What the research says about language and citations
Arabic answers draw on a different set of sources. Marta Warren ran 20 London hotel questions aimed at Gulf travellers, in Arabic and English, more than 1,400 times across ChatGPT, Gemini, Copilot, Google AI Overviews and AI Mode (PRmoment, 26 August 2026). The hotel shortlists were 79% similar. The evidence behind them was not: 28% of non-platform sources appeared in both languages. Arabic answers leaned on regional travel platforms such as Almatar (157 citations) and Almosafer (140), and on Instagram (94).
The Arabic web is small. W3Techs reported on 28 October 2026 that Arabic is the content language of 0.6% of websites whose language it can detect, against 49.5% for English, 1.6% for Turkish and 1.0% for Persian.
| Finding | Source | What it means for your content |
|---|---|---|
| 28% of non-platform sources shared between Arabic and English | PRmoment, August 2026 (hotels) | Your English coverage does not stand in for Arabic coverage. |
| Arabic is 0.6% of websites by content language | W3Techs, 28 October 2026 | Fewer Arabic pages compete for each Arabic question. |
Arabic, English or both: a decision table
The table below is Zaher's judgement, based on the research above and on client work. It is a starting point to test against your own measurements.
| Your situation | Where to start | Why |
|---|---|---|
| Most customers ask in Arabic (for example, mass-market retail in Egypt) | Arabic-first, with English kept for key pages | Arabic answers lean on Arabic sources, and the buyer's own words are Arabic. |
| Mixed Arabic and English audience (common in the Gulf) | Both, with the same facts and depth on every commercial page | Each language builds its own source pool, so a gap in one is not covered by the other. |
| Globally oriented sector such as hotels or travel | Both, and a presence on the regional platforms Arabic answers cite | In the PRmoment hotel study, Arabic answers leaned on regional travel platforms while sharing few sources with English answers. |
| B2B or international buyers | English depth, plus Arabic pages for local trust and procurement | Buyers research in English, but local decision-makers and listings often work in Arabic. |
| Regulated categories such as health or finance | Both, with identical regulated facts and wording reviewed in each language | A wrong price, dose or fee in one language is repeated by assistants that cite it. |
Two languages on one site: what Google asks for
Google's guidance on multilingual sites (Search Central, last updated 10 December 2025) applies directly, because a page must be indexed and eligible for a snippet to appear as a link in AI Overviews or AI Mode (Google, "AI features and your website", same date):
- One URL per language version, rather than switching language with cookies or browser settings. Google notes that its crawler usually comes from the US and sends no Accept-Language header, so versions reached by automatic switching alone may never be crawled.
- hreflang annotations so Google links each searcher to the right language version.
- One language per page. Google decides a page's language from its visible content, not from the lang attribute or the URL, and advises against side-by-side translations.
- No forced redirects by language. Let people switch language themselves.
Writing Arabic that assistants can quote
These are Zaher's working practices from client delivery, not published research.
- Write natively, do not translate. Start from the questions Arabic-speaking customers ask, then write the page that answers them. A translated English page answers English questions in Arabic words.
- Match the register your customers use. Write product and policy pages in clear, natural Arabic that matches how the store's own customers speak. Keep regulated statements, such as warnings and policies, precise and consistent whatever the register.
- Track dialect phrasing in your questions. Research on Arabic dialects (TARJAMAT, Kadaoui and colleagues, 2023) found that language models struggle more with dialects that have little public data, so measure the phrasing people actually type in Cairo, Riyadh or Dubai.
- Keep one set of facts. Prices, branch counts, years in business and policies should be identical in Arabic and English. When the two disagree, an assistant has two versions to choose from.
- Spell your name the same way everywhere. The PRmoment study had to consolidate hotel names that appeared in different spellings and scripts before rankings could be trusted. Decide your Arabic and Latin spellings and use them on every page and profile.
- Earn Arabic third-party mentions. Arabic answers cite Arabic marketplaces, directories, media and social pages. Find which ones are cited in your category and make sure they describe you accurately.
What this looked like for an Egyptian pharmacy chain
Al Fouad Pharmacies switched its online store to Arabic-first on 1 May 2026. Zaher built a bilingual master library of 1,498 products in eight fields: title, description, why choose it, when to use it, key ingredients, how to use it, warnings and precautions, and frequently asked questions. The 267 products that make up 80% of revenue were rewritten in full.
In the client's own GA4, AI-referred revenue grew 11.9 times between two consecutive 90-day windows (22 December 2025 to 21 March 2026, and 23 March to 20 June 2026). For a like-for-like read on a later window: chatgpt.com, the one AI source tagged the same way in both windows: revenue +667%, 5 May to 2 Aug 2026 vs the previous 90 days (GA4). Part of the growth reflects AI referrals being labelled more clearly (GA4 added a built-in AI Assistant channel on 13 May 2026, inside the second headline window), the results come from the period of the work rather than a controlled experiment, and the language switch was one of several changes made in that time. See the case study for the full sequence.
How to measure your own Arabic and English split
- Write the same buying questions natively in each language, including dialect phrasing for Arabic.
- Run each question several times on each assistant you care about, in each language.
- Compare three things per language: how often you are named, which sources are cited, and whether the facts are right.
- Start with the weaker language, then re-run the same questions to see what changed.
Zaher's GEO Analysis compares results by language and market, and Zaher Qalam, its AI content writer, produces Arabic and English pages from your own facts. For the vendor questions to ask before you buy any of this, read how to choose an AI visibility provider.
Zaher is the Attention Layer for Arabic-first brands: it measures and improves what AI assistants say about a brand, in Arabic and in English. Read how AI assistants choose brands to recommend in Arabic, or see our research.