No — there's no official way to see how often people ask ChatGPT, Gemini or Claude a particular question. AI companies don't publish prompt volumes the way Google Ads publishes keyword volumes, so every "ChatGPT search volume" figure you see is an estimate. That doesn't mean you're flying blind: combine public proxies with first-party data and prioritise prompts by business value. Here's how.
Why prompt volume is hard to measure
- No public data. OpenAI, Google (for Gemini and AI Mode prompts), Anthropic and Perplexity don't release per-prompt demand.
- Prompts are long and unique. "Best CRM for a 6-person real estate team in Pune that works with WhatsApp" may be asked once in exactly that form, yet thousands of similar questions are asked.
- Conversations, not queries. Follow-up turns depend on earlier ones, so counting single prompts misses how people actually research.
How prompt volume estimates are made
| Method | How it works | Blind spots |
|---|---|---|
| Panels | Opt-in users share their AI usage | Small, skewed samples; may under-represent some countries |
| Clickstream | Browsing data from extensions or apps | Privacy limits; mostly web, not apps |
| Modelling from search volume | Estimate AI demand from keyword volumes and trends | Assumes AI demand mirrors search demand |
| Topic-level trends | Track rising subjects rather than exact prompts | No per-prompt numbers |
All are useful directionally. None is precise for a single prompt.
Better ways to find prompts that matter
1. Use keyword data as a topic proxy
Google keyword volumes and People Also Ask questions tell you which topics people care about. Turn them into natural prompts. See prompt research.
2. Mine first-party conversations
Sales call notes, support tickets, chat logs and WhatsApp enquiries contain the exact questions and constraints buyers bring. See finding buyer questions for AI prompts.
3. Check site search and AI referral landing pages
Your site search shows what visitors look for; landing pages of visits from chatgpt.com and perplexity.ai show which topics AI already sends people to you for. See tracking AI referral traffic in GA4.
4. Read Bing's grounding queries
Bing Webmaster Tools' AI Performance report shows grounding queries — the searches Copilot ran when it cited you. They're a rare first-party glimpse of what AI looks for. See Bing's AI Performance report.
5. Watch communities
Reddit and Quora questions in your category mirror how people ask AI. See Quora and AI answers.
6. Include other languages
In India and other multilingual markets, add Hindi and Hinglish versions of key prompts. See AEO for Indian brands.
Prioritising without exact volumes
Score each prompt or prompt cluster:
| Factor | Question | Score 1–5 |
|---|---|---|
| Business value | How close is this to a purchase, and how valuable is the buyer? | |
| Demand signal | How often does it come up in keyword data and conversations? | |
| Winnability | Are you close to being mentioned, or far behind? | |
| Strategic fit | Does it match the products and markets you're prioritising? |
Track the highest scorers first. See branded vs unbranded prompts and building a prompt tracking set.
What to measure instead of volume
Because answers vary between runs, the useful metric is how often you appear across a well-chosen prompt set — your mention rate and share of voice — weighted by prompt importance. That tells you whether you're winning the questions that matter, whatever their exact volume. See AI visibility metrics and AI answer variability.
A worked example
A B2B payroll company wants to prioritise 100 candidate prompts. It has no prompt volume data, so it:
- Groups the prompts into 12 topics.
- Pulls Google keyword volumes for each topic's head terms as a demand proxy.
- Counts how often each topic appears in three months of sales call notes.
- Checks GA4 for which topics AI referral visitors land on.
- Runs each prompt three times in ChatGPT and Gemini to see where it's already mentioned.
Topics with strong conversation counts and near-misses in AI answers rise to the top — even where keyword volume is modest — because they're close to purchase and winnable.
Be sceptical of precise numbers
If a tool claims "this prompt is asked 12,400 times a month in ChatGPT", ask how it was measured, in which countries and over what sample. Use such numbers to compare topics, not as facts. For more on evaluating tools, see choosing an AEO tool.
