What Is Prompt Volume? (And How to Estimate It Without a Tracking Tool)

"Prompt volume" is the number the AI visibility tools put next to a prompt the way keyword tools put search volume next to a keyword: how many times a month, roughly, people ask ChatGPT or Perplexity or Gemini that question. Profound sells it as Prompt Volumes, Peec and Semrush have their versions, data vendors like DataForSEO sell a modelled AI search volume per keyword, and Ahrefs' Brand Radar estimates impressions in AI answers. The number looks exactly like the search volume you have trusted for fifteen years, and that is the problem with it.

Search volume comes from somewhere. Google publishes ranges in Keyword Planner, clickstream panels fill in the rest, and the tools calibrate against each other. Prompt volume has no equivalent source. None of the assistants publish per-prompt counts, OpenAI has shared only aggregate figures (it said in July 2025 that ChatGPT handles more than 2.5 billion prompts a day), and the prompts themselves are sentences, not two-word keywords, so there are thousands of ways to ask the same thing. Every prompt volume figure you will ever see is a model's guess about a question nobody can look up.

That does not make it useless. It makes it a ranking signal rather than a measurement, and once you treat it that way you can get most of its value without paying for it.

What the vendors are actually estimating

Each vendor builds prompt volume differently, and none of them publish the full method, but the inputs are the same handful: search volume for the keyword form of the question, extrapolated by an assumed share of searches that have moved to assistants; clickstream or browser-panel data showing what a sample of users typed into the assistants; the vendor's own logs of prompts run by its customers; and a language model clustering thousands of phrasings into one "prompt". The output is an estimate of an estimate, and two tools will give you numbers that differ by a factor of five for the same question.

That is not a reason to distrust the tools. It is a reason to read the number the way you would read a keyword difficulty score: fine for comparing two prompts against each other, meaningless as an absolute. "How much does an SEO audit cost" at 3,000 and "best SEO audit tool" at 400 tells you which one to write for. It does not tell you that 3,000 people asked.

Why absolute volume matters less in AI answers than in search

In Google, volume is destiny: a keyword with ten searches a month is not worth a page. In an assistant the economics are different, for three reasons.

First, prompts are long, so the demand is spread across phrasings. The question "which CRM should a ten-person agency use" is asked a hundred ways, and the assistant answers all of them from the same few sources. A page that becomes one of those sources is cited across the whole cluster, whatever the volume of any one phrasing.

Second, the answer has three to five names in it, not ten links. Being in the answer for a low-volume prompt is worth more than ranking seventh for a high-volume keyword, because the searcher reads the whole answer and rarely asks a follow-up.

Third, commercial prompts are disproportionately valuable. The people who type "best X for Y" into ChatGPT are further along than the people who type "X" into Google, and the assistant's answer is closer to a recommendation than a list. For most businesses the twenty prompts that matter are obvious once you see the answers, and their volume is beside the point.

How to estimate prompt volume without a tool

If you want a number anyway, here is how to get one that is about as good as the tools' and costs nothing.

• Start with the keyword form. Take the question, strip it to the two-to-four-word keyword a Googler would type, and look up its search volume in any keyword tool. SEO Stuff's Keywords Explorer does this on credits, with no subscription. That is your floor: the question is asked at least that often, somewhere.
• Apply a share. Industry estimates of how much search has moved to assistants vary widely and change monthly; the honest move is to pick a range (say 5% to 15% for commercial questions in 2026) and carry both ends. The number you get is "a few hundred a month", and that is as precise as anyone's.
• Ask the assistants. Put the question to ChatGPT, Perplexity, Gemini and Claude with web search on and look at the searches they run and the pages they cite. If four assistants cite the same three sources, the question is asked often enough that the sources have stabilized. If every answer is different, it is either rare or nobody has written for it yet, which is an opportunity.
• Check Search Console for the question form. Filter your queries to ones containing "how", "what", "best", "vs" or "should". Questions that already reach your site through Google are being asked of assistants too, and you have a head start on them.
• Count the real-world signal. Sales calls, support tickets, your site search, the questions in Reddit threads about your category. A question your prospects ask you every week has prompt volume whatever a tool says.

Put those together and you have a ranked list of questions with a rough order of magnitude for each, which is everything prompt volume is for.

What to do with the list

Pick twenty. Not two hundred; twenty questions a real buyer in your category asks, weighted toward the commercial ones. Ask each of them of the four assistants, with web search on, and record the answers: who is named, who is cited, and what the answer gets wrong. That is a baseline, and it tells you more than any volume figure, because it shows you which sources you need to be in.

Then write for the clusters, not the phrasings. One page that answers "which CRM should a small agency use" directly, with the trade-offs and current prices, is the page the assistants will quote for every version of the question. The research that coined the term generative engine optimization found that pages with statistics, quotations and citations in them were the ones that got cited; keyword density did nothing. Prompt volume tells you which twenty questions to write for. It does not tell you how to write, and the how is where the work is.

This is the method behind SEO Stuff's Done-For-You Package: the twenty-question baseline goes on your dashboard on day one, the strategy guide names the questions you can own, and the ten pages and three placements are built around them. If you want to see where you stand first, the free SEO + AI search audit includes the questions the assistants are already answering about your category, and who they are naming instead of you.

The short version

Prompt volume is a modelled estimate with no ground truth, useful for ranking prompts against each other and useless as a count. Get the ranking from keyword volume, the assistants' own answers, Search Console and your sales calls; pick twenty questions; write for the clusters; and measure what the assistants say before and after. The tools are worth paying for when you are tracking hundreds of prompts across markets every week. For deciding what to write this quarter, you already have what you need.