Before you change anything on your website, find out where you stand. This check takes about ten minutes, costs nothing, and gives you a clearer picture than most dashboards. It is the same starting point I use with clients, just in a lighter form.
Step 1: Write ten real questions
Write the questions your customers actually type or say, not the keywords you target. Mix four types:
- Category questions: "What is the best accounting software for small businesses in Saudi Arabia?"
- Comparison questions: "Brand A vs Brand B" and "alternatives to Brand A".
- Problem questions: "How do I pay remote employees in several countries?"
- Brand questions: "What is [your brand]?" and "Is [your brand] good for ...?"
If your market is bilingual, write some of the questions in Arabic and some in English. The answers can differ a lot between languages.
Step 2: Ask four AI tools
Run every question in ChatGPT, Gemini, Perplexity and Google AI Mode. Use a fresh conversation each time, and ideally a logged-out or private window, so earlier chats do not shape the answer.
Step 3: Record what you see
| Column | What to write |
|---|---|
| Mentioned? | Yes or no |
| Position | First, among several, or last |
| Description | How the tool describes you, word for word |
| Accuracy | Is the category, the location and the offer correct? |
| Sources | Which pages or sites it cites |
| Competitors | Who is mentioned instead of you |
Step 4: Read the pattern
- Not mentioned at all. The engines may not know you or may not see you as relevant to that question. Look at who is mentioned and what their pages do differently.
- Mentioned, but described wrongly. This is an entity problem. It is exactly what happened to Forlanso, which AI tools grouped with freelance marketplaces. I wrote about how that gets fixed.
- Mentioned only in one language or one tool. Your content or your sources are stronger in some places than others.
- Cited from other sites, not yours. Third-party pages describe you. Check that they say the right thing.
What this check can and cannot tell you
AI answers change from one run to the next, and they can differ by country and by user. Treat the result as a snapshot and a direction, not a statistic. Run important questions two or three times, and repeat the whole check once a month so you can see the trend.
The same method led to the Qoyod example in my work: asked in Arabic for the best Saudi websites offering free accounting templates, ChatGPT put Qoyod first. You can read it in the Qoyod article or the full case study.
What to do with the results
- Pick the three questions that matter most for revenue.
- Write one fixed, two-sentence definition of your brand and use it everywhere.
- Rewrite the pages that should answer those questions so each section starts with the answer. See answer-first writing.
- Re-run the check in four weeks.
Frequently asked questions
How many questions do I need?
Ten is enough to see a pattern. Twenty to thirty gives you a more reliable picture, especially if you cover several products or markets.
Do I need a paid tool to track AI visibility?
No. A manual check with a spreadsheet is a good starting point. Paid tools can help later, when you need to track many questions over time.
Why does each AI tool give a different answer?
They use different models, different sources and different ways of searching the web, and their answers also vary from run to run. That is why the check uses several tools and repeats the important questions.
Do you know how AI answers your customers' questions today?
I test your customers' real questions on ChatGPT, Gemini and Google AI Mode, and show you where your brand appears and what it takes to become the answer.