Weighing up AI Translation (discourse-ai) on a multilingual community, also for the SEO upside. With content_localization_enabled + content_localization_crawler_param on, each translation becomes a crawlable ?tl=<locale> URL, so in theory, more searchable content per topic.
For anyone who’s actually turned it on:
Did organic search traffic move, and how long before you saw it?
Did Google index the ?tl= variants cleanly, or any duplicate content issues appeared?
Worth the translation spend?
Any caviats?
Even quick “worth it / don’t bother” welcome. Thanks!
Hey there, great questions. Good thing I wrote a blog post about the experience - please have a read. I believe it would answer a lot of your questions
Another thing, there’s a small footgun I am going to fix this week - if you’re turning on the feature, besides enabling the settings listed above, please also enable set locale from param.
As a webmaster who decided to translate about 140,000 posts and 14,000 topics into several languages, I can tell you that from an SEO perspective, I still don’t have a concrete answer, although I have noticed that my main traffic has lost some percentage share to the countries in question.
I made some SEO mistakes while fixing some issues with the WAF, so since I started translating, I haven’t had a clear perception beyond what Google Search shows me.
Recently, I allowed AI search bots to crawl my instance, and apart from an interaction or two that @Lilly had in my community, international members have been few so far, as I focused on providing translations rather than creating multilingual content.
On the other hand, I think translating content with LLMs is an interesting step, as Twitter and Reddit have shown us that it is useful, of course, with the awareness of allowing users to have their choices respected when reading and writing content. For example, I still prefer that my content remains 100% written in Portuguese and then translated for reading, even though real-time translation is possible and at a low cost.
