hreflang without pain: implementation for multilingual sites

A client of ours ran 14 languages and 9 markets with organic traffic growing 3% a year while the catalog doubled. The audit surfaced 211k URLs with broken hreflang: missing return tags, wrong region codes (en-UK instead of en-GB), and x-default pointing to a 302. Sixty days of cleanup later, traffic in secondary markets jumped 38%. Hreflang isn't glamorous, but it's where global teams quietly bleed revenue. Before walking through the patterns, remember How to audit on-page SEO without falling into guesswork: without a baseline from logs and Search Console, you're optimizing blind.

The core rule is symmetry. If page A points to B with hreflang='es-MX', B must point back at A. Google silently drops one-way declarations, and Screaming Frog (SEO Spider 20+) ships a 'Hreflang > Missing Return Links' report that catches 80% of issues in a single crawl. A detail few teams nail: the attribute takes ISO 639-1 codes for language and ISO 3166-1 Alpha 2 for region. en-UK doesn't exist. pt-BR does. It trips people up because the .uk TLD is common, but hreflang follows its own standard.

Where do you declare it? Three valid options per Google: link tag in the HTML head, HTTP header for PDFs and non-HTML assets, or XML sitemap. For sites past 50k translated URLs, sitemap is the only sustainable option. Cramming 40 link tags into every page head bloats HTML weight and tanks LCP, something we already covered in Core Web Vitals: beyond LCP, what actually moves the needle. Use a sitemap segmented by market with consistent lastmod. If you still generate sitemaps by hand, read Modern XML sitemaps: priority, lastmod, and what to skip before continuing this conversation.

x-default is the most misunderstood piece. It is not 'the English version'. It's the page for users whose region or language matches none of the declared alternates. In global ecommerce, x-default usually points at a country selector or a generic international version (US or UK English, USD prices). I've seen x-default land on a URL that runs IP-based 302 redirects, which causes Google to shuffle SERP results unpredictably. If you run international PLPs, E-commerce on-page: PLP vs PDP without cannibalization covers how to avoid parallel cannibalization.

Tools I use in the monthly audit: Screaming Frog for the crawl, Ahrefs Site Audit to surface conflicts with canonical (canonical and hreflang pointing at different pages is a guaranteed bug, see Canonical tags: common mistakes bleeding your organic traffic), and custom BigQuery queries on top of the GSC export to cross country-level impressions with declared hreflang. That third technique is the kind covered in BigQuery + GSC: queries your agency won't run, and it's what separates diagnosis from guessing. On projects with more than 5 languages, build an adjacency graph of declarations and look for disconnected components.

Errors that still kill: subdomains with hreflang pointing at subdirectories (URL architecture mismatch), underscores instead of hyphens (pt_BR is invalid, must be pt-BR), and declaring hreflang on noindex pages. Google drops the noindex side and keeps only the other end, breaking symmetry silently. Another classic is duplicating hreflang across multiple alternates for the same language-region pair, generating 'multiple entries' in GSC. Check the 'International Targeting' report (still works in 2026) weekly for the first 90 days post-rollout.

Practical takeaway: before you add an eleventh language, audit the 10 you already have. Run Screaming Frog with hreflang validation on, export the errors, sort them by impression volume from GSC, and fix what already gets traffic first. Hreflang isn't advanced SEO, it's hygiene. Done well, it pays off in markets you didn't realize were leaving money on the table.