What is prompt volume in AI search, and how is it estimated?
Prompt volume is the AI-search cousin of keyword volume. Here is what it means, how vendors estimate it, where the numbers mislead, and which tools publish it.
Prompt volume is an estimate of how often people ask an AI assistant a given question, or a close variant of it, over a period of time. It is the AI-search cousin of keyword search volume. If "best CRM for a five-person agency" has a high prompt volume, many people are asking some version of that question in ChatGPT, Gemini, Perplexity or Google's AI features. If it has a low one, you can still track it, but you should not expect much demand behind it.
The word "estimate" matters. Nobody outside the AI companies sees a complete log of what users type, so every prompt volume you see in a tool is modeled. Understanding how it is modeled tells you how much weight to put on it.
Why prompt volume is harder than keyword volume
Keyword volume has a long history. Search engines expose query data through their own tools, clickstream panels fill gaps, and the whole industry has settled on a shared vocabulary. A keyword is also short and repetitive, so "running shoes" gets typed millions of times in nearly identical form.
Prompts behave differently. They are long, conversational and highly personal. Two people with the same need will phrase it differently, add different context, and get different answers. Many prompts are asked exactly once. So "volume" for a prompt is really volume for a cluster of similar prompts that share an intent, and the clustering method is part of the estimate.
There is a second wrinkle. A single prompt can trigger several background searches. Those sub-queries, often called query fan-outs, are what the engine actually sends to its search index. A topic can be well covered by fan-outs even if no user ever typed the exact prompt.
How vendors estimate it
Methods vary, and vendors rarely publish the full recipe. From what the tools in our catalog describe, three approaches show up.
Modeling from keyword data. Ahrefs Brand Radar offers an AI Visibility Index of more than 496 million prompts modeled from Ahrefs keyword data, covering AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot. That gives you scale, but the prompts are derived from Google-style keyword demand rather than observed chat behavior. An independent January 2026 test cited in our review found it reported 3 ChatGPT mentions where 123 were verified manually, so treat it as directional.
AI-generated prompt sets. Semrush's AI Toolkit tracks prompts that, per critics quoted in our review, are AI-generated approximations of real questions rather than observed ones. That is fine for a rough map of topics, less fine if you want to know which phrasing actually wins.
Dedicated volume datasets. Profound sells a Prompt Volumes dataset of query demand per prompt. It is a real strength, but the full dataset sits in the Enterprise tier, which has no published price and starts around $40k a year according to our review.
Promptwatch takes a blended route. Its Prompt Volumes show search volumes, difficulty scores and citation rates across AI search, Google and Bing, and its Prompt Explorer suggests prompts from Google Search Console, ranking keywords, People Also Ask or any page you point it at. Because it starts from signals you already own, such as your Search Console queries, the suggestions are tied to demand that has already reached your site.
What a prompt volume number can and cannot tell you
Use it for ranking. If you have 200 candidate prompts, volume helps you decide which 50 deserve daily tracking and content first. Used that way, an imperfect estimate is still useful, because relative order is more reliable than absolute counts.
Do not use it as a forecast. A volume of 1,000 does not mean 1,000 chances to be mentioned. The answer may cite five sources out of hundreds, and many AI answers satisfy the user without a click. Volume also says nothing about whether your brand is currently named. For that you need visibility tracking: Visibility Score, Share of Voice, sentiment and position in the answer.
Do not compare volumes across vendors. A Profound number, an Ahrefs number and a Promptwatch number come from different models and different prompt definitions. Pick one source for ranking and stay with it.
A practical workflow
- Build a candidate list from three places: your Search Console queries, your sales team's most common questions, and competitor comparison phrases.
- Attach volume and difficulty to each prompt, then sort. In Promptwatch this is done in Prompt Volumes, and you can tag prompts by topic, type (organic, brand, competitor comparison) and intent.
- Track the top group daily across models, countries and languages, and use query fan-outs to see what sub-questions the engines actually run.
- Check the Answer gap view, called content gap analysis in the product, to find high-volume prompts where competitors are cited and you are not.
- Publish against those gaps, then watch whether visibility moves.
How the tools compare on prompt demand data
| Tool | Demand data | Source of the numbers | Entry price per our catalog |
|---|---|---|---|
| Promptwatch | Prompt Volumes with difficulty and citation rates, Prompt Explorer | Blends AI search, Google and Bing signals, Search Console input | Free tier; paid plans from $95/mo |
| AthenaHQ | Prompt and response analysis, GEO score | Tracked responses | Free tier; from $295/mo |
| Scrunch AI | Prompt tracking, insights | Tracked responses | From $250/mo |
| Profound | Prompt Volumes dataset | Vendor dataset, Enterprise tier | Enterprise custom, around $40k a year |
| Ahrefs Brand Radar | AI Visibility Index of 496M+ prompts | Modeled from keyword data | Index from $199/mo on top of an Ahrefs plan |
| Semrush AI Toolkit | AI-generated tracked prompts | Approximations | $99/mo per domain add-on |
Who should pick which
Pick Promptwatch if you want volume to feed an actual workflow: volume, then tracking, then the content gap, then publishing, then crawler logs and visitor analytics to prove what changed. Pick Ahrefs Brand Radar if your team lives in Ahrefs and you want category-level research, accepting that it is modeled from keywords. Pick Profound if you have an enterprise budget and want its dataset specifically. AthenaHQ and Scrunch AI suit teams that want tracking plus actions and will not lean on volume figures.
The honest summary
Prompt volume is a modeled estimate of demand for a question, built from keyword data, generated prompt sets or vendor datasets, and it works best as a ranking signal. Use it to choose what to track and write, never as a promise of how many people will see your brand. Then measure the result with real visibility, citation and traffic data, which is where an end-to-end platform earns its keep.