Hreflang Either Works Completely or Breaks Everything
Most hreflang audits stop at “does the tag exist?” Wrong question. The real damage happens when the tags exist, validate cleanly in every crawler you own, and still silently route the wrong content to the wrong users — or trigger a canonicalization fight that Google resolves in a direction you didn’t intend.
We’ve worked through enough multi-regional setups to say this plainly: hreflang is the only technical SEO signal where a partial implementation is actively worse than no implementation at all.
The Bidirectionality Requirement Is Where Most Sites Fail
Every hreflang annotation must be bidirectional. If your /en-us/ page points to /en-gb/ as an alternate, then /en-gb/ must return a reciprocal tag pointing back to /en-us/. Every URL in the set references every other URL in the set — including itself via a self-referential tag or x-default.
Sounds simple. Breaks constantly. Usually because:
- CMS templates generate hreflang dynamically, and untranslated pages still get partial tag sets from a shared component
- Regional microsites run by different teams diverge — one team updates a URL, the other doesn’t
- Sitemaps and on-page tags disagree, and Google gets conflicting signals from both
When reciprocation breaks, Google doesn’t just ignore the tag. It falls back to canonicalization heuristics, and those heuristics heavily favor the URL with more inbound inbound PageRank. Your US site cannibalizes your UK site not because of content overlap, but because of an annotation error that looked clean in every validator you ran.
x-default Is Not Optional for International Sites
x-default tells Google which page to serve when no regional match exists — someone in Japan hitting a site that only has en-us and en-gb variants. Omitting it doesn’t just leave an edge case unhandled. It leaves Google making a canonicalization decision, not a soft fallback.
The mistake we see most often: pointing x-default at your highest-converting US page because it’s your “best” page. That tells Google to treat it as the global default, which means ranking signals from every unmatched locale consolidate into that single URL. Sounds fine until you realize you’ve undermined every regional variant’s ability to earn independent authority. x-default should point to your actual catch-all — a language selector, or your primary market page — not the URL you’d most like to rank globally.
Where Hreflang and Canonical Actually Conflict
Google’s documented position is that hreflang and canonical should agree. When they don’t, Google treats the canonical as dominant for indexing and hreflang as advisory for serving. That distinction matters more than most people realize.
Concrete example: if /en-gb/pricing carries a canonical pointing to /en-us/pricing, Google will likely index only the US version — but may still attempt to serve the GB version to UK users based on hreflang. You end up with a page that isn’t independently indexed getting served to the audience it was built for. It ranks on borrowed equity and gets no crawl attention of its own. This is a slow-motion ranking problem, which is exactly why it persists for months before anyone notices.
The fix is straightforward: canonicals within a hreflang set should be self-referential. /en-gb/pricing canonicals to itself. /en-us/pricing canonicals to itself. They’re separate documents serving separate markets. Treating them as duplicates defeats the entire point of regional targeting.
Sitemap vs. On-Page Implementation: Pick One, Do It Completely
You can implement hreflang via XML sitemaps, <link rel="alternate"> tags in the <head>, or HTTP headers for non-HTML files. Mixing these without a deliberate reason produces conflicts that are nearly impossible to debug without crawling all three sources simultaneously.
Sitemap-based implementation is common on large sites because it avoids bloating page HTML. The trade-off: Googlebot has to fetch the sitemap to discover the annotations, and if your sitemap isn’t crawled frequently, your hreflang signals lag behind your actual site structure. For e-commerce sites with seasonal regional variants, that lag can mean regional targeting is wrong for weeks before Google reconciles it.
On-page tags are crawled when the page is crawled — the annotation is always current. The cost is payload size, which adds up fast on sites with 20+ regional variants.
My default for most multi-regional setups: on-page tags for the core site, sitemap-based for any content type that doesn’t control its own <head>. The critical part either way is auditing both sources for consistency — not just checking whether one of them validates.
Language vs. Region: The Annotation Semantics Matter
Hreflang uses BCP 47 language tags. en means English in any locale. en-US means English as used in the United States. en-GB means English as used in the United Kingdom. These aren’t interchangeable, and using en when you mean en-US creates a signal mismatch — you’re claiming the page suits any English speaker globally, which competes with your en-GB variant unnecessarily.
The failure mode that follows: Google sees a generic English page and a UK-specific English page. The generic one tends to absorb ranking signals for both markets because it’s treated as the broader match. Your UK variant gets suppressed in UK SERPs by the page you never intended to rank there.
How to Actually Audit This
Standard tag validators — Screaming Frog, Sitebulb — check syntactic correctness and basic reciprocation. They don’t check:
- Whether canonical tags on the same page contradict the hreflang annotation
- Whether URLs referenced in hreflang return 200s (not 301s, not 404s)
- Whether sitemap-based and on-page implementations agree with each other
- Whether
x-defaultpoints to a page that’s actually indexed and not itself canonicalized away
A reliable audit requires a crawl that simultaneously extracts canonicals, response codes, on-page hreflang, and sitemap-declared hreflang — then cross-references all of them per URL. Most teams do this in pieces and miss the cross-source conflicts entirely.
Google Search Console’s International Targeting report catches obvious issues like missing return tags. It does not surface canonical-hreflang conflicts or the x-default misuse patterns above. Treat GSC as a first-pass signal, not a clean bill of health.
One Thing Worth Sitting With
If your multi-regional site has been running hreflang for more than a year without a full cross-source audit — canonicals vs. on-page tags vs. sitemap declarations, all three at once — there’s a reasonable chance you have at least one conflict costing a regional variant its independent ranking potential. The annotations look fine. The GSC report looks clean. And somewhere, your UK or AU or CA variant is borrowing equity from your US page instead of building its own.
Start with the pages where your US and UK variants have near-identical organic traffic curves. That’s usually where the canonicalization bleed is happening.