A common surprise in AI visibility audits is how uneven results are across languages. A company that appears in English answers for its category may be missing entirely when the same question is asked in German, Japanese or Spanish. For brands that sell in several markets, this gap matters: buyers ask assistants in their own language, and the answers they receive draw on different sources.
Why the gap exists
Answer engines retrieve and summarise sources that match the language and context of the question. When someone asks in German, the assistant tends to favour German-language pages, German publications and directories used in German-speaking markets. If your brand has strong English coverage but little in German, the assistant has less to work with, even if the underlying model knows your name.
Translation alone does not always close the gap. A translated page that nobody links to, that is not indexed properly, or that answers questions local buyers do not ask, carries little weight.
Step one: measure per language
Build a small prompt set for each target language, written by a native speaker in the way buyers actually ask. Keep the prompts comparable across languages but not literal translations. Run them monthly, record whether the brand is mentioned, whether a brand URL is cited and whether the description is accurate. A simple table per language shows where the gaps are.
Step two: check the technical basics
For each language version of the site, confirm that pages are indexed, that hreflang annotations connect the versions, that each page is self-canonical and that the main content is present in the server-rendered HTML. Many multilingual sites lose visibility because language versions are rendered only in the browser, or because canonical tags point every language to the English page.
Step three: build local sources
Assistants rely heavily on third-party sources. In each market, identify the directories, review platforms, industry associations and publications that appear in answers for your category. Make sure your company has accurate profiles there, in the local language. Contributed articles and interviews in local publications are slow to earn but valuable.
Step four: localise the pages that matter
Prioritise the pages that already earn visibility in English: comparisons, pricing explanations, glossaries, calculators and detailed guides. Rewrite them for the local market with local examples, currencies and regulations, keeping the facts and structure that make them citable.
Step five: keep entity facts consistent
Company names, descriptions and key facts should be identical in meaning across languages. If your German description says you are a software company and your English one says you are a consultancy, assistants may treat them as different organisations or blend them incorrectly.
How long it takes
In our experience, technical fixes show results within weeks, while third-party coverage in a new language takes several months to build. Track progress per language rather than as one global score, so that strong English performance does not hide weak local performance.
Summary
Multilingual AI visibility is not a translation project; it is a local presence project. Measure each language separately, fix the technical foundations, build local sources and localise the pages that already work. Our PandaWM GEO services page describes how we run this process for companies going global.
References
- Google Search Central documentation on localised versions and hreflang.
- Public documentation from major answer engines on how sources are retrieved and cited.
Published by PandaWM, an overseas marketing team based in Hong Kong.