Hreflang Mistakes That Are Costing You International Rankings

Your French page ranks in Canada. Your US page appears in the UK. Google swaps the English homepage into markets where you spent months building local content. Nothing looks broken, but organic traffic tells a different story. This is what hreflang mistakes often look like. They quietly give Google conflicting instructions about which page belongs in which market, leading to weaker visibility, the wrong search landing pages, and wasted localisation investment.

The Scale of the Problem: 75% of Sites Get Hreflang Wrong

Hreflang connects alternate versions of a page so Google can serve the right language or regional URL. In practice, every page in the set must agree, use valid codes, point to canonical URLs, and stay accurate as pages change.

A widely cited Hreflang.org study found that roughly 75% of websites using hreflang had at least one technical error, from invalid codes and missing return tags to redirected alternate URLs. Broken implementation is closer to the default than the exception.

The commercial impact goes beyond rankings. CSA Research surveyed 8,709 consumers in 29 countries and found that 76% preferred products with information in their language. Forty percent would not buy from a website in another language. Send a Spanish-speaking shopper to an English product page and you may lose the visit and sale.

The Mistake That Costs the Most: Missing Reciprocal Tags

Hreflang is a confirmed relationship, not a one-sided recommendation. If an English page names a French equivalent, the French page must point back. Without that return link, Google may ignore or misinterpret the annotations.

Imagine these two pages:

• example.com/en/pricing
• example.com/fr/tarifs

The English page points to French with hreflang=”fr”, but a template update removes the French page’s hreflang=”en” reference. Both pages still work for visitors, so the error survives unnoticed. To Google, the relationship is incomplete.

Each page must also include itself. Both pages should reference the English and French URLs. If you have a general selector, x-default can identify the fallback for users who match none of the listed values.

These hreflang implementation errors multiply with scale. A six-language product set needs the same complete annotations on all six pages. Add a seventh market without updating the first six and the cluster becomes inconsistent. Hashmeta reported an ecommerce expansion across six Southeast Asian markets that lost 64% of organic traffic after a broken implementation. The report offers limited detail, but it shows why reciprocity must be tested before launch.

The Canonical Conflict Nobody Checks First

Sometimes the hreflang looks correct and Google still shows the wrong page. The problem may sit in the canonical tag.

Suppose example.com/fr/produit declares the English page as canonical while claiming to be its French alternate. The canonical says “consolidate this duplicate into English.” Hreflang says “show this distinct version to French users.” Google must decide which URL belongs in its index first, so the French page may disappear before hreflang can help.

This often happens when developers copy an English canonical across every locale. Each indexable language page should normally use a self-referencing canonical, while hreflang points to canonical URLs in other markets. Redirected, blocked, noindexed, or non-canonical URLs do not belong in the set.

Check canonicals before treating the problem as hreflang alone. A technical SEO audit should compare canonicals, indexability, status codes, sitemaps, and hreflang together. Otherwise, you may repair the annotations but leave the real conflict untouched.

ISO Codes Are Stricter Than Most Teams Assume

Google expects an ISO 639-1 language code followed, when needed, by an ISO 3166-1 Alpha-2 region code. Language comes first, separated from the optional region by a hyphen.

Hreflang value	Valid?	What it means
en-GB	Yes	English for users in the United Kingdom
en-UK	No	UK is not the supported region code; use GB
es-MX	Yes	Spanish for users in Mexico
es-LATAM	No	LATAM is not an ISO Alpha-2 region code
zh-Hans	Yes	Chinese written in the Simplified script
zh-CN	Yes	Chinese content intended for China

The final two examples are not interchangeable. zh-Hans identifies a script, while zh-CN identifies a region. A Simplified Chinese page for Singapore may need a different target from one for mainland China. Three-letter language codes and region-only values are invalid.

A bad code does not break the page. Google ignores the unsupported annotation, so en-UK can survive for months while teams investigate content and links instead.

ccTLD vs Subdirectory: The URL Decision That Compounds Over Time

Hreflang cannot rescue a confusing site structure. First decide where each market will live.

A country-code domain such as example.de gives a clear German targeting signal and separates markets with different products or regulations. The trade-off is cost. Each domain needs its own setup, monitoring, content operation, and link authority.

Subdirectories such as example.com/de/ keep every market under one domain. Authority, infrastructure, and analytics are easier to consolidate, making this the practical default for many businesses. The central platform must still support every locale cleanly.

A retailer with separate inventory and legal entities may justify ccTLDs. A SaaS company selling one platform globally will often benefit from subdirectories. Decide before thousands of URLs and backlinks depend on the structure. Reversing it later creates redirect, canonical, and hreflang risks. Vicious Marketing’s enterprise SEO services guide covers the wider planning required at this scale.

How to Actually Audit Hreflang at Scale

The International Targeting report in Search Console was deprecated in 2022, so it cannot support a 2026 audit. Search Console can still show whether individual pages are indexed and which canonical Google selected, but another tool must validate complete hreflang clusters.

Screaming Frog can crawl hreflang in HTML, headers, and XML sitemaps. Sitebulb visualises broken relationships and indexability conflicts. JetOctopus suits millions of URLs and server-log data. Whatever you use, ask: does every page reference itself, do alternates return the relationship, are codes valid, and are all referenced URLs indexable, canonical, and returning 200?

Audit by template and locale, not isolated URL. If 20,000 French products lack German return tags, the fault probably sits in a template or deployment rule. Manual fixes only hide the pattern.

An international SEO audit should compare hreflang with sitemaps, canonicals, redirects, robots directives, internal links, and rendered HTML. A tag that disappears after JavaScript rendering is different from one pointing to a noindexed page. 

Frequently Asked Questions

What percentage of websites have hreflang errors?

A widely cited Hreflang.org study found errors on roughly 75% of websites using hreflang. Missing return tags, invalid codes, canonical conflicts, and non-indexable alternates remain common.

What happens if hreflang tags aren’t reciprocal?

Google may ignore the annotations. If Page A identifies Page B as an alternate, Page B must link back so Google can confirm the relationship.

Can a canonical tag break my hreflang implementation?

Yes. If a translated page canonicalises to another locale, Google may consolidate it. Use canonical, indexable URLs throughout the set.

Should I use ccTLDs or subdirectories for international SEO?

Use ccTLDs when country separation justifies separate authority and infrastructure. Subdirectories are easier to manage and consolidate authority for most international businesses.

What tools can audit hreflang across an entire site?

Screaming Frog, Sitebulb, and JetOctopus can validate codes, return links, self-references, canonicals, and indexability. Choose based on site size and implementation method.

Running hreflang across multiple markets and not sure it’s actually working? Get a technical SEO audit from Vicious Marketing →