AI Search in Multiple Languages: What It Means for Your Site
AI search in multiple languages works on top of ordinary web search: Google's AI Overviews and AI Mode draw on pages in Google's index, ChatGPT search rewrites your question into search queries it sends to search providers, and Perplexity searches with its own crawler. None of them publishes a rule for exactly which language's sources it cites. So the practical answer for a business selling in two or more languages is the same as for classic search: have a real, indexable, crawlable page in each language you sell in, written for that market, because an engine can only cite the version that exists and that it can reach.
Below is what each engine actually documents, what that implies for a multilingual site, and a simple way to test how you show up in each language.
How does AI search answer in different languages?
Each AI search product retrieves web pages through a search system first, then writes an answer from them, so its reach in a language depends on what it can find and fetch in that language.
Here is what the providers themselves say, and nothing more.
Google AI Overviews and AI Mode. Google's documentation on AI features and your website says there are no additional requirements to appear in AI Overviews or AI Mode: a page must be indexed and eligible to show in Google Search with a snippet. Both features may use a "query fan-out" technique, issuing multiple related searches across subtopics to build a response, and they may use different models, so the links they show vary.
Google announced in May 2025 that AI Overviews were available in more than 200 countries and territories and more than 40 languages. In September 2025 it added AI Mode in Hindi, Indonesian, Japanese, Korean and Brazilian Portuguese, noting that a global search experience "goes far beyond translation" and needs an understanding of local information.
ChatGPT search. OpenAI's help article on searching the web with ChatGPT explains that ChatGPT search sometimes partners with other search providers and typically rewrites a prompt into one or more targeted queries that it sends to them. It may also use an approximate location derived from the user's IP address to improve results. To be eligible for inclusion, a site must allow the OAI-SearchBot crawler and allow traffic from OpenAI's published searchbot IP addresses.
Perplexity. Perplexity's crawler documentation says PerplexityBot is designed to surface and link websites in Perplexity's search results, and recommends allowing it in robots.txt. A separate agent, Perplexity-User, fetches pages when a user's question requires it.
What none of them documents: whether a question asked in German prefers German sources, falls back to English ones, or translates. Treat any confident claim about that as a hypothesis to test, not a rule.
What each engine documents, side by side
The documented facts differ by engine, but the controls you have are almost the same everywhere: be indexable, be crawlable, and have a page in the language of the question.
| Engine | How it finds pages | Language and location behaviour it documents | What you control |
|---|---|---|---|
| Google AI Overviews | Google Search index; may fan out into several related searches | Available in more than 40 languages (May 2025 announcement) | Indexing, snippet eligibility, hreflang, localized content |
| Google AI Mode | Google Search index; query fan-out; may use different models from AI Overviews | Added Hindi, Indonesian, Japanese, Korean and Brazilian Portuguese in September 2025 | Same as AI Overviews |
| ChatGPT search | Rewrites prompts into queries sent to search partners; OAI-SearchBot | May use approximate location from IP address | Allow OAI-SearchBot and its IP ranges; clear pages in each language |
| Perplexity | PerplexityBot index; Perplexity-User fetches on request | Not documented on its crawler page | Allow PerplexityBot; clear pages in each language |
The common thread is retrieval. An answer engine that runs a search in Spanish can only cite a Spanish page if one exists, is indexed, and is not blocked. Your English page may still be cited for a Spanish question, but you have no documented way to make that happen, and a Spanish reader landing on an English page is a weaker visit.
What this means for a multilingual site
AI search raises the stakes on the same multilingual basics, because a missing or broken language version is now invisible in two places at once: the classic results and the AI answer above them.
Four implications follow.
1. Separate, crawlable URLs per language matter more, not less. Since Google's AI features draw on its index, anything that stops a language version from being indexed also keeps it out of AI Overviews. Automatic redirects by IP, cookie-based language switching and one-way hreflang all reduce what Google can see. The fundamentals are in the international SEO strategy and the hreflang tags guide.
2. Write extractable passages in every language. AI answers quote passages, not whole pages. A localised page with a direct answer under each heading, a comparison table and a short FAQ is easier to lift than a long translated essay. The principles in getting cited in Google AI Overviews apply language by language.
3. Localise, do not just translate. If query fan-out and query rewriting turn a question into the phrases people actually use in that market, a page built on translated home-market keywords may not match those searches. Native keyword research is what aligns your page with the queries an engine is likely to run.
4. Check your crawler rules. Firewall, CDN and security plugin rules can block AI crawlers without you noticing. If OAI-SearchBot or PerplexityBot is blocked, those engines cannot surface your pages in their search features, in any language. Review robots.txt and any firewall or CDN bot rules.
Ranking in classic search and being cited in AI answers overlap but are not identical. The difference between the two disciplines is covered in GEO vs SEO, and the language dimension simply multiplies the work: each market needs its own pages that an engine can retrieve and quote.
How to test AI search in multiple languages
Run the same small set of questions in each language you sell in, record which sources each engine cites, and repeat monthly.
- Pick five to ten buyer questions per market. Use real questions customers ask before buying, phrased the way locals phrase them, not translated from English.
- Ask each engine in the target language. Use Google (checking whether an AI Overview appears), AI Mode where available, ChatGPT with search, and Perplexity. Where you can, set the interface language or region to the market you are testing.
- Record the citations. Note whether your site is cited, which language version appears, and which competitors are cited instead.
- Diagnose gaps. If your English page is cited for a German question, the German version may be missing, unindexed or weak. If nothing of yours appears, check indexing and crawler access first, then content.
- Fix and retest. Improve or create the localised page, confirm it is indexed, then rerun the same questions next month.
Keep the log simple: date, engine, language, question, cited URLs. Over a few months it shows whether your localised pages are gaining ground. Answers vary between runs, so look for patterns across several checks rather than reacting to one result.
Common mistakes with AI search across languages
Most of the errors are classic multilingual SEO mistakes that now cost you twice.
- Blocking AI search crawlers by accident. A blanket bot rule can block OAI-SearchBot or PerplexityBot. Allow the search crawlers you want to appear in.
- Relying on one language to cover all markets. There is no documented guarantee that an English page will be cited for questions in other languages.
- Machine-translating everything for AI visibility. Unedited translations miss local phrasing and add little for readers. Fewer, well-localised pages are safer and more citable.
- Hiding language versions behind redirects. Googlebot mostly crawls from the US, so IP redirects can keep localised pages out of the index and out of AI Overviews.
- Testing only in English. If you only check English answers, you will not see that a competitor owns the German or Spanish answers.
- Treating AI answers as a separate project. The work that earns citations is the same work that earns rankings, done per language. See AEO vs GEO for how the terms relate.
Frequently Asked Questions
Does Google AI Overviews work in other languages?
Yes. Google announced in May 2025 that AI Overviews were available in more than 200 countries and territories and more than 40 languages. Google's documentation says there are no extra requirements to appear: a page must be indexed and eligible to show in Search with a snippet. Pages in each language you target therefore need to be indexable.
Does ChatGPT search in the language I ask in?
OpenAI does not document the exact language behaviour. Its help article says ChatGPT search rewrites a prompt into one or more targeted queries sent to search providers and may use approximate location from your IP address. The practical step is to test your key questions in each language and see which sources are cited.
Can an English page be cited in a Spanish AI answer?
It can happen, but no provider documents when or how often, so you cannot rely on it. A Spanish reader sent to an English page is also less likely to convert. If Spanish-speaking customers matter to you, publish a localised Spanish page that is indexed and crawlable, connected to the English version with hreflang.
Do I need hreflang for AI search?
Hreflang is a Google Search signal for choosing which language version to show, and Google's AI features draw on its search index, so correct hreflang supports the right version being available. Other AI engines do not document using hreflang. The safest approach is to implement hreflang correctly and keep each language version independently strong and crawlable.
Should I block AI crawlers on translated pages?
Only if you do not want those pages to appear in that engine's answers. OpenAI says sites must allow OAI-SearchBot to be eligible for ChatGPT search results, and Perplexity recommends allowing PerplexityBot to appear in its results. Blocking these search crawlers removes your pages from those features in every language, not just one.
Making AI search in multiple languages work for you
AI search in multiple languages rewards the same thing classic multilingual SEO does: a real, localised, indexable page for each market, with extractable answers and open crawler access. Be sceptical of anyone who claims to know exactly how each engine picks a language; test it with your own questions and fix the gaps you find.
The hard part for a small team is producing that localised, citable content in every language, month after month. Agent Solo's Growth plan publishes in two languages with hreflang, canonicals and per-market keywords, the Scale plan supports up to five, and every plan includes AEO and GEO optimisation. If you want your second market to appear in AI answers too, join early access.