How does GEO work in Dubai and the UAE?

The UAE did not drift into AI adoption; it legislated it. The country appointed the world's first Minister of State for Artificial Intelligence in 2017, runs a National AI Strategy 2031, and in May 2025 became the first country in the world to enable ChatGPT nationwide under the Stargate UAE partnership with OpenAI. When the state treats an assistant as infrastructure, the buyer on the other side of an AI answer is not an early adopter — they are the default customer.

The population that types those prompts is unlike any other market in this series: roughly 89% of UAE residents are expatriates, with the Indian community the largest single group at about a third of the population, and Dubai itself is over 90% expat. English is the lingua franca of commerce; Arabic is the official language and the language of government, Gulf-audience consumer content and local trust signals; Hindi, Urdu and Russian each carry buying communities big enough to move a category.

In practice two tracks decide most funnels. English owns B2B, due diligence, relocation and comparison prompts. Arabic owns the government-adjacent, Gulf-consumer and legitimacy layer — the sources an assistant leans on when the question touches regulation, licensing or local standing. The diaspora languages sit on top as high-intent niches: a brand that is citable in Russian for Dubai real estate reaches buyers the English track never sees.

Arabic is right-to-left, like Hebrew, and morphologically rich in the same way that hurts English-built retrieval: articles, prepositions and pronouns attach to the word, so one written token packs several units, and a page phrased one way can fail to match the same question phrased another. Arabic adds a second layer Hebrew does not have — diglossia. Users prompt in Gulf dialect; almost all written content is Modern Standard Arabic. The assistant has to bridge that gap, and self-contained passages that state the brand and the claim in plain, repeated forms survive the bridge far better than fluent marketing prose.

The Arabic web is thin in a way you can put a number on: Arabic is the content language of about 0.6% of websites — for a language with hundreds of millions of speakers — while English covers roughly half the web. The region has answered with Arabic-centric models like Jais, built in Abu Dhabi precisely because English-centric models underserve the language. For GEO the asymmetry cuts the same way it does in Hebrew: corroboration is harder to accumulate, and a single well-structured Arabic source can own an answer slot that would take dozens of pages to win in English.

Commercial prompts in Dubai carry the city name as a qualifier — "in Dubai", "في دبي" — and split by language along intent lines. English carries company formation, banking, property investment and relocation; Arabic carries consumer services and anything close to regulation; Russian and Hindi carry their communities' purchase decisions end to end.

Agree on a fixed set of relevant questions for each language and repeat them in ChatGPT, Perplexity and Gemini on a schedule. Report share of voice separately by language so progress in one does not hide a problem in another.

UAE brands live in two scripts, and often under two names: the Latin brand and the Arabic rendering, plus a registered trade-licence name that frequently differs from both. Models must learn all of them as one entity — otherwise citations split between half-known names and none accumulates authority. That means declaring the Arabic and Latin forms together in structured data, keeping profiles consistent, and making sure the licence name is connected to the brand name on pages models actually read.

Corroboration in this market runs through a recognizable set of high-trust surfaces: national business media — Gulf News, Khaleej Times, The National, Arabian Business, Zawya — government portals like u.ae and the emirate-level economy departments, and the directories of the free zones themselves. Because the Arabic pool is small, a handful of consistent independent descriptions moves entity recognition more than volumes of links would in English.

We produce English, Russian and Hindi content as part of our work in nine languages, which also includes right-to-left text. For Arabic, we work with a native-speaking partner and handle page structure, consistent brand information and measurement. Facts must agree across all versions so assistants receive a consistent account of the business.

Our published method starts with a fixed question set for each language. We repeat tests in clean, logged-out sessions across ChatGPT, Perplexity and Gemini, then report citation rate and share of voice against the starting baseline. We also check that the English and Arabic pages describe the business consistently.