How to Implement Hreflang on a Multi-Region Site Without Sending Conflicting Signals

A compass resting on a paper map on a desk

A company launches a Canadian version of its store, copies the English pages, changes the currency, and waits for Canadian shoppers to find it. Instead, search results in Toronto keep showing the US pages, and the Canadian pages compete with the originals for the same queries. The team adds hreflang tags, and for a few weeks things look worse, because the tags contradict each other in ways nobody noticed.

Hreflang is a small piece of markup with an outsized reputation for going wrong. The idea is simple: tell search engines which version of a page belongs to which language and region. The execution is where sites stumble, because every annotation is a claim that another page has to confirm. This guide walks through how to build those claims so they agree with each other, with your canonical tags, and with what is actually on the page.

compass on a paper map spread across a desk
Photo by Bahram Yaghooti on Pexels

What Hreflang Actually Does and Does Not Do

Hreflang tells a search engine that a set of URLs are alternates of the same content for different audiences. When someone searches in a given language or region, the engine can swap in the version that fits that person instead of showing whichever page happened to rank first. The hreflang attribute is a hint, not a command, and it does not boost rankings on its own.

It also does not translate anything, redirect anyone, or merge ranking signals the way a canonical tag can. Each alternate keeps its own URL and has to earn its own visibility. What hreflang changes is which of your own pages gets shown, which matters most when you have near-identical pages, such as English content for the US, UK, and Australia.

If your site has a single language and a single audience, you do not need any of this. Add hreflang when you have real alternates, and not as a general SEO ritual.

Decide Whether You Have Real Alternates

Before writing a single tag, list the pages that are genuinely equivalent across versions. A product page in German and the same product page in French qualify. A German landing page with no French counterpart does not, and pretending otherwise creates annotations that point nowhere.

Be strict about what counts. Pages should cover the same content, even when the wording, prices, and legal details differ. If the French version is a different article that happens to share a topic, it is not an alternate, and linking the two will confuse the engine rather than help it.

Also decide how you target audiences. Language alone works well when you serve one translation to everyone who reads it. Language plus region makes sense when the content changes by country, for example different pricing, shipping rules, or spelling. Keep that choice consistent across the site, because mixing approaches in one cluster makes the annotations harder to reason about.

Write Language and Region Codes Correctly

Hreflang values follow a specific pattern: a language code, optionally followed by a region code. The language comes from the ISO 639-1 list and the region from ISO 3166-1 alpha-2, which the IETF language tag standard builds on. So en-GB is English for the United Kingdom, and de-AT is German for Austria.

The mistakes here are boringly consistent. People write en-UK because it feels right, but the correct region code for the United Kingdom is GB. They write en_US with an underscore, or put the region first, or use a region code on its own, such as us, which is not valid because the language is required.

There is also a reserved value, x-default, which marks the fallback page for users who match none of your listed languages. A language selector page or your main global page usually fits this role. Use it once per cluster, and never as a stand-in for a specific audience.

library card catalog drawers labeled in rows
Photo by Daniel Forsman on Unsplash

Pick One Method and Stick With It

You can deliver hreflang three ways: as link elements in the page head, as HTTP headers, or inside your XML sitemap. All three are valid, and Google's documentation on localized versions describes each. The trouble starts when a site uses more than one for the same pages and the copies drift apart.

Head tags are the easiest to inspect and suit most sites with a template you control. HTTP headers are the only option for non-HTML files such as PDFs. Sitemaps work well when a large catalog makes per-page tags heavy, because the annotations live in one place and can be generated from your data.

Choose the method that matches how your pages are produced, and generate the annotations from a single source of truth, usually the same table that records which pages are translations of which. Hand-editing tags across hundreds of templates is how clusters end up with a missing link nobody can find.

Make Every Annotation Reciprocal

This is the rule that causes most failures. If page A says page B is its alternate, page B must say the same about page A. Without that return link, search engines may ignore the pair, because an unconfirmed claim is easy to get wrong or abuse.

Each page in a cluster also has to list itself. A cluster of three pages therefore has three annotations on every page: one pointing at itself and one pointing at each of the others. Add x-default and it is four. That scales quickly, which is exactly why generating the markup beats writing it by hand.

Reciprocity breaks in predictable ways. A page gets deleted but its siblings still point to it. A URL changes during a migration and only some pages are updated. A new market launches and the old pages are never told. Build the annotation list from current, live URLs every time you deploy, so a retired page disappears from every cluster at once.

"Hreflang fails quietly. The tags validate, the pages load, and nothing tells you the return link is missing until the wrong country is ranking. Treat reciprocity as something you test on every release." - Dennis Traina, founder of 137Foundry

server rack with organized network cables
Photo by panumas nikhomkhai on Pexels

Keep Canonical Tags and Hreflang in Agreement

Canonical tags and hreflang answer different questions, and they have to be consistent. The canonical says which URL is the preferred version of this exact content. Hreflang says which different URLs serve different audiences. When they disagree, the engine has to pick one, and it often ignores the hreflang.

The classic mistake is pointing the canonical of every regional page at the main version. That tells the engine the regional pages are duplicates to be folded away, which cancels the annotation you worked to add. Each regional page should carry a canonical pointing at itself, and the hreflang cluster should list those same self-referencing URLs.

Check the hreflang URLs against the canonical URLs character for character. A trailing slash, a www mismatch, an http versus https difference, or a tracking parameter is enough to make the two systems disagree. If you want a refresher on how canonicals go wrong, our guide to fixing canonical tag conflicts covers the common patterns, and the same discipline applies here.

Handle Redirects, Noindex, and Blocked Pages

Every URL in a hreflang cluster should return a clean 200 status, be indexable, and be crawlable. An alternate that redirects, carries a noindex directive, or sits behind a robots.txt block is a dead end, and the engine will drop that link from the cluster.

This is easy to break by accident. A regional page gets a temporary redirect to the global page during a campaign, and the cluster silently loses a member. A staging rule leaves noindex on a new market's templates after launch. A robots.txt pattern written for one folder swallows another.

Make the check automatic. After each deploy, request every URL in your annotation list and confirm the status, the robots directives, and the canonical. A short script that fails the build when any alternate is not a plain indexable page is cheaper than weeks of wondering why a market is invisible.

Don't Redirect Users by Language Alone

It is tempting to detect a visitor's location or browser language and send them to the matching version automatically. Automatic redirects cause two kinds of harm. Crawlers usually fetch from one location with default settings, so they may never see the other versions. And real users get trapped on a page they did not choose, with no easy way out.

A better pattern is to show a clear banner suggesting the better-fit version while leaving the current page in place, and to give people a visible language and region switcher. Let the hreflang annotations handle search results, and let the page handle the person already on it.

If you must redirect, for example on the bare root domain, make sure a crawlable page still exists for every version at its own URL, and that the redirect never applies to requests that identify themselves as a search crawler. Hiding content from crawlers that users can see is its own problem.

Match the Page to the Promise

Hreflang makes a claim about the page's language and audience, and the page should back it up. A page tagged fr-CA that is written in European French with euro prices undermines the claim. A page tagged en-GB that still says "color" and quotes dollars does the same.

You do not need to rewrite everything for every region, but the visible signals should fit: the language of the text, the currency, the address and phone formats, the shipping and returns information, and any legal notices. The more a page genuinely serves its stated audience, the better it tends to hold its place for that audience.

Keep an eye on machine translation that leaves untranslated fragments, such as navigation labels or form errors, in the source language. Search engines look at the content actually on the page, and a mismatch there can matter more than the tag.

magnifying glass resting over a printed document
Photo by cottonbro studio on Pexels

Audit the Result With Real Tools

After implementation, verify instead of trusting. Google Search Console reports on how your pages are indexed and which alternates it has found, and a crawler such as Screaming Frog can extract every annotation across the site and flag missing return links, invalid codes, and non-200 targets.

Start with the simplest checks. Do all codes follow the language-region pattern? Does every page list itself? Is every link returned? Do the URLs match the canonicals exactly? Then spot-check a few clusters by hand, fetching each alternate in a browser and reading the page the way a visitor would.

Run the audit on a schedule, not once. Sites change constantly, and hreflang is one of those systems where a single template edit can break hundreds of pages without any visible error. If you want a second pair of eyes on a sprawling multi-region setup, our technical SEO team does this kind of review as part of larger projects, and you can see how we approach it on the services page.

A Short Checklist Before You Ship

Put the essentials in a list your team can run before any international launch. Every page in a cluster lists itself and every sibling. Every annotation has a return link. Language and region codes are valid. Each page has a self-referencing canonical that matches its hreflang URL exactly. Every alternate returns a 200, is indexable, and is crawlable. There is one x-default per cluster, and no automatic redirects based on location.

Keep the annotation source in one place, generate the markup from it, and test the output on every deploy. Do that, and hreflang stops being a fragile pile of tags and becomes a boring, reliable part of your site, which is exactly where it belongs. For related groundwork, see our web development services and the about page for more on how we work.

Need help with Technical SEO?

137Foundry builds custom software, AI integrations, and automation systems for businesses that need real solutions.

Book a Free Consultation View Services