A do-it-yourself AI visibility audit for Dubai businesses: a clean ChatGPT test setup, UAE buyer prompts, a run count backed by research, a scoring sheet and the crawler check that decides whether you can appear at all.
Most Dubai business owners have tried it once. They open ChatGPT, type "best [my service] in Dubai", and scan the answer for their own name. That single test tells you very little. An AI visibility audit swaps the one-off glance for a repeatable check you can run yourself in an afternoon, without an agency or a paid tool.
The afternoon is worth it. Microsoft's report on AI diffusion in early 2026 ranks the UAE first in the world, with 70.1% of people aged 15 to 64 using generative AI products, against 17.8% worldwide. Your buyers are already asking these tools for recommendations. This guide shows you how to find out what they hear.
What an AI visibility audit measures for a Dubai business
An AI visibility audit measures how often an AI assistant names your business when a buyer asks a question you should win. It also checks how accurately the assistant describes you, which competitors it names instead, and which websites the answers lean on. For a Dubai business, that means testing the prompts local buyers actually type.
Four things come out of a proper audit:
- Mention rate: the share of test runs where your business is named, without you putting the name in the prompt.
- Accuracy: whether what the assistant says about your services, location and prices matches reality.
- Competitor share: which rivals appear, and how often.
- Source list: the websites the answers link to.
Rank position is missing on purpose. Step 3 explains why.
Step 1: Test in a clean ChatGPT session
Run every test in a Temporary Chat with personalization turned off. That way ChatGPT is not drawing on your saved memories or custom instructions, which may already mention your own company. A session that knows you run a gym in Dubai Marina will not show you what a stranger in Deira sees when they ask.
OpenAI's help centre describes the setup: open a new chat and click the Temporary button in the top-right corner of the page. A non-personalized Temporary Chat does not use your memory, custom instructions or plugins, and it does not create new memories. You choose personalization when you start the chat and cannot change it afterwards, so set it before typing. The no-new-memories point matters for an audit, because it stops one test run from colouring the next.
Open a fresh Temporary Chat for every run, not just every prompt.
Should you also ask ChatGPT about your business by name? Yes, but record it separately. A prompt like "What do you know about [your company] in Dubai?" tests accuracy. It cannot test visibility, because you have already handed over the name. Keep brand prompts and discovery prompts in different parts of your sheet.
Step 2: Write prompts the way Dubai buyers ask
Write 8 to 12 prompts in the words a real buyer would use. Include the Dubai area, the service, and the detail that matters to that buyer, such as a budget in AED, a language or a deadline. Mix the prompt types below, because each one tests a different moment in the buying decision.
| Prompt type | Example for a Dubai business | What a mention tells you |
|---|---|---|
| Category | Best physiotherapy clinic in Jumeirah Lake Towers | You are in the default shortlist for your service |
| Problem | The AC in my Business Bay office leaks every summer, who can fix it properly? | The assistant links you to the buyer's problem, not only your category |
| Budget | Affordable accountant for a small Dubai free zone company, with fees in AED | You are seen as a fit for that price range |
| Area | Which interior designers work on villas in Arabian Ranches? | You are tied to the neighbourhoods you actually serve |
| Language | Your category prompt, written in Arabic | Whether you exist for buyers who ask in Arabic |
Leave your brand name out of all of these. The moment a prompt contains your name, it stops measuring visibility.
Do not skip the Arabic row if your customers speak Arabic, and do not stop at ChatGPT. Google launched AI Mode in English across the Middle East and North Africa in August 2025, then added Modern Standard Arabic that October. Running the same prompts in Google's AI Mode gives you a second view of the same buyers.
Step 3: Run every prompt several times
Run each discovery prompt 10 times, every time in a fresh Temporary Chat, and count how often you are named. A single run tells you almost nothing, because AI assistants rarely return the same list twice. Ten runs also keeps the maths simple: each appearance adds 10 percentage points to that prompt's mention rate.
The research behind this is recent. SparkToro and Gumshoe.ai ran a study, reported by Search Engine Land in January 2026, in which 600 volunteers put 12 prompts through ChatGPT, Claude and Google's AI nearly 3,000 times. The odds of getting the same list of recommendations twice were under 1 in 100. The odds of the same list in the same order were closer to 1 in 1,000.
That is why this audit ignores rank. If the order reshuffles on almost every answer, "we came third" is noise. What held steady was presence: in some categories, the same names showed up in 60% to 90% of responses for a given intent. The takeaway in Search Engine Land's write-up is simple: "Track how often your brand appears across many prompts, run many times."
Answers also drift over longer periods as models are updated, which we cover in how often AI models change their answers. So one audit is a snapshot. Repeat it on a fixed schedule with the same prompts.
Step 4: Score results with an audit sheet
Score each prompt with one formula: mention rate equals the runs where you were named, divided by total runs. Then log every competitor named and every source linked in each answer. Those two columns matter as much as your own score, because they show who is winning the recommendation and where it came from.
A plain spreadsheet with these columns is enough:
- Prompt and prompt type
- Runs completed
- Runs where you were named
- Mention rate
- Competitors named, with a tally for each
- Sources linked in the answer
- Anything the assistant got wrong about you
Average the mention rate across your discovery prompts and you have one audit score. Put the competitor with the highest tally next to it, so the score has a benchmark rather than floating on its own.
Why does ChatGPT keep naming a competitor instead of you? Your sheet is the place to start looking. Filter for runs where a rival was named and you were not, then read the sources column for those runs. If the same directory, review site or news article keeps turning up and your business is not on it, that page is a concrete target.
Column 7 is a separate job. Wrong facts, such as an old address or a service you stopped offering, need fixing on the pages that carry them, not just on your own website.
Step 5: Check that ChatGPT's search crawler can reach your site
Open your-domain/robots.txt and check that OAI-SearchBot is not blocked. OpenAI's crawler documentation says sites that block OAI-SearchBot will not be shown in ChatGPT search answers. GPTBot is a separate crawler that OpenAI uses for model training. The one that governs ChatGPT search is OAI-SearchBot.
Do this before you rewrite any content. If the crawler is blocked, your pages cannot be shown in ChatGPT search answers, however good they are. Look for a User-agent: OAI-SearchBot line followed by Disallow: /. If you find one and want to appear in ChatGPT search, remove the block or ask your web developer to.
OpenAI notes that it can take about 24 hours after a robots.txt change for its systems to adjust. Wait at least that long before you rerun your audit.
What to do with your audit score
Treat your first audit score as a baseline, not a verdict. Rerun the same prompts on a fixed schedule, compare your mention rate with the competitor who appears most often, and work through the sources that keep appearing in answers where you are missing. Those sources are your clearest list of places to earn a mention.
If you want a lighter version to share with your team first, our five-minute manual check covers the basics. For choosing which numbers to report over time, read how to measure AI visibility when there is no rank one.
Running 80 or more tests by hand takes real time. If you would rather start with an automated baseline, GEOscanAI's free AI visibility audit sends five of your keywords to ChatGPT, Claude and Perplexity and reports whether your brand appears. You get one free audit per email, and your manual sheet becomes the way to check the result.