Enough to be representative, not a round number for show. A solid audit covers the buyer journey from early research to decision, runs across all three engines, and includes the real questions your customers ask. That usually means dozens of carefully built prompts, not thousands of shallow ones. Coverage and relevance beat raw volume every time.
The honest answer is that prompt count is the wrong thing to fixate on. People ask for it because it feels measurable, but a big number is easy to fake and easy to waste. What actually determines whether an audit is trustworthy is coverage: does the prompt set represent how real buyers ask AI engines about your market, across the whole journey?
A representative set spans the buyer journey. Early-stage prompts are broad and unbranded, the kind someone types when they barely know the category exists. Mid-stage prompts compare options and weigh features. Late-stage prompts are close to a decision and often name brands or specific needs. If an audit only tests one of these, it gives you a distorted picture. You might look strong on decision-stage queries and invisible on the research questions that feed the top of your funnel.
It also has to span the engines. ChatGPT, Google AI Overviews and Perplexity do not answer the same question the same way. They draw on different sources, weight them differently and phrase recommendations differently. A prompt set that only runs on one engine is not an AI visibility audit, it is a single-engine snapshot. The same prompts have to run across all three to be comparable.
So what does that mean in numbers? For most businesses, a genuinely representative audit lands in the range of tens of well-constructed prompts per engine, multiplied across the journey stages and any distinct products, personas or markets you serve. A small, focused business might need a few dozen. A business with several product lines, personas and geographies needs more, because each adds a slice of buyer questions that must be covered. The driver is the shape of your market, not a target on a slide.
Beware two failure modes. Too few prompts and the result is noise: one or two unlucky answers swing your whole picture. Too many shallow or duplicated prompts and you pay for volume that adds no new signal, while burying the findings that matter. The sweet spot is enough distinct, relevant questions to be stable, and no more.
When you commission or buy an audit, ask how the prompts were chosen, not just how many there are. A provider who can explain the journey stages, the personas and the engines they covered is giving you something defensible. A provider who leads with "we tested 10,000 prompts" and cannot tell you why those prompts is selling a vanity figure.
The right prompt set is also the foundation for tracking over time. If your prompts are representative and stable, you can rerun them each cycle and trust that movement in your results reflects real change, not a reshuffled question list. That repeatability is what turns a one-off audit into a measurement programme.
We design prompt sets around your actual buyer journey, products, personas and markets, then weight them so the questions that matter most to your funnel carry the most weight. Every prompt runs across all three engines, ChatGPT, Google AI Overviews and Perplexity, so the AI Visibility Score reflects how AI answers as a whole, not one tool. You can see the framework behind this on our methodology page.
Your free Exec Brief uses a focused, representative set to give you a reliable first read. The ongoing cadences, which are bespoke priced, expand and maintain that set so you can track movement cycle over cycle with confidence. To get your starting set built around your market, begin at /start. Be Known. Be Cited.
Book an AVS Exec Brief: a real, one-off measurement of how ChatGPT, Google AI Overviews and Perplexity talk about your business right now. Same methodology as the full AI Visibility Strategy, delivered manually, free of charge. Annual, Bi-Annual and Quarterly cadences are bespoke priced.
Book your AVS Exec Brief →