The quickest way to understand where a brand stands in AI search is not a keyword list. It is the questions customers actually ask, and a close look at where the answers come from.
The 2026 Semrush AI Visibility Index includes a short planning section that describes this as an afternoon AI visibility audit: ten customer questions, four platforms, and a careful log of the sources. PPC Land summarised it in its coverage of the report. Here is how I would run it, with a few additions for brands selling in India, the Gulf and Southeast Asia.
What you need
- Ten real buyer questions for one product or category.
- Access to ChatGPT, Gemini, Google AI Mode and regular Google Search.
- One spreadsheet with columns for question, platform, date, market, language, brands named, your position in the answer, and every cited URL.
Step 1: pick 10 questions buyers really ask
Pull them from sales calls, support tickets, reviews, WhatsApp chats and the search terms that bring people to your site. Mix question types: best option for a need, comparisons between two brands, price range questions, and problem questions where the buyer does not name a product yet. Avoid questions that already contain your brand name. Those tell you how AI describes you, which is useful, but not whether it finds you.
Step 2: run each question on four platforms
Ask every question on ChatGPT, Gemini and Google AI Mode, and search it on Google to see whether an AI Overview appears. Google says AI Overviews only show when its systems decide they add something, so a missing one is a result worth logging too. RemoteCMO also tracks Perplexity by default, so add it if you can.
Step 3: keep the market and language fixed
Run the whole set for one market and one language at a time. Mention the city or country in the question the way a buyer would. If your buyers ask in Hindi, Arabic, Bahasa Indonesia or Malay, run a second set in that language and keep it in its own tab. Answers can change with the language, so mixing them hides the real picture.
Step 4: log every cited URL
This is the step most teams skip. For each answer, copy every source link, not just whether you were named. The report's core finding is that being mentioned and being cited are different lists, so you need both. Our article on who gets quoted in AI search explains why.
Step 5: sort the sources into three groups
- Your own domains. Your site, and any country sites you run.
- Third parties that mention you. Marketplaces, review sites, comparison pages, news and forums that include your brand.
- Third parties that skip you. Sources that the AI trusts in your category but that do not mention you, or list only competitors.
The third group is usually the most valuable list you will make all quarter. It is the shelf you are missing from.
Step 6: match each gap to the right fix
The report makes a point I agree with: a generic plan to create more content is not yet a plan. Match the fix to what you found.
- Your page exists but is never cited. Check that it can be crawled and indexed first. That is a job for SEO and engineering. Google's own guidance on AI features says a page must be indexed and eligible to show with a snippet to appear as a supporting link, and that no special schema or AI text file is required.
- The facts AI needs are missing or out of date. That is a content job. Put clear, current answers on the page buyers would expect to find them.
- A third-party site describes you wrongly, or leaves you out. That is a communications and partnerships job: corrections, listings, reviews and PR.
One technical warning from the report is worth repeating. A chatbot telling you it can reach a page is not a technical test. Check robots.txt, indexing settings and any server or CDN rules yourself.
Step 7: repeat it on a schedule
AI answers vary from run to run, so one afternoon gives you a baseline, not a verdict. Re-run the same questions on the same day each month, in the same market and language, and look for what keeps recurring. The Semrush study itself reads its results as recurrence across months rather than as trends, which is a good habit for your own data too.
What an afternoon will and will not tell you
It will show you whether AI names you for real buyer questions, which sources it leans on, and which of those sources skip you. It will not give you a statistically solid share of answer, a view across many markets, or a picture of how things change over time. For that you need more questions, repeat runs and careful logging. If you want a second pair of eyes, our AI visibility audit runs this method across markets and languages, and the snapshot is free.