Almost every New Zealand business that wants Chinese-speaking customers starts the same way: install a translate widget, add a small flag to the header, done. It costs nothing and it is finished by lunchtime.
Then nothing happens. No Chinese enquiries. No traffic from Chinese searches. And the handful of Chinese speakers who do land on the page leave quickly. Not because they cannot read it. Because of how it reads.
What a machine translation actually produces
A translation engine preserves meaning and discards rhythm. English sentences are built long, with the point at the front and qualifiers trailing behind. Chinese marketing copy is built short, in balanced pairs, with the point landing at the end. Translate one into the other word-for-word and the meaning survives while the voice does not.
The result is a page a Chinese reader can parse but does not trust. It reads like a form letter from a bank. That instinct, “this was not written for me”, is the whole problem, and no amount of translation accuracy fixes it.
Meaning is the easy half. Voice is the half that decides whether anyone calls you.
The part search engines see
A widget translates in the browser, after the page has already loaded. What the crawler fetches is the English source. There is no Chinese URL to index, no Chinese title, no Chinese description in the results page. As far as Google is concerned your Chinese site does not exist, because in the HTML it does not.
A real bilingual site does the opposite. Each language lives at its own address. Each address carries its own title, its own description, its own written copy. The two point at each other with hreflang tags so search engines know they are a pair rather than duplicates competing with each other.
What a properly built bilingual site has
- Separate indexable URLs per language, not a client-side switch
- Titles, descriptions and headings written in each language, not translated
- hreflang pairs plus x-default so the two versions do not compete
- Forms, error messages and confirmation emails in the same language as the page
- Chinese-language search terms researched separately. They are not translations of the English ones.
The words are not the same words
This is the part that surprises people most. The Chinese term a customer types is rarely the translation of the English one. It carries a different scope, sometimes a different intent, and it competes against a different set of pages. Translating your English keyword list gives you a Chinese keyword list nobody searches.
There is a second layer under that. Chinese-speaking buyers in New Zealand often do not start on Google at all. They start in WeChat groups, on 小红书, in a comment under someone else's video. A perfect Chinese page still needs someone to have reason to link to it.
When a widget is the right call
It has one legitimate use: a support page, a policy page, a document nobody searches for but somebody occasionally needs to read. Rough comprehension is the entire requirement there, and paying to write it twice would be waste.
For anything meant to win a customer (a service page, a product page, the page an ad points at), write it twice. If that sounds like double the work, it is worth knowing that the second version usually turns out shorter and blunter than the first, because it is written by someone who knows which two objections actually matter.