Answer Engine Optimisation: What AI Search Means for UAE Businesses

The UAE is the most AI-saturated consumer market on earth, and most websites here are still built for a search behaviour that is quietly disappearing.

Microsoft’s AI Economy Institute put the UAE first globally at the end of 2025, with around 64% of the working-age population regularly using AI tools. Singapore came second at roughly 61%. The United States, which builds most of the world’s frontier models, ranked 24th at 28.3%. Deloitte’s consumer research found 58% of UAE and Saudi consumers have used tools like ChatGPT or Gemini, with 55% of those users engaging weekly or daily. A PYMNTS and Visa study fielded in March 2026 across 5,241 consumers found that 72% of UAE shoppers used AI in some form during their most recent purchase.

That’s the market your website is competing in. Not a future one.

This is a technical post about what changes when the thing reading your site is a language model rather than a person clicking a blue link, and what to do about it. We build for both at Tequila, and the practical work is less exotic than the terminology suggests.

What actually changed

Traditional SEO optimises for position in a list of links. The assumption is that a ranked result produces a click, and the click produces a visit.

That assumption is weakening. Zero-click searches reached 68% of Google queries between January and April 2026 in US data, up from 60.5% in 2024, driven largely by AI Overviews. Pew Research found that about one in five Google searches now produces an AI summary, and that users click a traditional result on only 8% of those visits, against 15% when no summary appears.

So the question shifts. It’s no longer only “does my page rank.” It becomes “when an AI system answers a question in my category, does it mention us, and does it get us right.”

Answer engine optimisation, sometimes called generative engine optimisation, is the work of making that happen. It overlaps heavily with good SEO. Google’s own guidance in May 2026 essentially confirmed that AEO is still SEO rather than a separate discipline. But the emphasis changes in ways that matter.

Why the UAE is further ahead than the guidance assumes

Most AEO advice you’ll read is written for US and European markets, where adoption is lower and the transition is slower. Applying it here understates the urgency by a considerable margin.

Three regional factors compound.

Workplace adoption is unusually deep. PwC’s Middle East workforce survey found 75% of regional employees used AI tools at work in the past year, against 69% globally. Stanford’s 2026 AI Index put regular workplace AI use above 80% in the UAE alongside India, China, Nigeria, Egypt and Saudi Arabia. B2B buyers here are researching suppliers through AI assistants as a default habit, not an experiment. Separately, 6sense found that 94% of B2B buyers used a generative AI tool during their most recent purchase process.

Mobile-first behaviour amplifies it. Mobile accounts for roughly 75.3% of UAE web traffic and 96% of internet users access the web via mobile. AI assistants are a natively mobile interaction. The behavioural shift lands harder in a market where the phone is already the primary screen.

Arabic is a live gap. AI assistants answer in Arabic, and the pool of well-structured, authoritative Arabic content in most commercial categories is thin. For businesses running a genuine Arabic presence rather than a machine-translated afterthought, this is currently one of the least competitive opportunities in the market. If you have an Arabic site, its structure and authority signals deserve the same attention as your English one.

How AI systems decide what to cite

Retrieval systems evaluate candidate sources before generating an answer. Broadly they favour content that is easy to extract cleanly, verifiable, attributable to someone identifiable, and corroborated elsewhere.

That last point produces the most important and least comfortable finding in the current data: roughly 84% of AI citations come from earned media, while a brand’s own website accounts for only 5 to 10% of what AI engines reference. Paid and advertorial content accounts for around 0.3%.

Your own site is necessary but not sufficient. If nobody independent writes about you, AI systems have nothing to corroborate against, and they will cite whoever has been written about instead. For most Dubai agencies and service businesses, that means directories and listicles are absorbing the visibility.

What the evidence does not support

There’s a claim circulating that adding schema markup produces large, measurable increases in AI citations. Figures of 36% and 40% get quoted freely.

Those figures come from vendors selling AEO tools. An Ahrefs study tracking pages that added JSON-LD schema found no statistically significant change in AI citations. That doesn’t mean schema is worthless. It remains genuinely useful for rich results, for entity disambiguation, and for making your content machine-readable at all. But the direct causal link to AI citation volume is not established, and anyone quoting a precise percentage lift is quoting marketing material.

We implement structured data on every site we build because the SEO case for it is solid and the cost is low. We don’t sell it as an AI visibility guarantee.

What actually to do

Six things, ordered by how much difference they make.

Earn third-party coverage. Given that 84% of citations come from earned media, this is the highest-leverage work available and it isn’t a website task. Regional trade press, industry publications, credible directories, and genuine expert contributions do more for AI visibility than any on-page change. Uncomfortable if you were hoping for a technical fix, but it’s what the data says.

Answer questions directly, at the top of the section. Lead with the answer in the first two sentences, then explain. Retrieval systems extract passages, and a passage that only makes sense with three paragraphs of preceding context is a passage that won’t get lifted. This is why FAQ sections perform well: they are structurally self-contained by design.

Use tables for anything comparative. Content with tables is cited noticeably more often, reportedly around 2.5 times. Comparison tables, pricing ranges, and specification breakdowns present structured relationships that models extract cleanly.

Attribute content to real, identifiable people. Author schema linked to a named person with verifiable credentials is the technical implementation of expertise signals. A page written by “Admin” carries less weight than one written by someone with a bio, a role, and a traceable professional history.

Keep content genuinely current. Freshness is a meaningful factor in citation selection. A post last updated three years ago competes badly against one updated last quarter, regardless of which is better.

Include specific, sourced numbers. Models favour content with concrete data over general assertion. A paragraph containing a named source and a dated figure is more citable than the same claim stated vaguely. This connects directly to our SEO work, where the content standard and the technical standard have to move together.

What this means for how sites get built

A few build-level implications worth naming.

Heading structure has to reflect actual content boundaries rather than visual rhythm. If your H2s exist because they looked right at that size, you’ve made your page harder to segment.

Content in JavaScript-rendered components may not be reliably retrieved. Anything you want cited should exist in the server-rendered HTML.

Answers buried behind accordions, tabs, or click-to-reveal interactions are less reliably extracted than visible text. That’s a design decision with a retrieval consequence.

And your llms.txt and robots.txt should not be blocking the crawlers you want reading you. Worth checking, because default configurations frequently do.

How to measure it

None of this appears in Google Analytics in a form you can read directly. Practical approach:

• Query the assistants your customers use, in both English and Arabic, with the questions they’d actually ask. Record whether you appear and whether the description is accurate.
• Track referral traffic from AI domains in GA4. Volume will be small. Quality tends not to be: Adobe found AI referrals converted 31% better than non-AI traffic during the 2025 holiday period.
• Monitor brand mentions rather than only links. Being named without a link still shapes the answer.
• Track share of voice across a set of category prompts, not a single query. Appearing once is noise.

Frequently asked questions

Want to know whether AI assistants currently mention your business, and what they say when they do? Talk to us. We’ll run the queries and show you what comes back.