--- title: "Does question language change who AI cites? | Len P. van der Hof" description: Ask an AI search tool the same question in English and in Dutch and it may cite different sources. What research shows, what it does not, and a 15-minute… image: "https://lenvanderhof.com/media/generated/blog-hero-does-language-change-who-gets-cited-v2.bdc8cef1ab60.wide.webp" --- [SEO, GEO and AEO](https://lenvanderhof.com/en/blog/category/seo-geo-aeo/) Research guide # Does the language of your question change who AI cites? An English-only test leaves your Dutch questions unchecked. Compare the cited pages, not just the language of the answer. Len P. van der HofPublished 20 October 20268 min read Same question, different language, different street. Direct answer It can. A 2025 University of Toronto preprint reported near-zero overlap in the domains cited by OpenAI's search-enabled GPT-4o model between English questions and their translations into Chinese, Japanese, German, French or Spanish. The study used developer versions, not the ChatGPT app, and did not test Dutch. Compare the same buyer question in both languages, repeat the runs, and record the answer's language separately from each cited page's language. A difference in your results does not reveal how the engine chooses sources. ## Key takeaways - In a 2025 preprint, GPT-4o's search model had near-zero overlap in cited domains across the tested language comparisons. This does not establish how today's ChatGPT app behaves. - Neither study discussed here tested Dutch. Run your own two-language test instead of assuming. - Change only the language. Keep the buyer's need, the country, and the kind of answer the same. - Log the answer's language and the cited page's language separately. A Dutch answer can cite an English page. - A difference between languages is a reason to check one missing page, not proof that you must translate the whole site. **The language of a question can change which websites an AI answer cites.** The language of the answer and the language of its sources are separate: an answer in Dutch may cite an English page. To check your visibility, read the citations as well as the answer. An English-only test tells you what appeared for your English questions. It does not establish what appears when buyers ask in Dutch. Research gives you a reason to compare the two, but neither study below tested Dutch. An **answer engine**, such as ChatGPT search, Perplexity, or Google’s AI Overviews, writes answers and can link to sources. Those source links are **citations**. In about 15 minutes, you can make a first comparison on one buyer question. ## What does the research show? **The Toronto study** by Chen, Wang, Chen and Koudas is a September 2025 University of Toronto preprint, not a peer-reviewed paper. The researchers translated 100 English questions, ten each on ten topics, into Chinese, Japanese, German, French and Spanish. They queried four AI engines through developer interfaces, plus Google, then compared the domains returned for each English question and its translated version. The [study](https://arxiv.org/abs/2509.08919) reported that: - OpenAI’s search-enabled GPT-4o model had near-zero overlap in cited domains across the tested language comparisons. - GPT and Perplexity drew a larger share of their sources from the question’s language than Google did. - Claude reused a much more stable set of domains across languages and drew more heavily on English-language sources. These are findings about the tested systems and questions, not a rule for every answer engine. **The [“Faux Polyglot” study](https://arxiv.org/abs/2407.05502)** by Sharma, Murray and Xiao, a NAACL 2025 paper, used fictional documents in English, German, Chinese, Hindi and Arabic. This was a controlled experiment in retrieving documents and generating answers, not a test of live search. The tested models favoured documents in the question’s language in both stages. When no document matched that language, they favoured languages with more available text, such as English. ## What does it not show? Neither study tested Dutch. Both are snapshots of changing systems. The Toronto study used developer versions of the engines, not the ChatGPT app. It also compared whole domains, while the check below looks at individual URLs. [OpenAI’s search help page](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt) says ChatGPT search typically rewrites a question into one or more targeted queries that it sends to search partners. It may use your general location, based on your IP address. If Memory is on, it may also use saved memories when it rewrites the question. The page does not describe any effect of the question’s language. A Dutch question might lead to Dutch search queries. That is an inference, not a documented rule, and it does not tell you which pages will be cited. [Google says](https://developers.google.com/search/docs/appearance/ai-features) AI Overviews and AI Mode may issue several related searches across subtopics and data sources. Its [multilingual-site guidance](https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites) says it determines a page’s language from visible content, not the `lang` attribute or URL. So when you log a page’s language, go by the text a reader sees, not by a label in the code or the URL. Neither these pages nor a handful of test answers will tell you how an engine chooses its sources. ## Myths and facts MythWhat the evidence says”The AI translates everything, so language does not matter.”The Toronto study found very little overlap in GPT’s cited domains across the tested language comparisons.”If I am cited in English, I am covered in Dutch.”Neither study tested Dutch. You need to check your Dutch questions separately.”The answer is in Dutch, so it read Dutch sources.”Answer language and source language are separate. Open the cited pages and record their languages.”Translate the whole site and you will be cited.”Neither study shows that translating your site causes citations. Start by checking whether a relevant answer is missing in one language. ## How do you run a fair two-language test? Keep the buyer’s need constant. An English question about “workflow software” and a Dutch question about replacing a shared spreadsheet ask different things. Their answers cannot isolate the effect of language. 1. **Write the need first.** Keep the country and the requested kind of answer the same: both questions should ask for an explanation, for example, or both for a supplier list. 2. **Write a natural pair.** Start in the language you know best. Have a fluent speaker check that the other version expresses the same need in wording a buyer would use. 3. **Check four hidden differences.** Did one version add a budget, ask for local suppliers, name a brand, or specify a feature? Match those constraints before testing. 4. **Keep the setup fixed.** Use the same engine, model where selectable, account, location and Memory setting. Save both exact questions and note those settings beside your log. 5. **Run each version in a fresh chat with web search enabled.** For a first pass, run English, Dutch, Dutch, English. Confirm whether search was actually used. Before acting on a pattern, extend the check to at least three runs per language and repeat in another session. Log one row per cited page, repeating the run number when an answer has several citations. Save the answers too. If a run has no citations, record that explicitly. RunDateQuestion languageWeb search used?Answer languageCited URLLanguage of that pageYour name in the answer?1(date)EnglishYesEnglish(URL)EnglishNo2(date)DutchYesDutch(URL)DutchNo3(date)DutchYesDutch(URL)EnglishYes *These rows are hypothetical, not results from an experiment.* Open the links before counting them. A page about a similarly named company is an [identity problem](https://lenvanderhof.com/en/blog/person-entity-seo/). A name without a supporting link is a [mention, not a citation](https://lenvanderhof.com/en/blog/chatgpt-mentions-vs-citations/). Record mixed-language or inaccessible pages as such; do not guess their language from the URL. Keep **market wording** in a separate test. Adding a cao requirement, the collective labour agreement for a sector, changes the scope of a staff-roster question. That may make the question more relevant to Dutch buyers, but it changes more than language. ## A worked example (hypothetical) Imagine a company selling maintenance scheduling to teams across several Dutch sites. Its stated need is: “a small maintenance team needs clean shift handovers and a visible record of changed assignments, in the Netherlands.” Suppose its English runs keep citing a broad American product guide, while its Dutch runs keep citing a Dutch competitor’s implementation note. This is an invented example; no English-versus-Dutch citation experiment was run for this article. The operator first checks whether the Dutch question added a constraint, such as a cao requirement. If it did, the pair needs correcting and rerunning. If the questions match and the difference repeats, the operator checks one thing: does the company have a useful Dutch page answering that need, with its identity clear? If that answer is missing, one page may be worth writing. The observation does not establish why the competitor was cited or justify translating the whole site. ## How should you read the result? - **Different pages in each language.** Check the question pair, then compare repeats within each language. If English runs also vary widely among themselves, the difference between English and Dutch is harder to interpret. Look for a repeated pattern before choosing a page to improve. - **The same pages in both languages.** This small set showed no difference in cited pages. Keep the result; it does not prove that language never matters. - **A narrow conclusion.** Name the questions, languages, runs and URLs. “These Dutch questions cited different pages in these runs” is defensible. “ChatGPT prefers Dutch websites” is not established by this log. You can observe a difference on your questions without knowing how the engine chose its sources. The [absence diagnosis](https://lenvanderhof.com/en/blog/why-chatgpt-does-not-name-you/) explains why a screenshot of one answer is not a finding; the same caution applies here. ## Try this today (15 minutes) 1. **Define the need (3 minutes).** Pick one question a Dutch customer asks before buying. Write the need, country and requested answer type. 2. **Prepare the pair (3 minutes).** Write English and natural Dutch versions. Check the four hidden differences. If you cannot judge both versions, get a fluent speaker’s check before interpreting the results. 3. **Run the first comparison (6 minutes).** Use ChatGPT with web search enabled and a fresh chat each time: English, Dutch, Dutch, English. Keep the setup fixed. 4. **Inspect and log (3 minutes).** Record each cited URL, the answer’s language and the page’s language. Note shared URLs and differences within each language as well as between them. Treat this as a first pass. If source checks take longer, finish them before drawing a conclusion. Repeat any apparent pattern before making a content decision. If the comparison exposes one missing answer you can support, start with that page, not with translating the site. [How to get cited by ChatGPT](https://lenvanderhof.com/en/blog/how-to-get-cited-2026/) covers what that page needs; this test only helps you decide which page to examine. Cite this:Does the language of your question change who AI cites?Len P. van der Hof. [https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/](https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/) · Published 20 October 2026. ## Terminology - [CITATION](https://lenvanderhof.com/glossary/citation/) ## Sources 1. [Generative Engine Optimization: How to Dominate AI Search (Chen, Wang, Chen and Koudas, University of Toronto, preprint, 2025)](https://arxiv.org/abs/2509.08919) · arXiv 2. [Faux Polyglot: A Study on Information Disparity in Multilingual Large Language Models (Sharma, Murray and Xiao, NAACL 2025)](https://arxiv.org/abs/2407.05502) · arXiv 3. [Searching the web with ChatGPT](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt) · OpenAI Help Center 4. [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features) · Google Search Central 5. [Managing multi-regional and multilingual sites](https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites) · Google Search Central 6. [Why am I not mentioned in ChatGPT?](https://lenvanderhof.com/en/blog/why-chatgpt-does-not-name-you/) 7. [How to get cited by ChatGPT](https://lenvanderhof.com/en/blog/how-to-get-cited-2026/) ## Further reading - [Why am I not mentioned in ChatGPT?](https://lenvanderhof.com/en/blog/why-chatgpt-does-not-name-you/) - [How to get cited by ChatGPT](https://lenvanderhof.com/en/blog/how-to-get-cited-2026/) - [How answer engines choose sources](https://lenvanderhof.com/en/blog/how-answer-engines-select-sources/) - [ChatGPT mentions vs citations: a name is not a source](https://lenvanderhof.com/en/blog/chatgpt-mentions-vs-citations/) About the author ## [Len P. van der Hof](https://lenvanderhof.com/en/authors/len-p-van-der-hof/) Entrepreneur, AI Innovator and Venture Builder Len P. van der Hof builds practical AI systems, digital ventures and evidence-informed tools for founders. ```json { "@context": "https://schema.org", "@graph": [ { "@type": "Person", "@id": "https://lenvanderhof.com/#person", "name": "Len P. van der Hof", "alternateName": [ "Len van der Hof", "L.P. van der Hof", "Leendert Pieter van der Hof" ], "honorificSuffix": "MSc", "url": "https://lenvanderhof.com/", "image": [ "https://lenvanderhof.com/photos/len-portrait-1.jpg", "https://lenvanderhof.com/photos/len-portrait-2.jpg", "https://lenvanderhof.com/photos/len-portrait-3.jpg", "https://lenvanderhof.com/photos/len-portrait-4.jpg", "https://lenvanderhof.com/photos/len-speaking.jpg", "https://lenvanderhof.com/photos/len-hero.jpg" ], "jobTitle": "Entrepreneur, AI Innovator and Venture Builder", "description": "Len P. van der Hof, MSc, is a Dutch entrepreneur and AI innovator in Zwijndrecht. He builds ReasonKit, MindSesh, Undominated.ai, books under his name, the fiction imprint LPH98.lifestyle, and technology ventures through LPH98.ventures. Eleven titles in Systems for the Strategic Self are available now, in English and Dutch.", "address": { "@type": "PostalAddress", "addressLocality": "Zwijndrecht", "addressCountry": "NL" }, "alumniOf": { "@type": "CollegeOrUniversity", "name": "Rotterdam School of Management, Erasmus University" }, "knowsAbout": [ "Artificial intelligence", "AI agents", "Agentic AI systems", "LLM routing", "SEO", "Generative engine optimization", "Answer engine optimization", "Venture building", "Founder performance", "Founder psychology", "Evidence-based decision-making" ], "sameAs": [ "https://www.linkedin.com/in/lenvanderhof/", "https://x.com/LenvanderHof", "https://www.youtube.com/channel/UCTG20buKqYYbitqqf7l3zJA", "https://www.instagram.com/Lenvanderhof/", "https://www.threads.com/@lenvanderhof", "https://github.com/Lenvanderhof", "https://huggingface.co/LPH98", "https://www.npmjs.com/~lenvanderhof", "https://www.goodreads.com/author/show/70983905.Len_P_van_der_Hof", "https://www.amazon.com/author/lenvanderhof", "https://www.bol.com/nl/nl/b/len-p-van-der-hof-msc/609879394/", "https://bsky.app/profile/lenvanderhof.com", "https://mastodon.social/@Lenvanderhof", "https://crates.io/users/Lenvanderhof", "https://cursor.com/@Lenvanderhof", "https://medium.com/@Lenvanderhof", "https://gitlab.com/Lenvanderhof", "https://hub.docker.com/u/lenvanderhof/", "https://dev.to/lenvanderhof", "https://www.facebook.com/Lenvanderhof", "https://soundcloud.com/Lenvanderhof" ], "affiliation": [ { "@id": "https://lenvanderhof.com/#publisher" }, { "@id": "https://lenvanderhof.com/#mindsesh" }, { "@id": "https://lenvanderhof.com/#lifestyle" } ] }, { "@type": "WebSite", "@id": "https://lenvanderhof.com/#website", "url": "https://lenvanderhof.com/", "name": "Len P. van der Hof", "description": "Len P. van der Hof, MSc, is a Dutch entrepreneur and AI innovator in Zwijndrecht. He builds ReasonKit, MindSesh, Undominated.ai, books under his name, the fiction imprint LPH98.lifestyle, and technology ventures through LPH98.ventures. Eleven titles in Systems for the Strategic Self are available now, in English and Dutch.", "inLanguage": [ "en", "nl" ], "publisher": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "Organization", "@id": "https://lenvanderhof.com/#publisher", "name": "LPH98.ventures", "url": "https://lph98.ventures", "founder": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "Organization", "@id": "https://lenvanderhof.com/#mindsesh", "name": "MindSesh", "url": "https://mindsesh.net", "founder": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "SoftwareApplication", "@id": "https://lenvanderhof.com/#reasonkit", "name": "ReasonKit", "url": "https://reasonkit.sh", "creator": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "SoftwareApplication", "@id": "https://lenvanderhof.com/#undominated", "name": "Undominated.ai", "url": "https://undominated.ai", "creator": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "Organization", "@id": "https://lenvanderhof.com/#lifestyle", "name": "LPH98.lifestyle", "url": "https://lph98.lifestyle", "founder": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "ImageObject", "@id": "https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/#primaryimage", "url": "https://lenvanderhof.com/media/generated/blog-hero-does-language-change-who-gets-cited-v2.bdc8cef1ab60.wide.webp", "contentUrl": "https://lenvanderhof.com/media/generated/blog-hero-does-language-change-who-gets-cited-v2.bdc8cef1ab60.wide.webp", "representativeOfPage": true }, { "@type": "BreadcrumbList", "@id": "https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/#breadcrumb", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://lenvanderhof.com/" }, { "@type": "ListItem", "position": 2, "name": "Blog", "item": "https://lenvanderhof.com/en/blog/" }, { "@type": "ListItem", "position": 3, "name": "SEO, GEO and AEO", "item": "https://lenvanderhof.com/en/blog/category/seo-geo-aeo/" }, { "@type": "ListItem", "position": 4, "name": "Does the language of your question change who AI cites?", "item": "https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/" } ] }, { "@type": "WebPage", "@id": "https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/#webpage", "url": "https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/", "name": "Does the language of your question change who AI cites?", "description": "Ask an AI search tool the same question in English and in Dutch and it may cite different sources. What research shows, what it does not, and a 15-minute test.", "isPartOf": { "@id": "https://lenvanderhof.com/#website" }, "primaryImageOfPage": { "@id": "https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/#primaryimage" }, "breadcrumb": { "@id": "https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/#breadcrumb" }, "inLanguage": "en-GB" }, { "@type": "BlogPosting", "@id": "https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/#article", "mainEntityOfPage": { "@id": "https://lenvanderhof.com/en/blog/does-language-change-who-gets-cited/#webpage" }, "headline": "Does the language of your question change who AI cites?", "description": "Ask an AI search tool the same question in English and in Dutch and it may cite different sources. What research shows, what it does not, and a 15-minute test.", "datePublished": "2026-10-20T19:00:00.000Z", "author": { "@id": "https://lenvanderhof.com/#person" }, "publisher": { "@id": "https://lenvanderhof.com/#person" }, "image": [ "https://lenvanderhof.com/media/generated/blog-hero-does-language-change-who-gets-cited-v2.bdc8cef1ab60.square.webp", "https://lenvanderhof.com/media/generated/blog-hero-does-language-change-who-gets-cited-v2.bdc8cef1ab60.landscape.webp", "https://lenvanderhof.com/media/generated/blog-hero-does-language-change-who-gets-cited-v2.bdc8cef1ab60.wide.webp" ], "articleSection": "SEO, GEO and AEO", "inLanguage": "en-GB" } ] } ```