English and Arabic AI search visibility in the UAE: how to get cited by ChatGPT, Gemini and Google AI Mode

On 28 October 2026 we ran 32 buyer conversations, five questions each, in ChatGPT and Google AI Mode with a 🇦🇪 UAE location. The conversations started from eight commercial questions about AEO and GEO agencies and AI visibility platforms in the 🇦🇪 UAE. A judge model from a different model family recorded, with verbatim quotes, why each answer kept or dropped each brand.

One question stood out: "Best AEO and SEO agency in MENA for English and Arabic AI search visibility". On that question, all 20 of 20 answer turns discussed Arabic. Language and market fit was cited as a reason 19 times — no other question went above 12. When a buyer names both languages, Arabic support becomes an entry condition for the shortlist, not a bonus.

The judge captured how the engines described the winners. One agency was called "a potential fit for 🇦🇪 UAE and GCC brands that need English and Arabic visibility across generative search". Another was praised for "bilingual GEO capabilities".

On the bilingual MENA question, the brands shortlisted most often were SEO Sherpa (9 turns), WAIM (5), VOCTOS (5), Omni Eclipse (4) and Digital Gravity (3). The engines leaned on the agencies' own pages for that question: seosherpa.com was cited 34 times, voctos.com 20, waimhub.com 16, digitalgravity.ae 14, omnieclipse.ai 12, and the review directory clutch.co 11.

Across all eight 🇦🇪 UAE questions, the reasons the engines gave for keeping a brand ranked like this:

Reason given in the answer	ChatGPT	Google AI Mode
Case studies with measurable results	52	44
Depth of expertise in the specific service	35	32
Third-party reviews	11	—
Tools and integrations	—	10

One more pattern matters. Self-published "best agencies in 🇦🇪 Dubai" listicles were never cited in these conversations. The agencies that won did it with their own service pages carrying numbers, and with independent directories carrying reviews.

Classic SEO ranks pages by links and relevance. AI answers work differently: the system retrieves a set of candidate passages, then composes an answer from the few it trusts most. A brand that ranks first in Google can still be missing from the answer if its pages do not state clear, extractable facts.

Google's own guidance for AI features in Search makes the same point from the publisher side: the content that appears is helpful, clear and well-structured, and there is no special markup that guarantees inclusion ^[1]. For a 🇦🇪 UAE B2B firm, that means the work shifts from "rank the page" to "make the page quotable".

Three quick checks before you start

First, ask ChatGPT "What does [your brand] do?" in English and then in Arabic. If the model invents details or declines, your entity signals are weak. Second, view the source of your top three service pages and search for schema.org. If nothing appears, you are giving machines no structured facts. Third, put the company description from your website, LinkedIn and Google Business Profile side by side. Any mismatch is a reason for an engine to trust you less.

AEO makes the brand recognisable and quotable inside answers. GEO builds source material broad and dense enough to be selected when the answer is composed. They target different failures.

A page that states "We operate in the 🇦🇪 UAE, Saudi Arabia and Qatar; founded 2017; 42 engineers; ISO 27001 certified" gives a retriever five facts in one sentence. A page that says "We are a leading regional partner for digital transformation" gives it none. Our measurements point the same way: the top reason for keeping a brand was case studies with measurable results.

Where to start: AEO or GEO

If the AI cannot describe you accurately, start with AEO: fix the canonical name, the schema and cross-platform consistency. If the AI describes you correctly but leaves you out of comparison answers, start with GEO: you have identity but not depth. Spending GEO budget before the entity is clear is a common mistake, because new content attaches to a brand the engine still cannot resolve.

Different sources. Arabic questions are mostly answered from Arabic pages, English questions from English pages. A strong English page does not automatically carry its visibility into the Arabic answer.

Different phrasing. Many Gulf buyers type in dialect, mix Arabic with English terms, or write the company name in Latin letters one time and Arabic letters the next. Your Arabic content needs to cover those phrasings in its FAQ, while the body copy stays in Modern Standard Arabic.

Translation does not get quoted. A machine-translated page carries English sentence structure inside Arabic words and reads awkwardly to a native reader. In our experience such pages rarely become the passage an Arabic answer quotes. That is why we write Arabic pages natively, from the buyer's question, in Modern Standard Arabic — not by translating the English version.

1. Fix the entity across both scripts. Use one canonical company name in Arabic and one in Latin letters, and use them identically on your site, LinkedIn, Google Business Profile, Crunchbase and local directories. In your Organization markup, list the Arabic name in alternateName and your official profiles in sameAs ^[2]. "ACME FZ-LLC", "ACME 🇦🇪 Dubai" and "ACME Group" used interchangeably look like three companies.

2. Link the two language versions. Every Arabic page needs its English counterpart and vice versa, declared with hreflang on both pages, each page also referencing itself ^[3]. Give each language its own URL rather than switching language by cookie, so both versions can be crawled and indexed ^[4].

3. Write the Arabic pages natively. Start from the Arabic buyer's question, not from the English text. Put the answer in the first paragraph, and state figures, timeframes and AED price ranges wherever you can. Cut generic phrases that carry no number, name or example — an engine has nothing to quote in them.

4. Publish measurable proof in both languages. Case studies with measured results were the top reason engines kept a brand in our conversations. In our work with Raqma, a B2B technology company serving MENA and international markets, AI answers named or cited the brand in 202 of 594 tracked prompts — a 34.01% mention rate — and 591 of those prompts did not contain the brand name. 564 of 2,934 AI answers linked to Raqma's own domain, with an average citation position of 2.75. Part of that work was bilingual entity markup kept consistent across both languages. Put numbers like these on the case page itself, with the date and method of measurement.

5. Be present where the engines already look. In our 🇦🇪 Dubai ChatGPT conversations, the most cited domains after agency sites were directories and review platforms: clutch.co (49), g2.com (42), thehovi.com (42) and dubailocal.ae (34). In Google AI Mode, linkedin.com appeared 7 times. Complete profiles on these surfaces, with service descriptions in both languages, are part of the visibility plan. In some sectors the answer layer sits almost entirely off your site: for our 🇦🇪 Dubai real-estate client Whitewill, 121 investment questions returned zero brand mentions at baseline, while Medium held about 97 citations and LinkedIn 44. The plan there started with publications, not the website.

The two engines also lean on different sources. Across our 🇦🇪 UAE conversations, ChatGPT drew most heavily on agency sites and directories — digitalgravity.ae alone was cited 96 times and seosherpa.com 74. Google AI Mode cited far fewer domains overall, and its list looked different: digitalmediasapiens.com (11), linkedin.com (7), globalsurf.ae (7) and searcheye.io (7). Being strong in one engine does not mean being present in the other, so check each engine on its own row of the measurement panel.

Run the English and Arabic tracks in parallel. "English first, Arabic later" usually means Arabic never ships.

What you measure	Arabic	English
Question panel	Questions written by Arabic-speaking buyers, in MSA and dialect	Questions as buyers write them in English
Location	🇦🇪 UAE	🇦🇪 UAE
Engines	ChatGPT, Gemini, Google AI Mode, Perplexity	Same
Metrics	Mention rate, citations of your domain, position among named brands	Same

Do not merge the two languages into one number. English can rise while Arabic stays at zero, and an average will hide it. Re-run the same panel every month so you compare like with like.

A spreadsheet is enough to start: query, language, engine, mentioned (Y/N), your domain cited (Y/N), position, date. Engines refresh what they retrieve on their own schedule, so judge the trend over several monthly runs rather than a single week.

When the numbers say act

If the brand is described correctly but still missing from comparison answers, add depth to the pages buyers compare: pricing ranges, regulation, implementation steps, and a dated case. If your domain is cited as a source but the brand is not named in the answer text, rewrite the opening paragraph of that page as a direct answer that names you.

If your buyers are global teams searching only in English, and your contracts and correspondence are in English, prioritise the English track and review platforms. The easiest way to know is to measure first: if your buyers do not ask in Arabic, an Arabic page will not change their shortlist. If you sell to government or semi-government entities, or to family businesses in the Gulf, Arabic is usually part of the buying decision itself.

[1] AI Features and Your Website — Google Search Central — https://developers.google.com/search/docs/appearance/ai-features

[2] alternateName — Schema.org — https://schema.org/alternateName

[3] Localized Versions of your Pages — Google Search Central — https://developers.google.com/search/docs/specialty/international/localized-versions

[4] Managing Multi-Regional and Multilingual Sites — Google Search Central — https://developers.google.com/search/docs/specialty/international/managing-multi-regional-sites