DistilledLive London | A few thoughts on hreflang

The Seattle office took the reins in the last DistilledLive video as Mike Pantoliano and Geoff Kenyon talked you through where to focus your e-commerce efforts in the coming months, as well as sending kudos to those brands already doing it well.

For the latest edition, we thought we’d take the opportunity to get a bit more technical - particularly following the news of our latest DistilledU module release; an interactive guide to meta information. So, without further ado, Distilled London consultant’s, Dave Sottimano and Tom Anthony get stuck into one of the key elements to consider when it comes to International SEO; the hreflang attribute.

Origins of Hreflang

Originally designed to avoid sending out duplicate content signals with content targeted at specific geo-locations, rel=”alternate’ hreflang=”X” is particularly useful mark-up for anyone running a site in multiple languages. However, the appropriate way to implement Hreflang is all circumstances is less than clear...

How the canonical tag can interfere with Hreflang settings and search results

When searching for the localised page, you’d expect to see the native language search page with a native URL but we’re often faced with the English result and the native URL. Why did Google recommend the canonical tag alongside this if we’re seeing this kind of interference?

In fact, Peter Handley delivers a great study on this through this sitemap implementation experiment detailing similar characteristics and issues. So, is there a reason to use the canonical tag? What problems can arise and how can you run tests to work out what’s best for your site and content?

As always, we’d love to throw this topic out for debate and will, as Dave mentions, be looking towards Pierre Far and John Mueller for more feedback. Get involved in the debate over at the Distilled blog or send a tweet along to us over on Twitter and let us know your thoughts.

Read on for a transcript of the above video: 

Dave: Howdy SEOmoz fans. I’m Rand. No, just kidding. I’m Dave. I’m from Distilled. This is Tom. He’s also from Distilled.

Tom: Hey. We’re going to be talking about hreflang today. So first stop I’m going to give a really quick introduction of the theory. So it’s very simple. If you’ve got an established page in the English language - for example, here’s my site- this would be my search results for selling widgets. OK, so now I want to launch a German site. Forgive my bad German here! I want to benefit from the link juice, the Google juice, that I’ve already established by getting links to my English site. Well, I want to serve the German language version to German-native speakers. Hreflang is basically a method that allows me to do that.

However, we found that it interacts funny sometimes with rel=canonical tags. We’re going to be talking specifically about that in this video. Dave?

Dave: All right. This goes out to Pierre Farr. Thanks for helping me out earlier but I’ve got a lot more questions for you after this video. So, we’re talking about the origins of hreflang, and, of course, it started off with an implementation that consisted with the canonical tag. Back on Google Webmaster blog, they said, “Okay, you can do this thing, use a canonical tag and everything will be happy.” Actually, it isn’t, and recently Google actually crossed that out from the documentation. Now they also say that you can still use it. However, you need to be careful of the effects, and you really need to understand what it’s doing. They’re absolutely right.

Let’s go with an example here. I’m going to try to go quickly, but hopefully you can follow along. We’ve got an English US version page, and we’ve got an English UK version page. You can see here I’m using isocodes. So, EN English. US is obviously the United States. Sorry. And GB . . .

Tom: He’s Canadian.

Dave: I’m Canadian. Sorry America. GB obviously UK. So what we’re doing here is we’re going to use the hreflang tag like Tom said. So you just signal. It’s like, oh we’ve got another alternate version of this page. It’s a duplicate copy, but this one is the UK version. Great. If we were going to do the canonical tag implementation, like Google said back in the day, we run into some funny stuff. So what  we’re using the canonical tag for is to make sure that we’re consolidating the link juice. What we’re going to do here is we’re going to canonicalize the UK page to the American page. Now the behaviour is not what you would expect.

Here we are in Google.co.uk, and we’re searching for our keyword, which is “widgets.” Let’s just pretend for a second that nobody else is ranking for widgets. It’s an isolated experiment. So what happens with the search snippet? Because we’ve canonicalized the UK version to the American version, this page no longer appears in Google’s index at all. The only time that it appears is when we’re looking for it specifically in the country and the language of choice. Like I said, we’re in the UK here, Google.co.uk. We searched for widgets and we get our result.

What’s going to happen is we’re going to see the American title. We’re going to see the American meta description or the American content, except we’re going to see the UK URL. Hreflang is causing the display URL to become the UK version. It only fires the UK URL at search time. Typically, you would see the American version, but because we used hreflang, that’s why we’re getting the UK URL.

Tom: On the UK site of Google.

Dave: Of course. So we have to be in the UK to make sure that we see this. So, what happens? It’s not that bad. In this case, both these pages are identical. So they’re going to have the same title and widgets, which is great. The only thing that changes is that we’re bringing the user directly to the British page, sorry the UK page. That’s actually great. Where you run into problems is if this is no longer English and this becomes a French page. So we have FR-FR. And, of course, widgets is not going to be the same translated in French. Of course, the content on the page is not going to be French.

So now we know that if we canonicalize the French version to the English version, the French version is now lost. It’s no longer in the index. What happens if we were to go to Google.fr, there’s not much of a difference except for we see the American title. We see the American description. Instead of seeing obviously the UK page, we see the French URL. That’s not really a good result at all, because widgets is not going to be the same translated in French. Of course, that doesn’t appeal to a French searcher.

Tom: Yeah. So the equivalent is that we’d expecting to see, if it was French or German, the native language result like this. But what we’re actually seeing is the English language result, but with a native URL, and that’s really a poor result for users.

But the question becomes sometimes we need to use the canonical tag. Google gave it to us a few years ago, and it’s become an established part of SEO. So what do we do, Dave, when you need to use the canonical tag, but we’re seeing this sort of result, this sort of interference between hreflang and canonical? How do we handle that?

Dave: Pierre Farr watching this, can you please step in on the comments? I don’t know. Hreflang, in my opinion, was designed to eliminate duplicate content and to consolidate link authority across different pages. It was the one answer. The thing that confuses me is the first implementation they recommended using the canonical tag, which makes me think that hreflang was only a signal to indicate there is another version of this page in a different country in a different language. So why would we need to use the canonical tag as well?

So, in my mind, I think that okay Google recommended canonical tag because it doesn’t consolidate the page authority. Now they’ve taken that recommendation away. So we’re stuck wondering if hreflang actually does consolidate or not. So Peter Handley actually did a good experiment. He was one of the first people to publicly talk about the sitemap implementation, and he found that he had the same characteristics I did in some of my tests.

So remember this, right? This is what happened with the canonical tag. It inherited some of the traits of the canonical target. Funnily enough, Pete’s implementation with sitemap had similar effects. So that makes me think that the hreflang tag actually inherited some of the canonical powers. Do we know for sure? No. Is there a reason to use the canonical tag?Yes, you can still do it. But here’s a golden rule. Never, ever, ever do it across languages, or you get things like this happening. So you get English content with a French URL, and that’s not good for your searchers.

Tom: I think, at the end of the day, the main message we’re going to give away here is going to be that you got to test this. We run tests like crazy on other sites, but when you’ve got your own site or a client’s site, it is always hard to test. But you can take a small portion. So if this is product pages or category pages or whatever it is, you can take a small percentage of those, run your own tests, wee what your results are, and work out what the best implementation is. Make sure you give enough time between tests so that you know they’re interfering with one another. From there make an informed decision yourself.

I’m sorry, Pierre and John Mueller and all the guys at Google, but normally our advice is to take Google’s advice with a little pinch of salt and test it for yourselves. Chris Semturs is one of the other guys working at Google, who said himself, that they’re not finding that it behaves consistently at the moment. So even when Google saying that, then our advice is pinch of salt.

Dave: And Pierre Farr, we’re looking at you to provide some more answers.Yes, we’re pointing at you and John Mueller too. Thanks.

Tom: Thank you very much.

Cheri Percy

Cheri Percy

Cheri joined Distilled as a community intern and now heads up the Marketing department in the London office. She has co-ordinated and project managed some of Distilled's biggest content pieces to date and has doubled its social media growth.  When...   read more

Get blog posts via email

18 Comments

  1. Very interesting find, thank you guys!

    I couldn't figure out why I kept getting the same results for my website, no matter what language I typed my query in, but I think you just solved the mystery for me.

    I'm in this unfortunate (but, I believe, quite common) situation where the content on my website is available in two languages: English, being the default one, and French, which is accessible through /fr/.

    After reading quite a lot about hreflang a couple weeks ago, I decided to add it to my sitemap and head tags. Since then, I've been patiently waiting to see how it would affect my search results. Needless to say, I was quite disappointed and perplexed to see that the results always came back in English, even though I typed my query in French.

    Now, I never really cared that much about SEO in general. I really only implemented hreflang because I thought it would help my visitors get relevant results for their searches, but I hardly think seeing results in English for a French search helps them in any way.

    At this point, one might wonder: why would Google treat hreflang tags as such? Wouldn't it make more sense to use the French meta title and description for searches that has French keywords? Most certainly, so why the strange behaviour?

    Maybe Google feels it would be unfair to give the translated content the same ranking as the original (English) content? Maybe that's what's happening? Perhaps the French content on my website is treated as a completely separate entity, with its own ranking? Food for thought.

    Either way, I'm still debating whether I should remove the tags or leave them up. I can't make up my mind about it...

    Thanks again.

    reply >
    • David Sottimano

      Hey Simon - don't give up on hreflang okay? The guys at Google are working hard to correct any issues but we need to help them out. There's a few quirks here and there, but I think I might be able to help you out. I'll personally take a look if you want to share? (david.sottimano@distilled.net)

  2. Thank you Tom and Dave for providing more insight on hreflang - will be putting this into action very soon.

    Also, I liked the previous Distilled "jingle" better... just sayin' :)

    reply >
  3. Great resource here guys. It seem that Google has acknowledged using the canonical in conjunction with the hreflang attribute doesn't result in the best experience when using the page tag and or the XML Sitemap implementation. I'm sure you're looking at different authority metrics to determine how juice is passed using the hreflang attribute. I'm looking forward to a follow-up post!

    reply >
    • David Sottimano

      I don't expect hreflang alone to pass any juice to it's alternate versions, and it would be exceptionally hard to convince any business to allow me to test without internally linking these pages into the architecture. But, it would be very interesting to see what would happen ;)

      I could do the test in isolation, but that would be pretty useless.

  4. Watched the video multiple time and i didn't understand why you canonicalized the French version to the English version, that by it self its a wrong implementation .
    I Can see implementing rel canonical for a English website between EN-US EN-GB and EN-CA
    on the other hand, lets say that the website is originally French so implementing canonical to FR-CA FR-CH FR-FR is right.

    and imo google removed the canonical mention because they saw that there is a confusion with the webmasters, where they thought that they HAVE to implement a canonical to the english version for everytime they use hreflang

    reply >
    • David Sottimano

      The reason why I used that example was to illustrate the WRONG implementation. So, you are right, this is the wrong implementation when using the canonical tag, but in the beggining, many webmasters were confused about this implementation, and I wanted to explain why it was wrong.

      Webmasters, and even myself (in the early days) thought that we could consolidate link equity through the use of rel canonical and that referencing the different language versions with hreflang would cause the right page in the right language to appear, without losing any link juice.

      Google removed the canonical reference because too many people were causing more harm than good. If hreflang consolidated, there wouldn't have been any use for implementation with the canonical tag.

  5. The joys of international SEO!

    From my understanding, with the French example, you would search for an English keyword, say 'laptop's, in Google.fr and be shown the English title and description with the French URL. What happens when you search for the French translation 'portatif' in Google.fr? Do you still get the canonical English title and description or the French version?

    reply >
  6. Martijn Hoving

    Hi Tom, Dave,

    Why would you canonicalize en-gb to en-us? I thought the hreflang tag itself was there to signal Google that there is a same page in another language for another market. And that the hreflang tag works well with a canonical tag seems rational to me. Yet not a canonical tag across countries, but a canonical tag that stays on the ccTLD, subdomain or subfolder of a country.
    Any thoughts?

    reply >
    • David Sottimano

      If the en-gb and en-us were identical (or very similar), you'd want to use the canonical tag to consolidate the link authority.

  7. Modesto

    In a private chat i had with Maile Ohye (Google's Developer Programs Tech Lead) at SES London she was adamant that hreflang anotations do not consolidate link equity.

    reply >
    • David Sottimano

      Had a chat with John Mueller in Italy, and he basically said the same thing. Hreflang was / is intended to get the right page to the right country / lang search.

  8. David Sottimano

    Hi all, my apologies for not getting to these comments earlier and thanks for getting involved in the conversation.

    reply >
  9. Kevin

    So we have FR-FR.


    That means you have a French page for the French market (FR-FR). If the content is in English, you should use the EN ISO for the language... as the content is in English. And then use the FR as second param to point out that it is for France. So in my opinion, it becomes en-fr. English for in France.

    reply >
  10. Sorry for the late comment, been doing tons of research on hreflang and rel-canonical and found a source on Moz recommending to self-reference each rel=canonical page once hreflang is implemented, and that using rel-canonical across sites at all is a mistake. Any thoughts on this?

    So every page rel=canonical points to itself, maybe this solves the over-riding title & met-d's in the search results..

    http://moz.com/blog/rel-confused-answers-to-your-rel-canonical-questions

    The comment is by gfiorelli1, he's a MOz associate and left the 1st comment.

    reply >
    • David Sottimano

      Hi Chris,

      Again, straight from Google engineers: Hreflang was not designed to consolidate link authority. Rel=canonical is, but should be used very carefully in conjunction with hreflang. Where hreflang annotations exists for EXACT duplicate content in the SAME language (English for example), you can point the rel=canonical to one version (EN-US for example). The page you canonicalize (EN-AU for example), will inherhit the content (title, meta description and page content) from the EN-US page. The ONLY thing that changes is the URL at search time in Australia. Otherwise, yes, you can self reference canonical tags without any problems. Canonicals are avoided because of the complexity of implementation, not for the fact that it causes damage.

      Hope that helps.

  11. reply >

Leave a Reply

Your email address will not be published. Required fields are marked *

*
*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>