GEO for UAE and Dubai Brands: What Changes When Your Customers Ask AI in English and Arabic

Four of the five most recent prompts we saw asking about GEO were about the UAE — Dubai agencies, Abu Dhabi clinics, a Sharjah logistics firm. We had zero regional content, so this is the first piece. It is written the way we write everything now: what is documented, what we have measured, and where we have not measured and say so rather than invent a number.

Which engines dominate in the UAE

We have not measured this yet. The market-share figures in circulation — for ChatGPT, Gemini, Perplexity, Copilot — are global or US-panel numbers, mostly built from trackable referral clicks. Those undercount Google AI Overviews entirely (they emit no referral) and depend on which platforms stamp their outbound links. We will not repeat a global share as a UAE share. What we can say without a study:

• Google Search is the default on Android and in Chrome, and AI Overviews and AI Mode are served in the UAE in both English and Arabic. Whatever the UAE mix is, Google's generative surfaces are in it.
• ChatGPT search, Perplexity and Claude all serve Arabic prompts. Whether a UAE customer's question was answered with retrieval on or off is not visible from the answer unless citations appear — that caveat is global, not regional.
• A UAE engine census is on our research list. It needs audit data across many UAE brands, which is what a fixing tool generates as a by-product. When we have it, it will be published with its sample size.

Bilingual prompt coverage

This is the part that is genuinely different, and it is mechanical. OpenAI documents that ChatGPT search "typically rewrites your query into one or more targeted queries" before sending them to search providers. It does not document the language of the rewritten queries. The safe assumption is that an Arabic prompt produces Arabic search queries, and an Arabic search query retrieves Arabic pages.

The consequence: if your site is English-only, there may be nothing for an Arabic prompt to retrieve. Not "ranked lower" — absent from the candidate list at the retrieval stage. That is the one stage where topical relevance is the factor three independent studies agree on, and a page in the wrong language has no topical relevance to the query.

What to check, in order:

1. Do your key facts exist in Arabic at all? Pricing, services, locations, licence numbers. Not the whole blog — the pages a customer would ask about.
2. Do the Arabic and English pages agree? A price stated in AED on the Arabic page and USD on the English page is not a translation, it is a second fact. Pick one currency per fact, or state both on both.
3. Is the Arabic page server-rendered? Many bilingual sites load the second language through a client-side switcher. Non-Google AI crawlers do not execute JavaScript, so a switcher-loaded Arabic page does not exist to them.
4. Do hreflang and canonical tags point where they should? Google documents these for its own index. No AI engine documents reading them; treat them as Google hygiene, not an AI lever.

We audit both languages by running the same prompt set in English and Arabic and reporting the two columns separately. Averaging them would hide exactly the gap you need to see.

.ae vs .com: entity signals

Google documents a country-code top-level domain as a strong geotargeting signal for its own search index (Search Central, managing multi-regional sites). That is the extent of the first-party evidence. No AI engine — OpenAI, Anthropic, Perplexity — documents treating a .ae domain differently from a .com. So:

• For Google surfaces, .ae plus consistent Arabic content is a documented geotargeting signal. A .com with proper hreflang and a UAE-targeted Search Console setting reaches the same place.
• For the other engines, we have no evidence either way. What every engine does read is your visible text. A brand whose About page says "based in Dubai, licensed by DED, serving the UAE and GCC" in plain server-rendered text on both language versions has stated its entity. A TLD has not.
• Do not run two sites. A .ae and a .com with slightly different facts is a contradiction with a domain name on each side. If you have both, one should redirect.

Local directories AI actually cites

We have not run the UAE citation census yet, so we will not list directories as "cited by AI" on the strength of a feeling. What we can say with confidence:

• Google Business Profile feeds Google's own surfaces, including AI Overviews for local queries. This is the one listing where the mechanism is documented by the operator. Keep the name, address, phone and hours identical to your website in both languages.
• Official registries — the Department of Economy and Tourism licence record in Dubai, the equivalent in each emirate — are the sources a model has no reason to distrust. Make sure your trade name on the licence is the name on your website.
• For everything else — property portals, food-delivery platforms, review sites — we will publish which ones appear in citations once we have the data, with the sample size next to the claim. Until then, treat any list you read elsewhere as unverified.

The Cloudflare change on 2026-09-15

This is dated, documented and imminent. From 2026-09-15, Cloudflare sets the Training and Agent crawler categories to blocked by default on pages that display ads — for all new domains, for new sites added to existing accounts, and for all existing free-tier customers. The Search category stays allowed by default. Cloudflare decides which bot sits in which category; check the AI Crawl Control panel for your zone rather than assuming.

Why UAE sites are exposed: a large share of small and mid-sized business sites everywhere sit on Cloudflare's free tier, and we have no reason to think the UAE is different. We have not measured the UAE share. What we can say is that a site on the free tier that displays ads will have its AI access changed on the 15th without its owner editing anything, and nothing on a dashboard will announce it.

What to do this week:

1. Log in to Cloudflare and open AI Crawl Control. Note which category each AI bot is in and what the setting will be after the 15th.
2. Decide deliberately. Blocking training bots (GPTBot, ClaudeBot, Google-Extended) is a legitimate choice. Blocking retrieval bots (OAI-SearchBot, PerplexityBot, Claude-SearchBot) removes you from those answers.
3. Request one of your pages as each retrieval bot and confirm a 200. OpenAI states eligibility requires both a robots.txt allow and that your host or CDN allows its published searchbot IPs. A 403 from the CDN fails the second half silently.
4. Run the same check on the Arabic pages. A rule scoped to a path or subdomain can pass one language and block the other.

What we can measure today, and what we cannot

Question	Status
What ChatGPT, Perplexity, Google AI and Claude say about your brand in English and Arabic	Measured per audit, retrieval state disclosed
Whether your Arabic pages exist in raw HTML and are reachable by each retrieval bot	Measured per audit
Where your English and Arabic pages contradict each other	Measured per audit
Which engines UAE customers use most	Not measured. No UAE-specific first-party data exists.
Which UAE directories AI cites	Not measured. Census pending.
Your rank or position inside an AI answer	Not measurable by anyone. Google says third parties have no ranking ground truth.

If you run a UAE brand, run a free audit in both languages before the 15th. It is the only way to get a pre-change baseline, and it tells you which of the checks above your site fails today. For the mechanics behind all of this, read how LLMs prioritise information about brands.