of September 2026.

One last thing. Don't treat hreflang as a one-time launch task. Treat it as infrastructure you maintain, like your sitemap or your SSL certificate. The sites that rank across ten markets are the ones that keep the plumbing clean while everyone else is arguing about translation vendors.

Frequently asked questions

Do I need a ccTLD to rank in a specific country?

No. Google Search Central documentation states that both ccTLDs and subdirectories can work for international targeting, and Google generally relies on content and hreflang signals rather than the domain extension alone. A subdirectory like example.com/de/ is usually cheaper and consolidates authority better for most B2B manufacturers.

How many hreflang tags does one page need?

One for every language and regional version in that page's cluster, including a self-referencing tag for the page itself, plus an x-default tag pointing to the fallback version. If you serve five language versions of a page, that page carries six hreflang tags: five language versions plus x-default.

Can I just use machine translation for localised pages?

You can publish it, but technical buyers notice. Localisation means adapting terminology, units, certifications and examples to the market, not swapping words. Machine translation with no native review tends to produce pages that neither rank for local queries nor earn trust from buyers comparing suppliers.

How long before an international SEO program shows results?

For most manufacturers, significant improvements in rankings and organic traffic can usually be observed within 3 to 6 months, though new markets and new domains take longer because they start with no authority. Measurement should be split by country and language from day one so you can see which market is moving.

Sources