Customer Impact

SEO

Fixing hreflang errors: return tags, x-default and many-to-many

Copy for AI

Most hreflang errors come from three mistakes: return tags, x-default and incomplete sets.

Your tags are in place, but the wrong language version still ranks. That’s the frustrating part: a tag that looks fine can be ignored completely.

The verdict: fixing hreflang errors is cluster work, not page work. Repair every language version in one go, or the problem comes straight back.

This page is the troubleshooting layer. If you’re not yet clear on what the tags do, read our explanation of hreflang tags first.

Or zoom out to the basics with our guide on what SEO is.

Already know all that? Then you’re in the right place. We’ll work through the errors behind nine out of ten problems:

  • missing return tags
  • a wrong x-default
  • broken many-to-many sets
  • canonicals that fight your hreflang

How do you diagnose hreflang errors first?

Find out which error you have before you change a single tag.

Changing tags blindly is the biggest waste of time with hreflang.

You end up fighting symptoms instead of the cause.

Three sources give you the answer

Each one shows a different layer of the problem:

  • Google Search Console. It used to list “no return tags” in the International Targeting report, but Google has deprecated that report. What’s left is URL Inspection, which shows the canonical Google picked per language version.
  • A crawl with an SEO audit platform. A crawler shows, URL by URL, which hreflang set is on a page and whether its counterparts point back. You’ll find this neatly structured in a proper SEO audit.
  • The page source. Open the <head> of two language versions side by side and compare. It’s slow, but it makes the problem tangible.

Google’s help page is explicit that the report is gone. It also says Google “will continue to support and use hreflang tags on your pages.”

So the tags still matter.

You just can’t lean on a ready-made error list in Search Console anymore.

A crawler does the heavy lifting

A desktop crawler such as Screaming Frog walks through every URL and lists the hreflang annotations it finds:

Screaming Frog SEO Spider product page, the crawler used to find hreflang errors such as missing return tags across all language versions, free up to 500 URLs Screaming Frog’s product page, October 2026. The free version crawls up to 500 URLs; a licence lifts the limit.

For a small B2B site with two or three languages, those 500 free URLs usually cover the whole cluster.

Then check what Google picked

URL Inspection tells you which canonical Google chose for each version. That’s where hreflang and canonical conflicts show up:

If the Google-selected canonical is a different language version, you’ve likely got the hidden fourth error below.

Quick checks per page

A fast first check per URL is our free hreflang checker. It also shows whether a translation tool like Weglot is running.

If your tags come from Weglot and you want to leave it, first read exporting your Weglot translations before you leave.

Moving off a plugin changes every URL in the cluster, so our guide to leaving a translation tool without losing SEO is worth a read too.

Rule of thumb: know the error type first, then repair. Never the other way round.

Error 1: why is the missing return tag the most common hreflang error?

Because two pages have to confirm each other, and it only takes one forgotten template to break that.

The rule is simple: if page A points to page B, B has to point back to A.

Google’s localized versions documentation is blunt: “If page X links to page Y, page Y must link back to page X.”

If the two pages don’t both point to each other, the annotations may be ignored. So it’s not just the one faulty line that stops counting.

A typical scenario

Your Dutch page is set up neatly and points to the French version. But a different team member built the French page from another template.

That page lacks the reference back. Google doesn’t see the relationship as mutually confirmed and does nothing with it.

Here’s what it looks like when it’s wrong. On example.be/nl/diensten/ you have:

<link rel="alternate" hreflang="nl-be" href="https://example.be/nl/diensten/" />
<link rel="alternate" hreflang="fr-be" href="https://example.be/fr/services/" />

But on example.be/fr/services/ there is only:

<link rel="alternate" hreflang="fr-be" href="https://example.be/fr/services/" />

The French page doesn’t point back to the Dutch one.

The fix: one identical set on both pages

Both pages have to carry the full, identical set:

<link rel="alternate" hreflang="nl-be" href="https://example.be/nl/diensten/" />
<link rel="alternate" hreflang="fr-be" href="https://example.be/fr/services/" />
<link rel="alternate" hreflang="x-default" href="https://example.be/" />

Don’t forget the self-reference

Here’s the detail many people overlook: the URL you’re on belongs in the set too.

That’s the self-reference, and Google requires it: “Each language version must list itself as well as all other language versions.”

So the nl-be page has to contain an hreflang to nl-be itself. If it’s missing, the set is incomplete and the return tag complaints come right back.

Two other causes of a broken return tag

Both are about the URL not matching exactly:

  • Notation differences: one tag writes https://, its counterpart http://, or one drops the trailing slash.
  • Redirected targets: the tag points to a URL that still passes through a redirect.

Google also wants fully qualified URLs, “including the transport method (http/https)”. So relative paths like /fr/services/ don’t count.

Watch out: hreflang always points to the final, indexable URL. Never to a version behind a redirect or a canonical.

Not sure whether a target redirects? Run it through our redirect checker, and read up on redirect chains if you find more than one hop.

Error 2: what goes wrong with a wrong or duplicated x-default?

Two things: several pages claim to be the default, or the default points to a dead or empty page.

The x-default is the version Google shows when no language or region matches. Google calls it the value “used when no other language/region matches the user’s browser setting.”

For the Belgian market, with Dutch speakers, French speakers and plenty of international visitors, that’s a useful safety net. But it’s also a source of two specific errors.

Mistake 1: more than one default

Your nl-be page carries an x-default pointing to itself. Your fr-be page carries an x-default pointing to itself too.

Now Google has two conflicting defaults and doesn’t know which one to show.

The rule: exactly one x-default destination per hreflang cluster, and every variant points to that same URL.

Mistake 2: a default that isn’t a real page

Sometimes the x-default sits on an old homepage that now redirects. Or on a language-picker page with no hreflang set of its own.

So choose your safety net deliberately:

  • A neutral landing page that lets the visitor pick a language
  • One of your main languages, usually English for an international audience

Whatever you pick, that destination has to be a live, indexable URL.

What a working setup looks like

This set sits, identically, on every variant:

<link rel="alternate" hreflang="nl-be" href="https://example.be/nl/diensten/" />
<link rel="alternate" hreflang="fr-be" href="https://example.be/fr/services/" />
<link rel="alternate" hreflang="x-default" href="https://example.be/" />

One x-default, one destination, the same on every page.

No page claims itself as the default while another one does the same.

Error 3: why does the many-to-many set break with three or more versions?

Because the number of tags grows much faster than the number of languages.

With two language versions, hreflang is still manageable. It gets treacherous once you add a third, for example an international English version next to nl-be and fr-be.

The maths behind the cluster

Every page has to point to all versions, including itself. Add an x-default and the count looks like this:

HREFLANG MATH Tags grow faster than languages 2 versions3 lines on each page6 tags 3 versions4 lines on each page12 tags 4 versions5 lines on each page20 tags 5 versions6 lines on each page30 tags Calculated: each page lists every version, itself and one x-default.

With three versions, that’s four hreflang lines on each page. Those sets have to match each other perfectly.

With five versions, you’re maintaining 30 tags across the cluster.

One forgotten line breaks the symmetry.

Where it goes wrong in practice

Someone adds a third language and sets up the new page neatly. But they forget to update the two existing pages.

The new English page points to nl-be and fr-be. Yet nl-be and fr-be still only point to each other, not to English.

The cluster is asymmetrical and falls apart.

The correct set for three versions

This should sit identically on each of the three pages:

<link rel="alternate" hreflang="nl-be" href="https://example.be/nl/diensten/" />
<link rel="alternate" hreflang="fr-be" href="https://example.be/fr/services/" />
<link rel="alternate" hreflang="en" href="https://example.be/en/services/" />
<link rel="alternate" hreflang="x-default" href="https://example.be/" />

Check the codes too. Google only supports language codes from ISO 639-1 and region codes from ISO 3166-1 Alpha 2.

And “you can’t specify the country code by itself.” So be alone is wrong; nl-be or nl is right.

What our own cluster looks like

Our experience: our own site runs in four languages, so we use exactly this pattern. Every page we checked in a cluster carries the same five lines.

Here’s the set we found on all three pages we checked on 20 October 2026 (Dutch, French and English):

<link rel="alternate" hreflang="nl-be" href="https://www.customerimpact.be/kennis/hreflang-tags-uitleg/">
<link rel="alternate" hreflang="fr-be" href="https://www.customerimpact.be/fr/blog/hreflang/">
<link rel="alternate" hreflang="en" href="https://www.customerimpact.be/en/blog/hreflang/">
<link rel="alternate" hreflang="de" href="https://www.customerimpact.be/de/blog/hreflang-tags-erklaert/">
<link rel="alternate" hreflang="x-default" href="https://www.customerimpact.be/en/blog/hreflang/">

Each version lists itself, all others and one x-default. The x-default points to English for every page in the cluster.

Treat the cluster as one whole

Change it in one go, not page by page. Add or remove a language, and you update every page in the cluster.

If you don’t, you create exactly the asymmetry that triggers return tag errors.

For larger sites, that’s an argument for managing hreflang in your XML sitemap rather than per page. Google confirms you “can use an XML sitemap to tell Google all of the language and region variants for each URL.”

You then maintain all relationships in one central place.

How to set that up within a scalable URL structure is covered in our guide on international SEO. The choice between folders and subdomains is in subdomain vs subdirectory.

The hidden fourth error: what happens when hreflang and canonical fight each other?

Google may ignore both signals.

Sometimes your hreflang tags are perfect and it still goes wrong.

Then the problem sits with the canonical.

Two signals that contradict each other

A canonical says “index this version”. Hreflang says “this is the variant for this audience”.

Say your nl-be page has a canonical pointing to the fr-be version. You’re telling Google both “this is a separate language version” and “this page doesn’t count, take the French one”.

If they contradict each other, confusion wins.

What Google’s canonical guide says

Google’s canonicalization guide says it directly: when you use hreflang, “make sure to specify a canonical page in the same language.”

The same guide warns against shortcuts: “Don’t use the robots.txt file for canonicalization purposes.”

The rule of thumb

Every language version gets a self-referencing canonical, plus the full hreflang set:

  • one canonical per page
  • pointing to itself
  • never to another language version

Our explainer on the canonical tag covers the basics. If you already have a mess, our guide to fixing canonical conflicts walks through the repair.

This consistency is a fixed checkpoint in our technical SEO approach. It’s exactly the kind of error that stays invisible until rankings collapse.

How do you repair hreflang errors in four steps?

Confirm, repair the whole cluster, check the URLs, then recheck after the next crawl. Hreflang is one layer of a wider international SEO audit.

Here’s the order we follow, and the decision for each error type:

ErrorWhat you seeThe fix
Missing return tagB doesn’t list ASame full set on every page
Missing self-referencePage doesn’t list itselfAdd its own hreflang line
Two x-defaultsEach page claims the defaultOne shared x-default URL
Broken many-to-manyNew language not on old pagesUpdate the whole cluster
Canonical conflictCanonical points to another languageSelf-referencing canonical

Step 1: Confirm the error

Use URL Inspection or run a crawl. Know which of the four errors you have before you touch anything.

Step 2: Repair the whole cluster at once

Update every variant with the same, complete set. Half fixes on one page solve nothing.

Step 3: Check URL notation and redirects

Every hreflang has to point to a final, indexable URL. Keep protocol and slash notation consistent.

Step 4: Recheck after the next crawl

Hreflang changes don’t take effect immediately. Google has to recrawl all versions before the relationship is confirmed.

Patience is part of the process. The Search Console indexing report shows whether the versions get indexed in the meantime.

Make it routine: add this check to your technical SEO checklist, so a new page never breaks a cluster unnoticed.

Frequently asked questions about hreflang errors

Why does Search Console say “no return tags” when my tags are right there?

Because the counterpart doesn’t point back, or points back to a slightly different URL (a difference in protocol, www or trailing slash). Google compares the exact URL. One character off and the return tag doesn’t count. Google has since retired the International Targeting report that showed this message, so a crawler is now where you’ll usually see it.

How many x-default tags am I allowed to have?

One destination per hreflang cluster. All variants point to that same x-default URL. Two pages each claiming themselves as the default is an error.

Should hreflang point to the canonical URL or to the actual page?

To the final, indexable URL, which is ideally also your self-referencing canonical. Never let hreflang point to a redirected or non-canonical version.

How long does it take for an hreflang fix to take effect?

Until Google has recrawled all the pages involved. That can take days to weeks, depending on your crawl frequency. A fix that goes live today won’t be validated tonight.

Does hreflang work outside Google too?

Google and Yandex document support for it. Yandex’s help says: “For each page version, specify links to all its alternative versions.” Other engines lean more on signals like the content-language tag, so if you target them, check their own documentation.

Does Google use hreflang to detect the language of a page?

No. Google says it “doesn’t use hreflang or the HTML lang attribute to detect the language of a page.” Hreflang only tells Google which versions belong together and for which audience.

Can’t get your hreflang right?

Then look at the cluster, not the page.

Hreflang is where Belgian B2B sites most often go wrong quietly. The tags look fine, but the right language version stays hard to find.

We’re a small team that looks at it with you. We’ll tell you whether it’s a quick fix or a symptom of a deeper structural problem.

For us, SEO is the acquisition layer of your growth engine, steered on pipeline, not on technical completeness.

See how we work as an SEO specialist, then send us one URL per language and we’ll check the whole cluster.

Contact us

Free website scan

Enter your website and get an automatic scan within minutes, with concrete technical and SEO improvements. No sales pitch.

Where should we send your report?

We only use your details for your scan. No spam, unsubscribe anytime.