The fragmentation pattern
The typical overseas Chinese business exists as 4–6 entities the machine can't merge: "Golden Dragon Restaurant" (Google), "Golden Dragon Chinese Restaurant" (Yelp), "Jin Long Restaurant" (an old directory), 金龙酒家 (WeChat), and a Facebook page from 2019 with a different phone number. Each fragment is findable; none is citable with confidence. Engines hedge — "there are several businesses with similar names" — and the customer moves on.
The canonical-name rules
- One English name, one Chinese name. Choose, document, enforce everywhere. If the registered name differs from the trading name, declare both in schema (legalName + alternateName) and pick the trading name for all listings.
- One romanization. Pinyin, consistently ("Jin Long", not "Jinlong" on one site and "Chin Lung" on another). Romanization variance is the classic entity-splitting bug.
- NAP byte-identical. Same address format, same phone format, across Google, Yelp, directories, your site, your Chinese-language profiles. "Suite 200" vs "Ste 200" vs "#200" are three entities to a strict resolver.
The schema pattern
LocalBusiness (or its subtype — Restaurant, Dentist, ProfessionalService) JSON-LD on your site with: canonical name, alternateName in the other language, address, telephone, geo, openingHours, sameAs pointing to every profile in both languages, and inLanguage-honest pages. This is the declaration layer that lets engines merge your fragments into one confident entity.
Linking the two webs
Your English site and your Chinese-language presence should reference each other: the site links the WeChat/Chinese profiles; the Chinese profiles carry the English name and website. sameAs does the machine-readable version of the same job. Engines that can verify "these two presences are one business" will state both correctly; unlinked webs drift into separate, weaker entities.