Gemini local search UAE results are not stable, and that is the single most important fact a UAE business can learn this month. A study published on 19 August 2026 analysed 14,472 citations from 1,487 local search queries and found that Gemini recommends the same top business only about 7% of the time when the identical question is asked again. Google’s traditional local pack returns the same top listing roughly 90% of the time. Run the same query twice in Gemini and the cited sources overlap only 40% to 46% of the time.
There is a second finding that changes where the work goes. Almost 60% of everything Gemini cited pointed at the business’s own website. Reddit came second on 13.7%. Every local-service directory in the study, combined, managed 10.3%. So the answer for UAE brands is not “get listed in more directories”. It is: publish the answers on your own site, in language a model can lift, and accept that visibility is now a probability rather than a position.
This post covers what the study found, why the UAE feels it faster than most markets, and the practical playbook for earning citations that hold up across repeated queries.
In this article
- What the 14,472-citation study found
- Grounding drift: why Gemini local search UAE results reshuffle
- Your own website out-cites every directory combined
- Why the UAE feels this faster than most markets
- Gemini and ChatGPT agree on the top business 4.2% of the time
- The local AI visibility playbook
- How to measure something that keeps changing
- Frequently asked questions
- The bottom line for UAE brands
What the 14,472-citation study found
Local SEO firm Steady Demand ran 1,487 local service queries through Gemini across the 50 largest US metro areas and 10 service categories, then logged every source the model cited. The dataset covers 14,472 citations, 3,611 unique domains and 4,410 unique businesses.
Three numbers do most of the work. Business websites took 59.9% of all citations. Reddit took 13.7%, beating every local-service directory in the study put together at 10.3%. And when the researchers repeated the exact same query with no change in wording, Gemini named the same top business only 7% of the time, against 90% for Google’s local pack on the same searches. The findings were reported on 19 August 2026.
The quality bar is worth noting too, because it is lower than most people assume. AI-recommended businesses averaged 4.75 stars. The plain Google baseline for the same queries averaged slightly higher at 4.84. Nine in ten AI-recommended businesses were rated 4.5 or above. Reviews get you into the pool, they do not get you picked.
| Consistency test | Result | What it means for a brand |
|---|---|---|
| Google local pack, same query repeated | 90% same top listing | A ranking you can hold and report on |
| Gemini, same query repeated | 7% same top business | Presence is probabilistic, not positional |
| Gemini, cited domains on repeat queries | 40–46% overlap | Over half the source set changes each time |
| Same question, three different phrasings | 40% average agreement | Wording changes who gets named |
| Gemini vs ChatGPT, same query | 8% domain overlap | Each engine needs its own strategy |
One caveat, stated plainly: the study ran in US metros. The engine, the grounding pipeline, and the retrieval behaviour are the same ones serving Dubai and Abu Dhabi, but the source mix in a UAE query will differ. Treat the percentages as the shape of the problem rather than a UAE benchmark.
Grounding drift: why Gemini local search UAE results reshuffle
The researchers call it grounding drift. When a generative model answers a local question, it runs its own retrieval pass, picks a handful of sources, and writes an answer from them. That retrieval pass is not deterministic. Ask twice, and the model can pull a different set of pages, then name a different business as the best option, with no change to your website, your reviews, or your Google Business Profile.
This is not the ordinary ranking volatility that SEO teams already manage. The same study measured Google’s local pack as a control and it held its top result 90% of the time. The instability is specific to the AI answer layer. Ben Fisher, Steady Demand’s co-founder, put it well: a single AI search should be treated as a snapshot, not a reliable measure of long-term visibility.
Phrasing matters as much as repetition. Asked the same question three different ways, source agreement averaged 40%, and it ranged from 3% in the worst category to 87% in the best. In practice that means your customers are not all seeing one answer. They are each seeing a draw from a distribution, and your job is to raise your odds across that whole distribution rather than to win a single position.
A single AI search should be treated as a snapshot, not a reliable measure of long-term visibility. Ben Fisher, co-founder, Steady Demand.
Your own website out-cites every directory combined
For a decade, local visibility work in the UAE has meant a Google Business Profile, a review programme, and a spread of directory listings. The citation data reorders that. Gemini pulled 59.9% of its citations from business websites, 14.4% from social and community forums, and 10.4% from general directories. Reddit alone, at 13.7%, out-cited the entire local-services directory category.
There is a long tail underneath that headline. Most cited domains appeared only once or twice across the whole dataset, which means individual businesses are reaching AI answers without a directory acting as gatekeeper. National platforms such as Angi, BBB and ConsumerAffairs showed up in all 50 metros, so they set a floor, but only 2.0% of businesses were recommended in more than one metro. Local AI visibility is won locally, page by page.
The practical read for a UAE brand: the service page is the asset. If your Dubai clinic, law firm, or fit-out contractor has one thin page covering nine services and three emirates, there is very little for a model to ground on. Split it. Give each service and each location a page that answers the real question, states prices or ranges where you can, names the neighbourhoods you serve, and carries the sort of specifics a model can quote without hedging. That is content and SEO strategy work, and it now pays into two channels at once.
Why the UAE feels this faster than most markets
The UAE is unusually exposed here, in both directions. Internet penetration sits at about 99%, mobile drives 77.6% of web traffic, and Google Maps reaches more than 84% of smartphone users. Roughly 78% of UAE consumers search online before visiting a local business, and more than 63% contact businesses directly from a Google Business Profile. “Near me” queries grew more than 180% between 2021 and 2024. Dubai alone accounts for around 65% of UAE local business search volume.
Now layer the AI surfaces on top. Google brought AI Overviews to MENA and to Arabic, AI Mode has been rolling out across the region, and Gemini’s Personal Intelligence features reached UAE users in April 2026. The Gemini app passed 900 million monthly active users by Google I/O in May 2026, up from 750 million at the end of 2025. A market that runs its local discovery through Google on mobile is, by definition, a market where the AI answer layer intercepts local intent first.
| UAE local search signal | Figure | Why it raises the stakes |
|---|---|---|
| Internet penetration | ~99% | Almost every buyer starts online |
| Mobile share of web traffic | 77.6% | AI answers dominate a small screen |
| Research before visiting a local business | 78% of consumers | The answer decides the visit |
| Contact a business from its Google profile | 63%+ of users | Discovery and enquiry happen in one step |
| Dubai share of UAE local searches | ~65% | Dense competition in one answer slot |
There is an upside hiding in the volatility. If the top spot reshuffles on nearly every query, then incumbency is worth far less than it was. A well-documented challenger with sharp service pages can enter the rotation in weeks, where climbing the map pack against an entrenched competitor could take a year.
Gemini and ChatGPT agree on the top business 4.2% of the time
The researchers put identical queries through both engines. The cited domains overlapped just 8% of the time, and the two models named the same top business in only 4.2% of cases. These are not two views of one ranking. They are two separate discovery channels that happen to answer the same question.
Their taste differs too. Gemini leaned on business websites at 59.9% of citations. ChatGPT, measured in July 2026, leaned on community forums at 41.7% and general directories at 34.6%, with business websites at 15.9%. That mix then shifted sharply through August, with forums falling away and business websites and directories rising. Both engines are still tuning what they trust, which is one more reason to hold your own site as the constant.
| Citation source | Gemini | ChatGPT (July 2026) |
|---|---|---|
| Business websites | 59.9% | 15.9% |
| Social and community forums | 14.4% | 41.7% |
| General directories | 10.4% | 34.6% |
The local AI visibility playbook
Five moves, in the order we would run them for a UAE client.
1. Rebuild the service pages as answers
One page per service, per emirate where it makes commercial sense. Open each page with a direct answer to the question a buyer would type, then support it with price ranges, timelines, qualifications, and the areas you cover. Models quote specifics and skip vague copy.
2. Make the facts machine-readable
LocalBusiness and Service schema, consistent name, address and phone details across every profile, opening hours that match reality, and an FAQ block on each service page. If a model has to guess at your details, it will find a competitor whose details are unambiguous.
3. Earn presence where people actually discuss you
Reddit carried 13.7% of Gemini’s citations. In the UAE the equivalent conversations also live in local community forums, WhatsApp groups that spill into public posts, expat groups, and industry associations. Show up as a named expert answering questions, not as a brand account dropping links.
4. Keep the review base above the bar, then stop optimising it
Ninety per cent of AI-recommended businesses sat at 4.5 stars or above, and the AI-recommended average of 4.75 was actually below the 4.84 Google baseline. Get above the threshold and keep it there. Chasing a higher star average past that point buys you very little in the answer layer.
5. Diverge by engine
With 8% domain overlap between Gemini and ChatGPT, one strategy cannot serve both. Your site does the heavy lifting for Gemini. Third-party mentions, directories and community threads matter more for ChatGPT. Budget for both.
How to measure something that keeps changing
The old local SEO report, a rank for a keyword on a date, does not survive contact with a 7% repeat rate. Checking your visibility once and calling it a result is how brands end up either falsely reassured or needlessly alarmed.
Measure share of voice instead. Pick 20 to 30 queries a real customer would ask, run each one 10 times across Gemini and ChatGPT, and record how often you are named and how often you are cited. That gives you a percentage you can move and compare month to month. Track phrasing variants separately, because agreement between phrasings ran as low as 3% in some categories. Then pair it with the AI performance data now surfacing in Google Search Console so you can see the traffic side of the same story.
One honest warning. Anyone promising you a guaranteed position in an AI answer is selling something that the data says does not exist. What you can influence is frequency, and frequency compounds.
Frequently asked questions
What is grounding drift in Gemini local search?
Grounding drift is the tendency of a generative search engine to retrieve a different set of sources each time it answers the same question. In the August 2026 Steady Demand study, repeated identical Gemini queries produced only 40% to 46% overlap in cited domains, and the same top business was named just 7% of the time.
Does Gemini local search UAE visibility replace my Google Business Profile?
No. The Google Business Profile still drives the map pack, and more than 63% of UAE users contact businesses directly from it. Gemini local search UAE visibility is an additional layer that draws heavily on your own website, so both need attention rather than one replacing the other.
Which sources does Gemini cite most for local queries?
Business websites accounted for 59.9% of Gemini’s local citations in the study. Social and community forums took 14.4%, with Reddit alone at 13.7%, and general directories took 10.4%. All local-service directories combined reached only 10.3%.
Do I need a different strategy for ChatGPT and Gemini?
Yes. The two engines cited the same domains only 8% of the time and named the same top business in 4.2% of queries. Gemini favours business websites, while ChatGPT drew more from community forums and directories in July 2026, so each channel needs its own emphasis.
How many reviews do I need to be recommended by AI search?
There is a threshold rather than a target. Ninety per cent of AI-recommended businesses were rated 4.5 stars or higher, and recommended businesses averaged 919 reviews against a baseline of 846. The AI-recommended star average of 4.75 was slightly below the 4.84 Google baseline, so pushing your rating higher does not reliably buy more AI visibility.
Does the US study apply to Dubai and Abu Dhabi?
The queries ran in 50 US metros, so the exact source mix will differ in the UAE, where local directories and community platforms are different. The retrieval behaviour being measured belongs to the model rather than the market, so the instability itself carries over. Treat the percentages as the shape of the problem and run your own query set locally.
How should I report AI visibility to management?
Report a share of voice percentage, not a rank. Run 20 to 30 customer questions 10 times each across Gemini and ChatGPT, then report how often the brand is named and cited. A single check is a snapshot and can swing wildly between runs.
Is instability good or bad for a smaller UAE business?
On balance it favours challengers. When the top recommendation changes on almost every query, an entrenched competitor cannot hold the slot, and only 2.0% of businesses in the study were recommended across more than one metro. A smaller business with strong, specific service pages can enter the rotation far faster than it could climb the traditional map pack.
The bottom line for UAE brands
Local visibility has stopped being a position and started being a probability. Gemini names a different top business on roughly 93% of repeat queries, cites your own website more than every directory put together, and disagrees with ChatGPT almost entirely. The response is not to chase the answer box. It is to build a site so specific, so well structured, and so consistently referenced elsewhere that you turn up in a large share of the draws.
For UAE businesses that is a genuine opening. The market is mobile-first, Google-dominant, and moving to AI answers quickly, and most competitors are still measuring a map pack rank once a month.
Find out how often AI actually names your business
Rothian Digital runs repeat-query AI visibility audits across Gemini and ChatGPT for UAE brands, then rebuilds the service pages, schema and content that decide whether you get cited. You get a share of voice baseline and a plan to raise it, not a one-off screenshot.
Sources
- Search Engine Land, “Business websites dominate Gemini citations in local AI search: Study”, 19 August 2026
- Steady Demand, AI Citation Ledger: 14,472 local AI citations analysed
- Google, “Bringing AI Overviews to MENA, and in Arabic globally”
- UAE SEO and local search statistics 2026
- Demandsage, Google Gemini user statistics 2026
- Khaleej Times, Gemini Personal Intelligence in the UAE
