Prompt Research for AI Search

In keyword research, you look up how often people search for a phrase. In prompt research, you collect the questions people ask AI assistants about your field.

Prompts are longer than keywords. They often include context: who is asking, where, and with what budget. That’s why they are so hard to measure.

Not to be confused with prompt engineering, which is about writing prompts that make AI do SEO work for you. That’s covered in our prompt engineering guide.

OpenAI, Google and Perplexity don’t publish what people type into their assistants. Google’s Search Console report for AI features shows impressions, but no queries.

Tools that show a “prompt volume” estimate it. Here’s how three of them describe their method.

Checked on October 19, 2026. Vendors change their methods, so check their pages before you rely on the numbers.

So use these numbers for direction: is a topic big or small, growing or shrinking? Not for exact counts. Google puts it plainly: “No third-party tool has access to our internal ranking or AI systems.”

• Google Search Console. In the Performance report, filter queries with a regular expression to find long or question-style searches. That’s the closest thing to conversational queries in your own data. The AI features report has no queries, so use the main report.
• Bing Webmaster Tools, AI Performance. This report shows “grounding queries”: the phrases the AI used to find content it then cited. It covers Copilot and Bing’s AI summaries. It’s a sample, but it’s first-party data about your own site.
• People Also Ask and autocomplete. Type your topic into Google and write down the questions it suggests.
• Your customers. Sales calls, support tickets and your site search are full of real questions in real words.
• Reddit and forums. Read threads in your field and note how people describe their problems.

Example: this regular expression finds queries that start with a question word. Paste it in Search Console under Query → Custom (regex).

^(who|what|why|how|when|where|which|can|does|is|are|should)\b

Swap in the question words of your own language if your site isn’t in English.

A search query is short. A prompt is a full question with context. Rewrite each question the way a real customer would ask it: who they are, where they are, and what they need.

• Discovery: “What are good invoicing tools for freelancers?”
• Local: “Which accountants in Rotterdam work with small online shops?”
• Problem: “My invoices keep getting paid late. What can I do?”
• Shortlist: “I need invoicing software under €15 a month that handles VAT. What are my three best options?”
• Alternatives: “I’m leaving [competitor]. What else should I look at?”

Test each prompt without your brand name first. If AI names you anyway, that’s a real signal.

Our AI prompt generator builds a set of prompts like these for your industry and region. Copy them into ChatGPT, Gemini, Claude or Perplexity and check if you’re named.

To be clear: it uses fixed templates, not data from real users. It saves you typing. It doesn’t tell you what people actually ask.

AI search often splits one question into several smaller searches. This is called query fan-out.

After you pick a prompt, our free query fan-out tool shows which sub-questions your page should answer. Our fan-out guide explains how it works.

There’s no magic number. Start with a small set per topic that you can actually check by hand.

Then run the same set again over time, with the same wording. AI answers vary from one run to the next, so one test proves little.

Some vendors suggest fixed numbers, like 25 to 50 prompts. Those are their suggestions, not research findings.

Ready to measure how often AI names you compared to competitors? Read our share of voice guide.

Don’t build a separate page for every prompt. Google says you don’t have to worry about capturing every long-tail variation. Creating content for every variation to manipulate AI answers breaks its rules on scaled content abuse.

Instead, cover a topic well on one strong page that answers the main question and its follow-ups.

1. Google Search Central: AI performance reports in Search Console (opens in a new tab)
2. Search Console Help: filtering with regular expressions (opens in a new tab)
3. Bing Webmaster Blog: AI Performance in Bing Webmaster Tools (opens in a new tab)
4. Ahrefs Help: what AI adjusted volume is and how it’s calculated (opens in a new tab)
5. Ahrefs blog: how Ahrefs estimates AI search demand (opens in a new tab)
6. Semrush Knowledge Base: AI visibility data (opens in a new tab)
7. Profound docs: about Prompt Volumes (opens in a new tab)
8. Google Search Central: AI optimization guide (opens in a new tab)
9. Otterly: how to do prompt research for AI search (vendor guide) (opens in a new tab)

• Free AI prompt generator
• Measure your AI share of voice
• Free query fan-out tool
• Prompt engineering for SEO