The real A/B test is: custom translation versus automatic in-browser translation. Huge percentage of people use the latter now. Edit: I haven't done careful ana

The real A/B test is: custom translation versus automatic in-browser translation. Huge percentage of people use the latter now. Edit: I haven't done careful analysis, but I run a large site with global appeal and 50%/50% US versus non-US traffic. Everyone is using auto-translate, I can detect it easily in the logs. Engagement metrics are near-identical. Sometimes even better, for example in some Asian countries where auto-translation is being heavily used. Regular people use the browser auto-translate now and are quite comfortable with it. Even bigger question: will you be harming the auto-translators if you don't make your site available in English? Automatic translators are quite decent at the hundreds of English<->Rare-Languages combinations, but much worse at the thousands of Rare-Language<->Rare-Language combinations. It's all due to the available training sets, and where the most machine learning investment has been made, and that simply this size relationship: combinations(number_of_languages,taken_2_at_a_time) = i.e. all pairs. is much greater than: 1 * number_of_languages = i.e. English paired with everything else. That relation is the power of a lingua franca, at least until translators vastly improve for all the obscure language pairings.