To choose the prompts to track for AI visibility, collect the full questions your buyers actually ask, cover every stage from discovery to comparison, group them into subcategories, and keep your own brand name out of most of them. A prompt is a complete question in a buyer's words, and the set you choose decides everything you measure afterward.
Good prompts make the numbers meaningful. Weak prompts produce a dashboard full of confident figures about questions nobody asks.
What is a prompt in AEO?
A prompt is the question you put to an AI engine, tracked over time. It is the answer engine optimization version of a keyword, but longer and more specific. "Payroll software" is a keyword. "What payroll software handles tipped staff in a restaurant?" is a prompt.
Because engines answer questions in full sentences, buyers have started asking them that way. The AI search glossary defines prompt, subcategory and the other terms used here.
Where do the prompts to track come from?
Start with real language, not guesses.
- Sales calls and demos. The first question a prospect asks is often a prompt.
- Support tickets and onboarding questions. They show what customers did not understand.
- Reviews and forum threads. Buyers describe their problems in their own words.
- Search Console queries. Rewrite your best commercial queries as full questions.
- Competitor pages. Note the comparisons and use cases they build pages around.
Write each prompt the way a person would type it into ChatGPT, not the way a marketer would title a page.
Which stages of the buying journey should prompts cover?
Mix three kinds so you do not only measure people who already know you:
- Discovery: "best project management software for a small agency"
- Comparison: "Brand A vs Brand B for a small agency"
- Brand checks: "is Brand A worth the price?"
Discovery prompts show whether you are in the consideration set. Comparison prompts show how you stack up against specific rivals. Brand checks show what the engine says when someone asks about you directly, which is where wrong or outdated claims usually surface.
Should prompts include your brand name?
Leave your name out of most of them. A discovery prompt that names you only measures buyers who already know you. Keep a smaller group of brand-name prompts to monitor how you are described, and keep the two groups separate so the numbers stay interpretable.
How do you group prompts into subcategories?
Group related prompts under a subcategory such as "pricing and value," "multi-site groups" or "sensitive skin." Averages across every prompt blur where you win and lose. With subcategories you can see that you lead one topic and are invisible in another, and then decide where to work. AI visibility metrics explains why breaking numbers down by subcategory beats a single average.
Why should prompts cover sub-questions?
Google says AI Overviews and AI Mode may use a "query fan-out" technique, issuing multiple related searches across subtopics and data sources to build an answer (Google Search Central). A single buyer question can therefore draw on several narrower ones. To reflect that, include prompts for the sub-questions around your main topics: price, alternatives, who it suits, what it integrates with and common problems. The difference between the two Google features is covered in Google AI Overviews vs AI Mode.
How many prompts do you need?
There is no universal number. A reasonable approach:
- Start focused. A few dozen well-chosen prompts across your main subcategories is enough to find the first patterns.
- Grow where it pays. Add prompts for subcategories you are working on, or for topics that sales hears about often.
- Retire overlap. Two prompts that ask the same thing in slightly different words count your result twice. If you want to measure the effect of wording on purpose, do it deliberately with a wording test. See Why AI answers change with wording.
What mistakes should you avoid?
- Writing prompts like keywords. Use full questions.
- Only branded prompts. You will see good numbers and learn nothing.
- Too many near-duplicates. They inflate the sample without adding information.
- Changing the set constantly. Add and retire prompts on purpose, and note when you do, so trends stay comparable.
- Ignoring location and language. A question asked in one market may get different answers elsewhere, so track the markets you sell in.
How does Verte choose and manage prompts?
Verte writes the prompts for your business from its name and website, grouped into subcategories, and you can edit them. For businesses, it researches your company, suggests goals and then proposes the subcategories to track before it writes the prompts. For agencies, it does the same for each client, with prompts shared from one pool across clients. The Verte Engine can also propose new prompts or removal of overlapping ones. Nothing changes until you confirm.
The Platform page explains subcategories, and the business and agency pages show the setup flow.
Start tracking your business or start as an agency. See current plans.