UAE local businesses are ranking on Google and disappearing from AI answers

Most UAE businesses that invested in SEO over the last decade are still being told their search performance is healthy — and by the old measure, it is. But a growing share of buying questions never reach a results page. Someone asks ChatGPT which garage in Al Quoz to trust, or asks Gemini for affordable apartments in Dubai Marina, and receives a short answer naming three businesses. If you are not one of the three, you were not outranked. You were not present at all.

We wanted to know how large that gap actually is for UAE businesses, so we measured it rather than estimated it. This report publishes the method, the per-engine results, and the specific mechanism that produces the gap.

What we measured

Two UAE businesses, tracked daily across six AI answer engines — ChatGPT, Perplexity, Gemini, Claude, Grok, and Google AI Overviews — using buyer-intent prompts written the way customers actually phrase them, not the way marketers write keywords.

Both businesses show the same shape: a functioning traditional search presence and a near-zero citation rate in AI answers. The real-estate listings site is the sharper case — it is competing in Dubai real estate, one of the most heavily marketed verticals on earth, against Bayut, Dubizzle and Property Finder, and it appears in generated answers for effectively none of the eighteen buying questions we tested.

The branded-prompt trap

This is the finding that matters most, and it is the one that misleads business owners most often.

The clearest example comes from outside the two sites above: for a software brand we track, the branded prompt (‘is <brand> the most affordable tool for AI visibility’) was cited at position one on four of six engines; the unbranded version (‘best AI visibility tracking tool’) returned zero citations.

A branded prompt measures whether an engine can look your name up. It does not measure whether the engine recommends you to someone who has never heard of you.

Every buyer who matters is in the second category. They do not know your name yet — that is the entire reason they are asking. So a dashboard that reports healthy visibility on branded prompts is reporting a vanity metric. We treat unbranded buyer-intent prompts as the only commercially meaningful measure, and we recommend you do the same regardless of which tool you use.

Why traditional SEO does not carry over

The instinct is to assume AI visibility is downstream of SEO — rank well, get cited. Our data does not support that. The car-servicing directory has nearly double the listings site's traditional search score and a materially identical AI visibility score. The correlation across our measurements is weak enough to be unusable as a predictor. We go through which SEO work does carry over in From SEO to AEO.

The mechanism is that the two systems reward different objects. Traditional ranking evaluates pages against a query. Generated answers assemble a response from entities the model already recognises and sources it already trusts. A well-optimised page can rank on results page one while contributing nothing to whether a model considers the business a real, established entity worth naming.

What appears to drive the gap

• Thin entity footprint. The business exists as a website but not as a structured, corroborated entity across the wider web — no consistent directory presence, no third-party corroboration of what it does and where.
• Nothing quotable. Pages are written to rank, in scannable marketing fragments. Generated answers need a self-contained passage that survives being lifted out of context and still makes a factual claim.
• No third-party sourcing. Engines overwhelmingly cite sources other than the business's own site. A business with no independent coverage has nothing for a model to cite about it. Our later source-anatomy study put a number on this for one category: the vendor's own site supplied 4.9% of the evidence.
• Engine-by-engine variance. Coverage is not uniform. The same prompt returns different sources on different engines, which is why measuring three engines and extrapolating produces a misleading number.

Method, in full

We are publishing the method in enough detail to be repeated, because a benchmark nobody can repeat is worth little.

• Instrument. A commercial AI visibility tracker, running daily automated scans across ChatGPT, Perplexity, Gemini, Claude, Grok and Google AI Overviews.
• Prompt design. 18 buyer-intent prompts per business, phrased as customer questions. Both branded and unbranded prompts are included so the two can be compared directly.
• Citation detection. A prompt counts as cited only when the business is named or linked in the generated answer. Appearing in an engine's underlying search results without appearing in the answer text does not count.
• Cadence. Daily. Answer engines re-rank continuously and the same prompt can return different sources on consecutive days, so single-point audits cannot separate signal from variance.
• Period. Figures in this report are from scans dated 7 and 8 August 2026.
• Error handling. Engine timeouts are recorded as errors and excluded from the denominator rather than counted as non-citations.

What this means if you run a UAE business

Three things follow from the data, in order of how much they matter.

1. Your current reporting probably does not cover this

Google Search Console shows impressions and clicks from Google. It cannot show that a customer asked ChatGPT for a recommendation, received three business names, and none of them were yours. That traffic does not appear as a loss anywhere in your analytics. It appears as nothing at all — which is why most owners have no idea it is happening.

2. Measure unbranded prompts or do not bother

Whatever tool you use, insist on unbranded buyer-intent prompts and on coverage of all the major engines. A tool measuring three engines on branded prompts will hand you a comfortable number that means nothing.

3. The fix is entity work, not more blog posts

Publishing more pages into the same thin entity footprint does not move these scores. What moves them is making the business legible as an entity — consistent structured data, corroborated presence across directories the models actually read, and content written in self-contained, quotable passages that answer a real question in full. That is the work our AI search visibility service is built around.

We are extending this

This report covers two businesses across two sectors. We are expanding the measurement to a wider set of UAE local businesses — garages, salons, movers, clinics, contractors — across all seven emirates, and will publish the full dataset under CC BY 4.0 when it is complete.

If you run a UAE business and want your own baseline included, we will measure it and send you the raw grid at no cost. We publish aggregates; your individual scores stay yours unless you tell us otherwise.

This report is licensed CC BY 4.0: cite it freely with attribution.

Questions people ask