Write the questions your buyers ask ChatGPT, Claude and Gemini before they ever talk to sales. Grouped by intent, with your brand left out on purpose.
The question that matters is the one a buyer asks before they know your name: “what’s the best expense tool for a 200-person company?” If the assistant’s shortlist leaves you out, the buyer never searches for you. So every question here leaves your brand out, and the groups follow how buyers actually narrow a choice.
Our guide to the buyer questions AI assistants get asked goes further into phrasing.
One answer is a sample, not a measurement: ask Claude the same question twice and you can get two different shortlists. What tells you something is the same questions, asked the same way, over time. That’s what a Mentioned tracked keyword does. Its five questions are locked when you create it and asked on every run, so when your number moves, the answers changed and not the questions. A free account tracks 5 queries weekly on Claude and Gemini.
Because buyers don’t, until they’ve heard of you. A question with your name in it measures whether the assistant knows you exist. A question without it measures whether the assistant recommends you, which is the thing that wins deals.
Start with five that map to real pipeline: the shortlist question for your core category, one or two problem-first questions, and the alternatives question for the competitor you lose to most. Answers vary from run to run, so asking the same five every week tells you more than asking fifty once.
All three if your buyers use all three, and most B2B buyers use more than one. A brand can lead on one assistant and be missing from another, so a question tracked on one model only tells you about that model.
From patterns, not from any assistant’s logs. Nobody outside OpenAI, Anthropic or Google sees what people actually type. The groups follow how buyers research: ask for a shortlist, describe the problem, compare names they know, then check price and fit.