How to check if ChatGPT recommends your business

To check if ChatGPT recommends your business, write 10 questions your customers would ask, put each one to ChatGPT, Gemini, Claude, Perplexity, Google AI Overviews and Grok in a clean session, and log four things per answer: were you named, in what position, which source was cited, and which competitor appeared. It takes about 30 minutes. Repeat it every month.

You don't need a tool or a budget for the first pass. You need a list of questions, six browser tabs and a table. The table is below.

How do I write the 10 questions?

Write them the way a customer types or speaks, not the way your website describes you. A buyer never asks for "premium dental services". They ask "best dentist near JLT" or "who does teeth whitening in Dubai Marina, price".

Split the 10 into three groups:

1. Four "best X in place" questions: "best dentist near JLT", "best immigration lawyer in Business Bay", "WhatsApp chatbot agency Dubai".
2. Three comparison or price questions: "how much does a Golden Visa lawyer cost in Dubai", "Property Finder vs Bayut for agents".
3. Three problem questions: "how do I stop missing WhatsApp enquiries after hours", "who can build an AI receptionist for a clinic in the UAE".

Add the Arabic version of two of them if your buyers use Arabic. A buyer in Sharjah may ask in Arabic and still call you in English, and engines often answer the two languages from different sources.

Don't put your brand name in any question. A question that already says "PHII Labs" tests whether the engine knows you exist. It does not test whether the engine recommends you to a stranger.

How do I run them so the results are fair?

Use a clean session for every engine, so your own history doesn't flatter you. Log out, or use a private window, or use the temporary or incognito chat mode where the engine offers one. Set your location to the city the customer is in, or add "in Dubai" to the question itself.

Run the same 10 questions in all six places:

• ChatGPT
• Gemini
• Claude
• Perplexity
• Google AI Overviews (a normal Google search; the overview appears only for some queries)
• Grok

Copy the question exactly each time. Paste, don't retype. Take a screenshot of each answer, because answers change and you will want proof later.

Google AI Overviews won't show up for every question. Google says in its guide to AI features that these features appear only when its systems judge them additive, so a blank is a result too. Mark it "no overview" and move on.

What do I record for each answer?

Record four facts and one note. The table below is the sheet we use. Copy it into a spreadsheet, one row per question and engine, which gives you 60 rows.

Field	What to write	Example
Question	Exact text, numbered 1 to 10	Q3: best dentist near JLT
Engine	One of the six	Perplexity
Mentioned	Yes / No / Named but wrongly described	Yes
Position	1st, 2nd, 3rd or "in list of 5+", or "n/a"	3rd of 5
Cited source	The page the engine linked or named for you	Your Google Business Profile, a directory listing
Competitor named	Every other business in the answer	Clinic A, Clinic B
Note	Wrong phone, old price, wrong district	Lists a closed branch

Then add up. Count the rows with "Yes" and divide by 60. That is your mention rate. Count how often each competitor appears. The competitor who shows up most is the one the engines currently prefer in your market.

The "cited source" column is the most useful one. If the engines name you but always cite a directory and never your own site, your site isn't the source of truth yet. That points to a specific fix.

When a business tells us "ChatGPT has never heard of us", the sheet usually shows something narrower. It names them in Perplexity because of one directory page, and skips them everywhere else because the website doesn't answer the question in plain words.

Why do the answers change between runs?

Answers change because language models generate text each time, and because the engines search the live web before they answer. Ask the same question twice and you can get a different list of names.

A single run is therefore a sample, not a verdict. Two habits make the sample useful:

1. Run each question twice on different days if you can, and count "named in at least one of two" as a mention.
2. Look for patterns across 60 rows rather than one answer. A business named in 3 of 60 is invisible. A business named in 25 of 60 is not, even if question 7 missed it.

Nobody can promise that an engine will name you on a given day. The same goes for a vendor who says otherwise. What you can do is measure the rate and watch it move.

How often should I repeat the check?

Repeat it monthly, with the same 10 questions. Changing the questions each time makes the months incomparable. Keep a column for the date, and keep the screenshots in a folder per month.

A monthly rhythm also catches the cases where a competitor moves in. If a clinic opens a new page about teeth whitening prices and appears in four answers next month, you will see it in the sheet before you see it in your enquiries.

Tracking the leads that follow is a separate job. Our post on tracking leads from AI search covers it. This article only answers whether you are being recommended right now.

What if I'm not being recommended?

Fix the cited-source column first. Look at who the engines cite for your competitors and what those pages have in common. Usually it is a page that answers the exact question in the first two sentences, with a real price or a real number, and a company name spelled the same everywhere. When the cited source is a Maps listing, our guide to ranking a Google Business Profile in Dubai shows how to get yours into the same shape.

Then work through three steps:

1. Make the pages the engines read. Your service pages should answer the question in the first 40 to 60 words. Our guide on how to get cited by ChatGPT and Perplexity shows how.
2. Check the basics behind it: indexing, crawler access, structured data, consistent name and address. The AI search visibility audit checklist lists every item.
3. Re-run the sheet after four to six weeks and compare the mention rate.

On a Dubai property platform built on a fresh domain, the same approach (entity-structured pages, answer-first articles, FAQ markup, JSON-LD) produced 1,182 AI citations in three months. That is one project, in one market. It shows the method can work. It is not a forecast for your business.

If you want the full program behind this, see our AI search visibility service.

Can someone run this check for me?

Yes, and we do it for free. The AI test does what this article describes, with 10 buyer questions put to 6 engines (ChatGPT, Gemini, Claude, Perplexity, Google AI Overviews and Grok). You get the answers and the fixes on WhatsApp within 48 hours.

The DIY version is enough if you have 30 minutes and want to learn how your market looks. The AI test saves the time, uses a question set we have refined across UAE sectors, and comes with a fix list. Both give you the same kind of evidence, a table of who gets named and why.

The takeaway

Ten real customer questions, six engines, a clean session and one table tell you whether ChatGPT recommends your business. Count your mention rate, note the cited sources, and repeat monthly so you can see change. If the number is low, start with the pages the engines cite for your competitors.

Get the free AI test and we'll run the 10 questions across the 6 engines for you and send the results on WhatsApp.