hreflang on Multilingual Stores: The Reciprocity Rule
One missing annotation can void the whole set. A ten-minute verification matrix, the ecommerce-specific failures, and what x-default should actually point at.
hreflang is the most commonly half-implemented tag in ecommerce. It is also the one where a single omission can void the whole set, which makes it worth checking properly rather than trusting a plugin.
The rules that actually matter
Every page must reference every language, including itself. A German product page lists all eight languages plus itself, plus an x-default.
Annotations must be reciprocal. If the German page points at the English one, the English page must point at the German one. One-way annotations are ignored.
Every page in the set needs at least one annotation without hreflang — the self-reference — or the cluster is considered incomplete.
URLs must be absolute and canonical. Point hreflang at the canonical URL, not a tracking variant, not a redirected URL.
The reciprocity rule is where implementations fail. A plugin that outputs tags only on translated pages produces a set where the default-language page is silent, and search engines may then ignore the annotations for the group.
How to verify in ten minutes
- Pick one product that exists in every language
- Extract its hreflang block from each language version
- Build a matrix: does every row list every column?
- Check that each listed URL returns 200 and is self-canonical
- Check that x-default appears exactly once per page and points at the same target everywhere
If the matrix has gaps, the implementation is broken regardless of what the plugin reports.
Which pages need it, and which do not
Annotations belong on pages that have genuine translated equivalents: home, categories, products, key content pages, contact, policy pages.
They do not belong on pages that exist in one language only. If a blog post is English-only, do not annotate it as available in German. A page that points at a non-existent translation is worse than no annotation.
The practical consequence for most stores: translate the commercial pages properly, and leave the blog in one language rather than machine-translating it into eight. That decision is covered in the multi-language structure notes.
Common ecommerce-specific failures
Products not translated but annotated. The plugin assumes every product exists in every language, so a product added only in English gets annotations pointing at German URLs that render the English content. Duplicate content across eight URLs, annotated as distinct.
Pagination inside the set. Category page 2 in German should annotate to category page 2 in French, not to the French page 1. Mapping everything to page 1 creates a cluster where most translations are unreachable through hreflang.
Language switcher that drops the context. A switcher that sends the user to the homepage in the target language is a UX failure, and it also removes the internal link path between equivalents. The switcher should link to the equivalent page, and those links are the crawl paths that keep the cluster coherent.
Currency-based language selection. Switching language when a customer changes currency is a common convenience feature that breaks hreflang, because it changes URLs based on session state. Keep them independent.
x-default, explained properly
x-default is the fallback for users whose language does not match any annotation. Point it at the version you would show a stranger: usually the English page, sometimes the largest market’s language.
Two rules: it must be the same throughout the set, and it should be a page that exists. Pointing x-default at a language the customer cannot read is not an improvement over letting search engines decide.
What hreflang does not do
It does not translate anything. It does not improve ranking in a country on its own. It does not stop duplicate content issues if the translations are machine-generated and identical in structure.
What it does: tells search engines these pages are the same offer in different languages, so the right one appears for the right user. On a store where each market matters commercially, that is worth having correct — and worth verifying rather than assuming.
If you maintain eight languages, check the matrix after every plugin update that touches SEO output. These tags break quietly: nothing on the storefront looks different, and the damage only shows as slowly declining visibility in non-default markets. That kind of silent failure is exactly what a technical SEO review is meant to catch.