Why it matters
Assistants answer full questions, not keywords, and you can only measure or improve your presence in answers to questions you have written down. Keyword tools report what people type into a search box, not what they ask an assistant, so the list has to be built from buyer intent rather than copied from a keyword report.
Google says appearing in AI Overviews and AI Mode takes the same fundamentals as Search, no special trick. The work, then, is knowing which questions matter and whether you show up when they are asked.
The workflow
Start from what you sell and to whom
Write one paragraph: the product, the customer, the market, the problem it solves. Every question on the list should be one that customer could plausibly ask an assistant before buying.
Walk the buying journey by intent
Cover each kind of question a buyer asks on the way to a decision: how-to, discovery, suitability, trust, local, comparison, alternatives, pricing and the final decision. Two or three questions per intent, about 60 intents in all, is a workable list for one market.
Phrase each question the way the market asks it
Write in Mexican Spanish for Mexico, Brazilian Portuguese for Brazil, Indonesian for Indonesia, with the local currency, the local institutions and the local vocabulary inside the question. Do not translate an English list; a translation keeps the wrong assumptions. Drop any phrasing that names your brand: you want to know whether you appear when the buyer has not asked for you.
Borrow the demand signals you already have
Export queries from Search Console, import a keyword CSV from Keyword Planner, Semrush or Ahrefs, and check Google Trends for the category terms. None of these is AI prompt volume; nobody publishes that. Treat them as a proxy for how common the underlying need is.
Score and trim
Score each question on commercial intent, the demand proxy, how far you are from appearing in the answer and how relevant it is to your brand. Keep the ones worth measuring every week and park the rest.
Let VozRadar draft the list, then review it
In VozRadar, onboarding crawls your site, drafts the candidates and scores them. Connect Search Console or import a keyword CSV to feed the demand proxy, approve each prompt or stop tracking it, then start the baseline measurement.
Run it in VozRadar
Questions
- Crawls your site (robots.txt respected), extracts a brand profile and drafts about 60 intents with 2–3 phrasings each, written natively for the market
- Drops phrasings that name your brand and merges near-duplicates before you review
- Scores every prompt on commercial intent, demand proxy, visibility gap and brand relevance, using keyword CSV imports, a Google Trends lookup and Search Console (imported daily, with candidates tagged “from Search Console”)
- For each additional market, “Draft 12 prompts from Mexico” creates same-intent variants of your approved questions, written natively for that market
In the app: Project → Questions. Approve or stop tracking each prompt, then start the baseline. A baseline consumes answers from your plan: prompts × surfaces × repetitions.
See how AI sees your brand.
Start with one market. Expand when you're ready.