Multilingual and international GEO

If you serve customers in more than one language or country, the dependable GEO work is the same work Google documents for international search: one language per page, a separate URL for each version, real localization, hreflang annotations that link the versions together, and local business details for each market. GEO (generative engine optimization) is the work of getting your business described accurately and cited in AI answers.

There is a gap you should know about first. As of October 2026, we found no AI vendor documentation of a way for a site owner to target a language or a country in AI answers. Google publishes lists of where AI Overviews and AI Mode are available. The OpenAI and Anthropic help pages we read say only that an assistant may use the user's approximate location. We found no documented "AI hreflang."

What vendors document about language in AI answers

Start with what exists, so you can judge any claim you hear.

"Not documented" does not mean "doesn't matter." It means nobody outside these companies can tell you how it works. The practical step is to follow Google's international guidance. It is the only detailed, published guidance we found, and Google says a page must be indexed and eligible to be shown in Search with a snippet to be a supporting link in its AI features, so a language version that Search handles badly starts at a disadvantage there. For background on those features, see Google AI Mode explained.

One language per page, and a URL for each version

Google's guide to managing multi-regional and multilingual sites defines the terms. A multilingual website "offers content in more than one language." A multi-regional website "explicitly targets users in different countries." You can be one, the other, or both.

Its core rules:

• Separate URLs. "Google recommends using different URLs for each language version of a page rather than using cookies or browser settings to adjust the content language on the page."
• One language per page. Google says you help it determine the language "by using a single language for content and navigation on each page, and by avoiding side-by-side translations."
• No automatic redirects by language. "Avoid automatically redirecting users from one language version of a site to a different language version of a site." Offer links to the other versions and let the visitor choose.
• No content swapped by IP address. "Don't use IP analysis to adapt your content."

The reason behind the last two is worth understanding. Google's page on locale-adaptive pages says the default IP addresses of Googlebot "appear to be based in the USA" and that the crawler sends requests "without setting Accept-Language in the request header." A page that shows German only to visitors from Germany may never show its German text to Google. We found no documentation of whether AI crawlers behave the same way, which is one more reason to give every version its own fixed URL.

Choosing a URL structure

Google lists four options and their trade-offs.

Two notes. Google treats some country-code endings, including .ai, .co and .io, as generic, so they don't target a country. And Google's SEO Starter Guide says the domain ending "only matters if you're targeting a specific country's users, and even then it's usually a low impact signal."

hreflang: what it does and the codes that work

hreflang is an annotation that tells Google which pages are versions of the same content for different languages or regions. Google's page on localized versions says it helps Search "point users to the most appropriate version of your page by language or region."

It is not a language label. Google says: "Google doesn't use hreflang or the HTML lang attribute to detect the language of a page; instead, we use algorithms to determine the language." A lang attribute will not fix a page whose visible text is in the wrong language.

You can add hreflang in the page HTML, in HTTP headers or in a sitemap. Google says the three methods are equivalent and there is no benefit in using more than one. The rules are the same for all three:

• "Each language version must list itself as well as all other language versions."
• Links must go both ways. "If two pages don't both point to each other, the tags will be ignored."
• URLs must be fully qualified, including http or https.

The codes trip people up. The first part is a language code in ISO 639-1 format. The optional second part is a region code in ISO 3166-1 Alpha 2 format.

hreflang is easy to break during a rebuild. If you are changing URLs, read our guide to protecting AI visibility during a redesign or migration first.

Translated and machine-translated content

Google's position is narrower than most summaries of it. As of October 2026, its spam policies contain no standalone rule about machine translation. Translating appears once, as an example of an automated transformation inside the scaled content abuse policy: scraping content to generate many pages "where little value is provided to users." That policy applies to mass-produced pages made to manipulate rankings "no matter how it's created."

So the documented test is value to users, not the method. It would be wrong to say Google penalizes machine translation, and just as wrong to say Google approves of it.

Two more statements from Google are relevant:

• On AI output in general: "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." A translation produced by an AI tool is AI output.
• On partial translation: translating only the boilerplate text while keeping the main content in one language "can create a bad user experience." Google also says localized versions are only considered duplicates "if the main content of the page remains untranslated."

A workable standard for a business:

1. Translate the pages that matter to that market, not the whole site by default.
2. Have a fluent speaker who knows your field review every page before it goes live.
3. Localize, don't only translate: prices and currency, units, phone formats, legal terms, examples and customer questions.
4. Translate titles, descriptions and structured data values too.

Our article on the role of human editors in GEO content covers the review step. A mistranslated fact is still a wrong fact, and any system that reads the page can repeat it.

Local business details for each market

Google names the signals it uses to work out which locale a page targets: country-code domains, hreflang statements, server location (which it calls "not a definitive signal"), and other signals such as "local addresses and phone numbers on the pages, the use of local language and currency, links from other local sites, or signals from your Business Profile."

Most of that list is ordinary business information. For each market you serve:

• Show a local address and phone number on that market's pages, if you have them. Don't invent an office.
• Use the local currency and local spelling.
• Keep the business name identical across languages unless you trade under a different name there.
• Claim a Google Business Profile for each real location.
• In Organization markup, Google says "You can provide multiple addresses if you have a location in multiple cities, states, or countries," and asks for the country code in phone numbers.

Google also says it "ignores locational meta tags (like geo.position or distribution) or geotargeting HTML attributes," so skip those. Our guides to GEO for multi-location businesses and mapping entities across your website go deeper on keeping details consistent.

The same language in different countries

An English site for the US, the UK and Australia is the common case. The pages are close to identical, which raises a duplicate question. Google's answer: if you provide similar or duplicate content on different URLs in the same language, "pick a preferred version and use the rel="canonical" element and hreflang tags to make sure that the correct language or regional URL is served to searchers." If the only difference between your versions is spelling, one version is easier to keep accurate.

Crawler access on every version

International setups multiply the places where a crawler rule can go wrong.

• Each subdomain and each country domain needs its own robots.txt. Google's documentation says a robots.txt file "applies only to paths within the protocol, host, and port where it is posted," and Anthropic asks site owners to set opt-outs "for every subdomain that you wish to opt out from." Check every host.
• Google says that on locale-adaptive sites, "robots meta tags and the robots.txt file must specify the same rules in each locale."
• A regional CDN or firewall rule that blocks traffic by country can block crawlers that come from elsewhere. Google's crawlers mostly originate from the US.

Measuring by market

Search Console's Generative AI performance report has a Countries tab, so you can see where your impressions in AI Overviews and AI Mode come from. Our guide to reading that report explains what it counts.

For other assistants, test by hand. Write a short set of customer questions in each language, as a native speaker would phrase them, and have someone in that market run them. Location can change answers, since OpenAI and Anthropic both say their assistants may use the user's approximate location. Our guide to building a prompt set shows how to record results.

Conclusion

International GEO rests on documented international SEO: a URL per language version, visible content in one language, correct hreflang codes with return links, reviewed translations and real local details. We found nothing published on what AI assistants do with language and location beyond that, so be wary of anyone who claims to control it. If you would like a review of your site's structure and visibility, request a free GEO audit. More guides are in our GEO and AEO Guide.

Frequently asked questions