Four of these six questions don't have your name in them

When we built the question generator inside AskScope, we expected the hard part to be coming up with enough questions. It wasn't. The buyer questions that decide whether an AI assistant recommends a business turn out to be a short, boring, finite list — six shapes, and you can write them all down on an index card.

Here they are, exactly as the product generates them, with your category and competitors filled in:

• What is the best {category} in {city}? — someone with a need and no shortlist yet.
• Which {category} should I choose for a small business? — the same person, closer to buying.
• Best {category} near {city} — the local version, which behaves differently enough to be its own question.
• What tool or service should I use for {problem}? — the person who hasn't even named the category yet.
• {You} vs {competitor} — which is better? — the comparison.
• What are the best alternatives to {competitor}? — the switcher.

Six shapes. Two of them contain your brand name. The other four are being answered about your category, on your behalf, without you in the sentence — and those are the ones that matter most.

Brand monitoring finds the questions you've already won

If you track anything today, it's probably your name: mentions, alerts, maybe a search for your brand in an AI assistant to see what it says. That's worth doing, and it will systematically miss the problem.

A question with your name in it is a question where you're already in the buyer's head. They know you exist; they're deciding between you and someone else. That's a good position to be in — it's the end of the funnel, and it's the part you can influence with a comparison page.

The unbranded questions are the ones that decide whether you ever reach that stage. "What's the best bookkeeping service in Austin" gets answered with three names. If you're not one of them, no alert fires anywhere. Nothing shows up in your analytics, because there's no click and no referrer — the buyer never left the chat window.

“You lose those silently, which is a different and worse kind of losing than a lost deal.”

That's the actual gap. Not "AI is talking about you and you can't see it" — more often, AI is answering questions in your category and you aren't in them at all.

When you only get five, which five?

The generator orders the questions deliberately, because most scans don't run all of them. A free scan asks five. When we had to choose, the ranking came out like this: discovery first, then the purchase-intent version, then local if the business has a city, then the problem-shaped question, and only after all of those, the comparisons.

That ordering is the opposite of what a founder wants. Everybody wants to see "me vs my competitor" first — it's the most emotionally satisfying question and the easiest to argue with. But it's also the question you're most likely to already appear in, because it contains your name; the model has been handed you as an option. The unbranded discovery question is where you find out whether you're an option at all.

There's a second reason, less noble: comparison questions multiply. Each competitor generates a comparison and an alternatives question, so a business that lists five rivals could fill an entire scan with nothing but questions about rivals and never ask the one that matters. The generator interleaves them so the first competitors get covered before the limit cuts in, and puts the unbranded questions ahead of all of them.

What this list isn't

It's a taxonomy, not data.

Nobody publishes what people actually type into ChatGPT. There's no keyword tool for assistant prompts, no search-volume column, no way to know that "best bookkeeping service in Austin" gets asked four hundred times a week and the alternatives question gets asked twice. Anyone who shows you AI prompt volume for your category is showing you a model of their own.

So these six shapes are a considered guess about how people ask for a recommendation, generated deterministically from your category and competitor names. They're the shapes we'd ask in if we were buying, and they cover the intents that map to real purchase moments — discovery, purchase, local, problem, comparison, switching. That's a defensible basis for a measurement. It is not the same thing as knowing the distribution, and we'd rather say so than dress a template up as intelligence.

Run them yourself this week

You don't need a product for the first pass. Open an assistant you'd expect your buyers to use, ask the four unbranded questions with your category filled in, and score each answer as one of four outcomes:

• Recommended — you're named as a suggestion.
• Mentioned — you appear, but as background, not a pick.
• Competitor won — a rival is named and you aren't.
• Ignored — nobody in the answer, including you.

That vocabulary is what AskScope stores per answer, and it's useful on paper too, because it separates those two cases the moment you write them down.

Two things will make the exercise more honest. Ask each question a few times rather than once, because you'll get different answers and the variation is the point. And check where the answer's citations point, if the assistant shows them — the sources it leans on for your category are a more actionable list than the answer itself, since "get listed on the thing it already trusts" is a task you can actually do.

Then do it again next month. That's the part we automated, and the honest reason we did: the measurement isn't secret, it's just tedious to repeat across several assistants while keeping the old answers around so you can prove what changed. If you'd rather do it by hand with a spreadsheet and twenty minutes a month, that genuinely works.