Who Holds the AI Answers in Dubai: Nobody Does

The assumption underneath most AI-visibility spending is that you are competing with a handful of rival websites for the answer. We measured it on our own panel and that is not the shape of this category at all. There is no incumbent to displace. There is a crowd.

The short version

Across our 62-query panel, run through Google AI Mode on 16 and 18 September 2026, the engine cited 301 sources drawn from 178 different domains. One hundred and thirty-one of those domains were cited exactly once. The largest single source in the entire panel is LinkedIn, at 6% of citations. YouTube, LinkedIn, Reddit and Quora together hold 54 of the 301 citations — 18%, the biggest bloc by some distance, and they appear in 27 of the 56 answers that cited anything at all. UAE domains hold 10%. Our own domain holds 1.7%. No rival agency domain exceeds 2%.

What we counted

Source class	Share of 301 citations	What it means for you
Platforms (YouTube, LinkedIn, Reddit, Quora)	18%, and present in 27 of 56 answers	Participation, not optimisation
UAE domains (.ae)	10%	The local field, fragmented
Our own domain	1.7%	Protect and extend
Everything else — a long tail	70%, across 178 domains	131 domains cited exactly once

The method is the one behind every number on this site: we ran real buyer questions through Google AI Mode and saved every answer and source list as a dated receipt. Our receipts are the input to the figures above. Counts are domain-level, and both figures above are recomputed from those receipts by a check that refuses any published number it cannot reproduce.

A correction, because the alternative is worse. An earlier version of this research circulated internally as "platforms hold 57% of citations", and a later reading was published on this site as "49% of citations". Both were wrong in the same way: that is roughly the share of answers citing a platform somewhere, not the share of citations sitting on one. The citation share is 18%. The answer-level figure, about half of all answers touching a platform, is the real and still-striking finding. We corrected it here rather than leave the more dramatic number up.

Why the tail is so long in a young category

The category's own content is thin

Dubai AEO is new enough that few sites have deep coverage. The engine reaches for whatever answers the question, which is why a detailed Reddit reply frequently beats a service page that only sells.

Platforms carry the discussion format

Buyer questions in this space are comparative and sceptical. Forums and video answer in that register natively; a service page does not. That is why platforms are the largest bloc even while no single one dominates.

The engine trusts corroboration over assertion

A vendor page claiming excellence and a thread of practitioners arguing about vendors are different evidence types. One is a claim; the other reads as independent corroboration.

What this changes about budget

1. Split your target questions before spending. For each, read the source list and mark it on-domain-winnable or held elsewhere. Fund them differently.
2. Stop buying on-domain work for answers assembled from platforms. No page you publish reaches an answer built from a Reddit thread and a YouTube video.
3. Fund participation as a line item. Answering the specific threads the engine quotes is reachable, and almost nobody staffs it.
4. Treat the long tail as pitchable. Seventy per cent of citations sit on domains that are not platforms and not competitors — publications, tool blogs, roundups. Those accept contributions.
5. Do not fear the competitor you were worried about. No rival agency domain held more than 2% of citations. You are not being beaten by one site; you are absent from a crowd.

This is what AI Search Visibility actually looks like when you measure it rather than buy a score for it. The citation count is a retrieval fact, not a ranking; treating it as a leaderboard is measuring the wrong thing. And because most of the sources sit somewhere you do not own, the work is the away game before it is an on-site project.

The uncomfortable part for agencies

An AEO retainer that touches only your website is structurally capped at the share of answers your own domain can reach. That is not a criticism of the work — passage structure genuinely moves retrieval, as the Princeton GEO research measured at up to 40% under controlled conditions. It is a statement about reach. Ask any agency what proportion of your target answers are held off your domain before you sign, and see whether they have counted at all.

Questions buyers ask

Is this split stable? No, and we would not claim it is. It is the reading on two capture dates and we expect it to move, because everything in this category does. Our churn measurement found complete turnover among agency leaders inside 52 days. Read it as a trend across repeat runs, never as a single snapshot.

Should I make YouTube videos then? If the answers you care about cite YouTube, yes — and the measurement tells you, not the trend. Check your own source lists before committing to a format.

Does posting on Reddit work? Answering the specific threads the engine quotes works. Posting promotional content does not, and communities punish it faster than engines reward it.

Why is your own share only 1.7%? Because we are early on our own coverage and publish the number rather than hide it. Across the same 62 questions we were absent from 59. As Firoz Azees puts it: "a measured zero is a work order; a flattering estimate is a story you tell yourself."

Do these shares differ by industry? Almost certainly. Ours is a young, contested, advice-heavy category, which favours discussion surfaces. A category with established trade publishers would likely tilt towards them instead.

How do I count this for my own market? Ask your eight buyer questions, read the source list under each answer, and classify every domain. It takes under an hour and it is the single most decision-changing hour in this discipline.

What if my competitor is in the tail? Then they won a narrow slice — no domain in our panel held more than 6%, and theirs is smaller. Study the page that got cited, then look at how much of the answer neither of you holds.

At Ivanooo, Firoz Azees runs Distinctiveness Engineering for the AI-answer era: measuring who the engines name, cite and recommend, then engineering the gap between listed and chosen. Start with a free AI visibility check and count your own split.