What prompt volume means in AI visibility tools, how vendors estimate it, why the estimate is not a search volume, and how Rightcited chooses the questions to check without it.
Prompt volume is a number some AI visibility tools attach to a tracked prompt to estimate how often people ask an assistant something like it. It is the AI-answer analogue of keyword search volume, and it is a much softer number, because no assistant publishes how often any question is asked. This page explains what the vendors say they measure, what the number can and cannot tell you, and how we choose questions without leaning on it.
Read from peec.ai/pricing, ahrefs.com/brand-radar and tryprofound.com/pricing on October 8, 2026.
Relative demand across the topics in a category: that comparison questions about one rival are asked more than about another, that a pricing question outweighs a feature question. Modeled from search demand, as Ahrefs does it, the number carries the same strengths and weaknesses as keyword volume: strong for established categories with searchable phrasing, weak for new products, local services and the long, specific questions buyers put to an assistant and would never type into a search box.
How often any assistant was actually asked your question, because none of them publishes it. Whether the people asking are buyers. Whether the answer named you, which is the number that matters and which only the answers can give. A prompt with a high volume and an answer that names you first is not a problem; a prompt with no volume estimate and an answer that sends your buyer to a rival is.
The 50 questions in a plan are chosen with the client in onboarding from five families, in the buyer's words: category questions (best X for Y), product and feature questions, comparison and alternatives questions for the rivals the answers already name, the trust question (is X legit, X reviews), and the local or market variants. The free scan runs six to eight of them first. Where a volume estimate exists we read it as a tie-breaker between candidate questions; the ordering of the work comes from the answers: the questions where the brand is absent, where a rival is named first, and where a record is wrong, in that order.
A leather goods store's question set does not need a volume estimate to be ordered. The four category questions where it was absent on all three assistants (slim wallets, women's crossbody bags, full-grain brands, MagSafe wallets on two of three) come first; the brand question where three assistants could not find where the goods are made comes next because the answer is being written for the brand by whichever record the assistant found; the alternatives question where every assistant pivots on the one-year warranty page comes third. Volume would not change that order, and the weekly check tells us whether each page moved its answer.
No. Search volume is counted from search logs; prompt volume is estimated, usually by modeling from search demand or scoring relative topic demand. Treat it as a relative signal, not a count.
No. We report what the answers did: named, named first, cited, sentiment and sources, each a count with a denominator and the answers behind it. Where a client's tool reports volume we read it beside those counts as a tie-breaker.
From the category: the questions the assistants already answer with shortlists, the comparison and alternatives questions for the rivals those shortlists name, and the trust and market questions, chosen with the client in onboarding and revised when the answers show a question we missed.
Or start smaller: the AI Page Check runs one page through all four assistants for $2.99.