Why translated pages are not enough
A business in New Jersey may have English service pages, Korean reviews, Spanish social posts, and a directory listing someone else created years ago. Search engines and AI tools assemble one picture of the business from all of those sources. When the name is spelled three ways, the hours differ between the website and the profile, and the services are listed differently in each language, the picture is fragmented and the answer a customer receives is unreliable.
Entity consistency means the same verified facts appear everywhere, written naturally for each language. Translation is one step of that work, not the whole of it.
One business, one fact sheet
Start by writing a single fact sheet: the name as it appears on the storefront and legal documents, any other-script spelling customers use, the address, phone, hours and holiday exceptions, the services actually offered, the service area, the languages spoken at the counter and on the phone, and how people book or order. Every language version of every page and profile is derived from this sheet.
When a fact changes, update the sheet first and then every derived page and profile in one pass. Most multilingual inconsistencies come from updating one language and forgetting the others.
Give each language its own URL
Google recommends separate URLs for each language version rather than switching content with cookies or browser settings, and advises against automatically redirecting visitors from one language version to another. Google also determines a page's language from its visible content, not from code attributes or the URL, so a page that mixes two languages is classified unpredictably.
Keep a visible link between language versions so a person can switch by choice.
Use hreflang so the versions are recognized as one page
Each language version must list itself and every other version in its hreflang annotations. If two pages do not both point to each other, the tags are ignored. Use x-default for the fallback version. Google treats localized versions as duplicates only when the main content stays untranslated, so shared navigation and templates are fine as long as the main content is written for that language.
Do not promise a language you cannot deliver
A Korean page implies Korean service. If the phone is answered only in English, or Korean-speaking staff are present on certain days, say so on the page. A precise statement of which languages are available in person, by phone, and in documents is more useful to a customer and to an AI system than a page that quietly implies full bilingual service.
Business profile rules for bilingual names
Google Business Profile requires the name to reflect the real-world name and does not permit repeating the name in multiple scripts or languages in the name field, even when the storefront sign shows both. All locations of the same business in one country must use the same name, and there should be one profile per business. Put the other-language name in the business description and on the website instead, and keep categories to what the business is, not what it has.
Reviews and mentions in each language
Customers who write reviews in Korean or Spanish describe different details than those who write in English. Reply in the reviewer's language, do not translate or repost reviews as new ones, and check that community directories and local media in each language carry the same address, phone, and hours as the fact sheet. Inconsistent third-party listings are a common reason an AI answer describes a business incorrectly.
How to check whether it worked
Ask the same customer question in each language, record the tool, date, answer, and cited sources, and compare which facts each language version gets right. Fix factual mismatches before adding new content, then repeat on a later date with the same questions. This does not guarantee that an AI tool will mention the business, and results differ by language and by tool, but it shows whether the facts a customer receives are now consistent.
Related pages
Sources
The technical statements above were checked against these Google documents on September 18, 2026. They describe Google's systems and should not be generalized to every AI provider.