hreflang in a Four-Language Country, What Actually Breaks

Two studies are worth putting next to each other before anyone touches a Swiss multilingual site. An Ahrefs analysis of 374,756 domains found that 67% of hreflang implementations carried issues. A separate study by Search Engine Land and NerdyData across 18,786 multilingual sites found 31.02% had conflicting hreflang directives. Those numbers describe the general web. Now apply them to a country that routinely runs three or four language versions of the same commercial page.

Switzerland is the market where hreflang errors compound rather than sit still. A mistake in a single-language site affects one page. A mistake in a de-CH, fr-CH, it-CH and en-CH cluster propagates across four, and because the versions reference each other, one broken reciprocal declaration can quietly unpick the whole set.

This article covers what hreflang does and does not do, the specific misunderstanding that causes most Swiss damage, how to structure locales, and what nobody has measured.

hreflang is a targeting annotation, not an authority pipe

This is the sentence that fixes most of the problem, and it deserves to be read slowly.

hreflang tells a search engine which language and regional audience a given URL is intended for. It does not transfer ranking authority between versions. It does not designate one version as the source of strength for the others. It does not consolidate signals the way a preferred-URL declaration does within a single language. Every language version accumulates its own authority, earns its own links, and competes on its own.

Get that wrong and a predictable sequence follows. A team decides the German version is the master, points the French and Italian versions at it as their preferred URL, and expects the German page's authority to carry the others. What actually happens is that the search engine is told the French and Italian URLs are duplicates of a German page, and it treats them accordingly. The versions that were supposed to inherit strength instead stop competing at all.

The correct configuration is unglamorous. Each language version declares itself as its own preferred URL. Each version lists every other version in its hreflang set, including itself. The declarations are reciprocal, so if the German page points to the French page, the French page must point back. And no version is canonicalised into another version's URL, because that is a statement about duplication, not about language.

What actually breaks, in order of frequency

From the error patterns those two studies describe, and from what we see on Swiss sites specifically, the failures cluster into a short list.

Missing return references. Version A points at version B, version B does not point back. The declaration is ignored. This is the single most common hreflang error on the web and it is more likely in a four-version cluster simply because there are more pairs to get wrong.

Preferred-URL declarations fighting the hreflang set. The most damaging pattern, and the one described above. The site simultaneously tells the engine that four pages serve four audiences and that three of them are duplicates of the fourth. The two statements contradict each other, and the duplication signal usually wins.

Wrong or invented locale codes. Switzerland invites this. The country code is ch, the language codes are de, fr, it and rm, and the combinations are de-CH, fr-CH, it-CH, rm-CH. Codes like ch-de, swiss, or a bare de intended to mean German-speaking Switzerland do not work. A bare de targets German speakers everywhere, which in practice means competing directly against a German web roughly ten times the size.

No x-default. With four locales and an international audience, users arrive who match none of the declared targets. Without an x-default declaration there is no stated fallback, and the engine picks one.

Partial coverage. hreflang declared on the homepage and top-level category pages, absent on the deep commercial pages where the money is. The cluster looks correct in an audit that samples the top of the site and is broken exactly where it matters.

Structure: one domain, four folders, no shortcuts

The structural question comes up before the annotation question, so it is worth settling. Three options exist for a Swiss multilingual build, and they are not equally sensible for most businesses.

Structure	What it does well	What it costs
Subdirectories on one .ch domain example.ch/de/, /fr/, /it/, /en/	All four locales build authority on a single domain. Simplest to maintain, easiest to get hreflang right, clearest Swiss signal from the ccTLD itself	Requires disciplined internal linking so one locale does not starve the others
Subdomains de.example.ch, fr.example.ch	Useful where locales are operationally separate teams with separate release cycles	Authority is more fragmented in practice, and four subdomains means four properties to maintain and monitor
Separate country domains example.ch alongside example.de	Strongest possible geographic separation, appropriate for genuinely distinct legal entities and product ranges	Each domain earns authority from zero. Only justified when the businesses really are separate, not just the markets
One version with an auto-translate widget	Nothing that survives scrutiny in a market this quality-sensitive	No indexable locale URLs, no per-locale keyword targeting, and machine German in a country that notices immediately

For most Swiss businesses the answer is the first row. Subdirectories on a .ch domain, four folders, complete reciprocal hreflang, each folder with its own researched keyword set. The ccTLD is already doing geographic work that no annotation can replicate, and keeping everything on one domain means the authority you earn in German is at least on the same property as the French pages, even though it does not transfer to them.

The measurement nobody has done

Practitioners routinely claim that a well-configured cluster passes some proportion of authority between language versions. We could not find that figure anywhere, in four independent research passes, from any named source.

No published study quantifies how much authority moves between de-CH, fr-CH and it-CH versions of the same page. The reason is that hreflang is not a transfer mechanism in the first place, so the question is partly malformed. What links across versions do is more mundane: they help crawling, they clarify relationships, and internal links between locales pass the same signals internal links always pass. That is not the same as a language version inheriting the strength of another.

Two consequences follow for planning. First, budget each locale as its own project, because each one earns its own position. Second, be suspicious of any agency that quotes a percentage here. If the figure existed, all four of our research passes would have found it, and none did.

An audit order that finds real problems

Most hreflang audits sample the homepage, find it correct, and stop. That is how a broken cluster passes inspection.

A better order starts at the bottom. Pull the ten deepest commercial pages, the ones a buyer lands on with intent, and check the hreflang set on each of those first. Then verify reciprocity by walking every declared pair in both directions rather than trusting the tool's summary. Then check whether any locale URL is pointed at another locale as its preferred version, which is the failure that costs the most and shows up least. Then check the locale codes character by character. Then look for x-default.

Only after that is it worth looking at the homepage, which is almost always the page that was configured most carefully and tells you least about the rest of the site.

Tessar Napitupulu covers multilingual and multi-market search architecture, including the trade-offs between structural separation and consolidated authority, in Found Before They Search, available as a free gated edition, with retailer editions on Amazon, Google Play and Apple Books.

---

Frequently Asked Questions

Does hreflang transfer ranking authority between language versions?

No. hreflang is a targeting annotation that tells a search engine which language and regional audience a URL is intended for. It does not consolidate or transfer ranking signals between versions. Each language version accumulates its own authority, earns its own links and competes on its own. The common failure is treating one version as a master and pointing the others at it as their preferred URL, which tells the engine those versions are duplicates rather than audience variants. No published study quantifies any percentage of authority moving between de-CH, fr-CH and it-CH versions, and we could not locate one in four independent research passes.

How common are hreflang errors in practice?

Common enough that assuming your implementation is correct is the wrong default. An Ahrefs analysis of 374,756 domains found 67% of hreflang implementations carried issues. A separate Search Engine Land and NerdyData study of 18,786 multilingual sites found 31.02% had conflicting hreflang directives. Those figures describe the general web rather than Switzerland specifically, but the risk is higher in a four-locale cluster because there are more reciprocal pairings to get wrong and one broken return reference can cause declarations to be discarded.

Should we use subdirectories, subdomains or separate country domains for Switzerland?

Subdirectories on a single .ch domain suit most Swiss businesses. All four locales build authority on one property, hreflang is simplest to configure correctly, and the ccTLD itself supplies a geographic signal no annotation can replicate. Subdomains make sense where locales are operationally separate teams with separate release cycles, at the cost of more fragmented authority and four properties to monitor. Separate country domains are justified only where the businesses are genuinely distinct entities with distinct product ranges, since each domain then earns authority from zero.

Which locale codes should a Swiss site use?

de-CH, fr-CH, it-CH and en-CH, with rm-CH only where Romansh content genuinely exists. Codes such as ch-de, swiss or a bare de do not do what teams expect. A bare de targets German speakers everywhere, which places a Swiss page in direct competition with a German web roughly ten times the size, and that is close to the opposite of the intended effect. An x-default declaration should also be present, since a four-locale cluster will receive visitors who match none of the declared targets.

Can we launch all four language versions at the same time?

You can, and it usually performs worse than sequencing them. Four thin folders published together compete for attention, share production capacity, and give you no clean performance window to attribute results to any single launch. A better order is de-CH built properly, then the locale your revenue argues for, then the remaining commercial locale, then it-CH or a documented decision to skip it. The step teams miss is that every new locale changes the required hreflang set on all existing locales, so declarations need updating at each phase rather than at the end.

Is machine translation acceptable for the French and Italian versions?

For a Swiss commercial site, not as the finished product. The market is quality-sensitive and readers notice, and machine output misses the vocabulary and formatting conventions that signal local production, including Swiss French numerals and the apostrophe thousands separator. Machine translation as a first draft under native review is a defensible compromise if the client understands which locales are natively produced and which are adapted. Presenting adapted content as native production is where the credibility damage happens, and it happens on the deep commercial pages where buyers are paying closest attention.

Where should an hreflang audit start?

At the deepest commercial pages, not the homepage. Pull the ten pages a buyer lands on with purchase intent and check their hreflang sets first, because the homepage is almost always the page that was configured most carefully and therefore tells you least about the rest of the site. Then walk every declared pair in both directions to verify reciprocity manually rather than trusting a tool summary, check whether any locale URL is pointed at another locale as its preferred version, verify the locale codes character by character, and confirm x-default is present.