According to these chaps here, for example, the reason why the likes of Google and various other imitators big and small use mostly GeoIP techniques for choosing the UI language, even overriding the mechanism provided by HTTP for this purpose, is that “people don't know how to change that setting”.

Can anybody please point me to:

  1. Quantitative research that shows that a majority of browser users do not know how to change language settings?
  2. Quantitative research that shows that, in a majority of cases, GeoIP based language negotiation is a better predictor than HTTP-based negotiation. This includes the possible use of techniques other than Accept-Language, notably scanning the User-Agent string for cues as to the language of the browser interface and whether language preferences have been set to other than the default (e.g., multiple languages set in Accept-Language is a strong indicator that it has been explicitly set, as is having a language different than that deduced from User-Agent)?
  3. Qualitative research showing that the implications of getting it wrong with these GeoIP based methods are not worse than getting it wrong via HTTP negotiation? A typical example is when different linguistic communities live in the same State or share the same IP block, such as users based in Switzerland, Finland, Norway (bokmål vs nynorsk), etc. In some cases, this may provoke a fairly hostile reaction from the user, e.g., a Flemish-speaking user in Belgium, or a Catalan-speaking user in Spain.

The above questions deal with the initial choice of language for a new visitor. It does not concern the many other situations when one must deal with language negotiation after first contact, such as:

  • Registered users who have specified language preferences in their profile.
  • Return visitors who have explicitly or implicitly made a language choice.
  • Multilingual users (i.e., a majority of us in the EU and anecdotally, in the world) who at any point may switch between various versions of a multilingual site for a variety of reasons.
  • 1
    As a Flemish-speaking Belgian I get most sites in French. I appreciate the question and want to know an answer as well. My personal preference is OS-language-based and then by language choice, but that is not backed by any facts to justify a full answer. Apr 15, 2016 at 9:24
  • 4
    This is a great topic, but you've included multiple big questions that are likely to result in either 1 few useful responses or 2 an unwieldy discussion of opinions. Can you reshape the question to focus on the experience of language selection and a single question about the right solution in a way that is less about implementation and more UX? Apr 15, 2016 at 15:59
  • 1
    Regarding “people don't know how to change that setting”, to paraphrase George Carlin: "Think about how dumb the average user is, now realize half of em are dumber than that!" Also, this is a self-perpetuating problem because the more content providers attempt to relieve users of having to make a choice, the more resistance users will express to thinking and learning for themselves in the future. The website has no means of inferring user intent; there's no "Language-I-actually-asked-for-myself" HTTP header, so they opt to think for the user and provide their own language selector. +1 to you.
    – Patrick M
    Apr 19, 2016 at 18:32
  • 1
    "showing that the implications of getting it wrong with these GeoIP based methods are not worse than getting it wrong via HTTP negotiation" No research, but if I go to an event in Finland, I don't want to suddenly start seeing everything in Finnish when I browse the web! And that's to not even mention Tor and friends. It seems perhaps reasonable to fall back to GeoIP-based selection if the client does not indicate a preference, but if it does, then please respect that choice if at all possible! (And serve some kind of language selection page if for whatever reason that's not possible.)
    – user
    Apr 19, 2016 at 21:28
  • This article suggests asking the user to select their preferred language and then remembering their choice (and flags up the problems you mention with using IP address etc) smashingmagazine.com/2014/12/… Apr 25, 2016 at 12:28

1 Answer 1


Quantitative and Qualitative research will be hard to find. The Norman Nielsen Group do have an article on some aspects of what you are looking for, and also contains some guidance: https://www.nngroup.com/articles/international-web-usability/

The best advice seems to be, if you can detect the users language, then use this to default the language selector field. In this way there is a good chance the user will not need to change it, however, if it is not correct, the user can still change it to something more suitable.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.