Ask three AI engines who's best in Dubai: the overlap is low, not zero
By Joe Hickman, Founder, BrandureResearch conducted Published

We asked ChatGPT, Perplexity and Gemini the same “best in Dubai” question across seven categories on the same day. Between them they named 147 businesses. Ten appeared on all three engines. In three of the seven categories, not one business made all three lists.
The overlap is low, but it isn’t zero, and where it exists it has a cause: a ranking body or a body of trade press that all three engines read. The bigger surprise was inside each engine. ChatGPT asked the same question minutes apart named the same clinic first both times, then a different set of clinics below it.
For a Dubai business, the practical point is that a visibility score from one engine on one run tells you about that engine on that run. It won’t tell you what the other engines say, or what the same engine says next time.
This is the second of three pieces from our September test. Piece 1 covers companies writing the articles AI quotes about them. Piece 3 covers what each engine reads before it answers.
What did we test?
We asked ChatGPT, Perplexity, Gemini and Google Search “Best [category] in Dubai” across seven categories: law firms, B2B SaaS, aesthetic clinics, boutique hotels, industrial manufacturers, universities and car dealerships. All logged out, free tier, UAE location, 44 answers in total. Clinics and manufacturers were run three times on every surface. The overlap figures below compare the three chat engines; Google’s AI Overview appeared on four of the seven categories, so it’s covered separately.
How often do the AI engines name the same businesses?
We counted every business an engine recommended, including “also worth considering” lists and separate tiers, and excluded passing mentions in caveats. For clinics and manufacturers we pooled the three runs on each engine.
| Category | Businesses named across the three engines | ChatGPT and Perplexity | ChatGPT and Gemini | Perplexity and Gemini | On all three |
|---|---|---|---|---|---|
| Law firms | 24 | 2 | 4 | 4 | 0 |
| B2B SaaS | 18 | 5 | 3 | 6 | 3 |
| Aesthetic clinics | 25 | 2 | 1 | 3 | 0 |
| Boutique hotels | 14 | 3 | 1 | 1 | 1 |
| Industrial manufacturers | 34 | 2 | 2 | 5 | 2 |
| Universities | 14 | 6 | 6 | 7 | 4 |
| Car dealerships | 18 | 1 | 6 | 2 | 0 |
| Total | 147 | 10 |
The ten businesses on all three engines:
| Category | On ChatGPT, Perplexity and Gemini |
|---|---|
| Universities | American University in Dubai, Canadian University Dubai, University of Wollongong in Dubai, University of Dubai |
| B2B SaaS | Bayzat, GrubTech, Dukkantek (DTEK) |
| Industrial manufacturers | Emirates Global Aluminium (EGA), Ducab |
| Boutique hotels | Beach Walk Boutique Hotel |
About 7% of the businesses named made every list. A business can lead one engine and be missing from another, and in law firms, clinics and dealerships that was true of every business named.
Why do Perplexity and ChatGPT disagree?
They read different things. In almost every category, Perplexity grounded its answer on editorial coverage and ChatGPT grounded its answer on listings and directories.
| Sources it leaned on | |
|---|---|
| Perplexity | GRAZIA, Country & Town House, What’s On, Upgraded Points, Manufacturing Digital, Legal 500 |
| ChatGPT | Bayut, Best in Hood, Doctify, Dubai Industrial City, dubicars.com |
Industrial manufacturers show it most clearly. Across three runs each, ChatGPT’s shortlist was SME fabricators: Al Shirawi, SIGMA, NOVA Composites, CAMTECH, Johar, Hotpack. Perplexity’s was heavy industry: EGA, Ducab, Jebel Ali Cement, Jubaili Bros, All Metals, Invention Steel. EGA and Ducab were the only names on both, and on ChatGPT they sat in a separate “large companies” section rather than the shortlist. The two engines didn’t share a single source either. ChatGPT cited Bayut, Dubai Industrial City, Dubai Holding, Dubai Investments and ICD. Perplexity cited mazeed, UAE Business Blog and Manufacturing Digital.
ChatGPT, logged out, UAE. SME fabricators in the shortlist, heavy industry set apart. Recaptured for publication; the shortlist differs from the logged runs, which is the next section’s point.
Same prompt, two readings of what a “manufacturer” is. Gemini sided with Perplexity: it shared five manufacturers with Perplexity and two with ChatGPT.
Does the same engine give the same answer twice?
Not below first place. We ran ChatGPT’s aesthetic clinics answer three times, minutes apart.
| Position | Run 1 | Run 2 | Run 3 |
|---|---|---|---|
| 1 | Biolite Clinic Dubai | Biolite Clinic Dubai | Biolite Clinic Dubai |
| 2 | Los Angeles Aesthetic Medical Center | Glow Aesthetics Dermatology Clinic | CLINICA Cosmetic Poly Clinic |
| 3 | CLINICA Cosmetic Poly Clinic | Los Angeles Aesthetic Medical Center | New Me Clinic |
| 4 | Athena Dermatology Clinic | Athena Dermatology Clinic | Five Cosmetic |
| 5 | EDEN Aesthetics Clinic | Modern Aestheticss Dermatology & Laser Clinic | Athena Dermatology Clinic |
| 6 | Five Cosmetic | — | Premium Cosmetic Laser Center |
Biolite was first every time. Athena was the only other clinic in all three runs. Ten different clinics filled 17 slots.
The engines don’t all vary the same amount:
| Engine | Category | Same #1 in all three runs? | In every run | Different businesses named |
|---|---|---|---|---|
| ChatGPT | Clinics | Yes (Biolite) | 2 | 10 |
| ChatGPT | Manufacturers | No (Al Shirawi twice, SIGMA once) | 4 | 9 |
| Perplexity | Clinics | No (Hortman once, another clinic twice) | 5 | 14 |
| Perplexity | Manufacturers | Yes (EGA) | 7 | 11 |
| Gemini | Clinics | Yes (Hortman) | 2 | 7 |
| Gemini | Manufacturers | Yes (EGA) | 4 | 11 |
Gemini’s clinic answer was the steadiest: Hortman first and The Nova second in every run. Google’s own results page didn’t move at all. The organic results and local pack were identical across three runs; only the AI layer varied.
What happens when one business goes by two names?
The engines split it. One Dubai aesthetic clinic in our results trades under two spellings of its name, which we’ll call Name A and Name B. Each surface settled on a different version.
| Surface | How it named the clinic |
|---|---|
| Google Search | Name A, in the local pack and via its own site in organic results |
| Perplexity | Both names in one entry, ranked first in two of three runs |
| Gemini | Name B only, in one of three runs |
| ChatGPT | Name B only, as a passing mention in one of three runs |
Perplexity was the only engine to connect the two names, and it ranked the clinic first. Google shows one name, the other two AI engines show the other.
A business with inconsistent naming gives each engine a chance to treat it as someone else, or to pick up only part of its reputation. The fix is practical: one canonical name across the website, Google Business Profile and directory listings, with the alternate spelling declared in structured data so every engine resolves to the same record.
Where do the engines agree, and why?
Where all three read the same kind of authority. Universities had the most overlap, with four institutions on all three engines, and it’s the category where independent rankings (Times Higher Education, QS, uniRank) are everywhere in the sources. B2B SaaS came next, with trade coverage like eChai Ventures shared across engines.
Car dealerships showed a different pattern. ChatGPT and Perplexity shared one dealer, Gargash. Gemini shared six with ChatGPT (including ALBA CARS, The Elite Cars and the franchise groups) and two with Perplexity (Al Tayer Motors, F1rst Motors). Gemini sat between the two engines that otherwise barely overlapped.
Google’s AI Overview mostly picked from the same consensus names:
| Category | Named in the AI Overview | Also named by |
|---|---|---|
| Industrial manufacturers | Ducab, EGA | All three engines |
| Industrial manufacturers | NAFFCO | Gemini |
| Universities | Canadian University Dubai | All three engines |
| Universities | University of Birmingham Dubai | Perplexity, Gemini |
| Law firms | Al Tamimi & Company, Hadef & Partners | ChatGPT, Gemini |
| Law firms | Clyde & Co | Perplexity, Gemini |
| B2B SaaS | Dukkantek (DTEK) | All three engines |
In manufacturers, universities and law firms, every business the AI Overview named was on at least one of the three engines, and most were on two or more. B2B SaaS is the partial exception: of the three companies its AI Overview led with, only Dukkantek made all three engines. Nazm.ae and Verofax did not.
What does a single-engine, single-run score miss?
Most of the picture. A score taken from one engine once can’t show:
| What it misses | Evidence from this test |
|---|---|
| The other engines | 10 of 147 businesses named appeared on all three |
| The next run | ChatGPT named 10 different clinics across three runs of the same prompt |
| Which source to fix | Perplexity reads editorial, ChatGPT reads listings |
| Naming splits | One clinic appeared under one spelling on Google and another on Gemini and ChatGPT |
Even the most anchored category reorders. ChatGPT’s law firm answer was grounded entirely on Legal 500 in both our August scan and in September. Between the two, A&O Shearman went from first to absent, Dentons and Norton Rose Fulbright dropped out, King & Spalding came in, and Al Tamimi moved from fourth to first. A directory holds the source steady. It doesn’t hold the ranking steady.
Methodology
- Prompt: “Best [category] in Dubai”, one prompt shape, English only.
- Categories: law firms, B2B SaaS, aesthetic clinics, boutique hotels, industrial manufacturers, universities, car dealerships.
- Surfaces: ChatGPT, Perplexity, Gemini, Google Search.
- Conditions: data collected 10 September 2026. Four surfaces, logged out, clean profile, geo verified to the United Arab Emirates, free or default model tier. Fresh session per cell. Signed-out Gemini served Flash-Lite.
- Volume: 44 answers. Clinics and manufacturers ran three times on every surface (n=3); the other five categories ran once (n=1). Single-run categories may overlap more or less on a repeat, and the n=3 categories have larger pooled lists, which lowers their overlap share.
- Overlap counting: every business recommended, including “also consider” lists and separate tiers; passing mentions in caveats excluded; Perplexity counted from its written answer, not its Places module. One Perplexity hotel entry was truncated in our log, so hotel overlap may be slightly understated.
- Market: Dubai only.
- Date: re-run logged 10 September 2026. The law firm comparison uses the ChatGPT cells from our August scan, which were logged out on the same conditions as September.
- What this doesn’t cover: signed-in or paid tiers, other phrasings, Arabic queries, Claude and Copilot.
- Screenshots: recaptured after the logged run. Answers shift between runs, so every number here comes from the run log.
See your own spread across the engines. Our free AI visibility report shows where your business appears across the AI engines, how you compare with three competitors, and which sources are recommending them instead of you. Delivered on a 15-minute call.
Frequently asked questions
Do ChatGPT, Perplexity and Gemini recommend the same businesses in Dubai?
Rarely. Across seven categories they named 147 businesses between them and 10 appeared on all three. In law firms, aesthetic clinics and car dealerships, no business appeared on all three.
Does ChatGPT give the same answer if you ask twice?
At the top, often yes. Below it, no. Across three runs of the same clinic question, ChatGPT put Biolite Clinic first each time but named ten different clinics in total.
Why do Perplexity and ChatGPT give different answers?
They read different sources. Perplexity mostly grounds on editorial coverage such as GRAZIA and What's On; ChatGPT mostly grounds on listings and directories such as Bayut and Doctify.
Which Dubai categories do AI engines agree on most?
Universities, where four institutions appeared on all three engines, followed by B2B SaaS with three. Both have strong third-party rankings or trade coverage that all three engines read.
About the author
Joe Hickman is the founder of Brandure, a Dubai agency focused on answer engine optimisation: how businesses get named in ChatGPT, Gemini, Perplexity and Google’s AI Overviews. He is also a Client Partner at Snapchat MENA and has a background in paid social across Meta, Google, TikTok and LinkedIn. More about Joe. Brandure’sfree AI visibility report shows where a business currently appears across the engines.