How to Find the Prompts Your Buyers Ask AI

Last updated: November 2026

To find the prompts your buyers ask AI, start with their own words: sales calls, support tickets and long questions in Google Search Console. Add People Also Ask, Reddit threads and competitor comparison searches. Prompt volume tools from Profound, Semrush and Ahrefs can rank the list. Then pick 25 prompts and track them on a fixed schedule.

Key takeaways🔗

• No AI company publishes prompt data the way Google publishes keyword data. Every "prompt volume" number is an estimate from a third party.

• Your best free sources are your own: sales calls, support tickets, onboarding forms and long, question-style queries in Google Search Console.

• People Also Ask, Reddit and "X vs Y" searches show how buyers phrase comparison and "best of" questions.

• Profound (Prompt Volumes), Semrush (Prompt Research) and Ahrefs (Brand Radar) all sell prompt data. Each builds it differently, so treat the numbers as directional.

• A good tracking set has about 25 prompts spread across category, comparison, alternative, problem and brand questions, with the wording kept fixed.

Why are prompts different from keywords?🔗

Prompts are longer and more specific than keywords, and AI tools often break them into several shorter searches before answering. That changes what you should track.

A buyer who types "crm" into Google might ask ChatGPT, "What's the best CRM for a 10-person real estate team that already uses Gmail?" Same intent, very different wording. And the AI does not search for that full sentence. Google says AI Overviews and AI Mode "may use a 'query fan-out' technique, issuing multiple related searches across subtopics and data sources" (Google Search Central). Microsoft says Copilot sends Bing a query that "consists of a few words informed by the user's prompt," not the prompt itself (Microsoft Learn).

So you need two lists. The prompts are what you track, because they show whether your brand gets named. The short searches behind them are what you rank for. This post is about the first list.

Where can you find the prompts your buyers ask?🔗

You can find buyer prompts in seven places, and the free ones are usually the most accurate because they use your buyers' real words.

Work through them roughly in that order. Collect everything in one spreadsheet with three columns: the prompt, where you found it, and the buying stage.

How do you mine sales calls and support tickets for prompts?🔗

Sales calls and support tickets are the best prompt source you have, because they capture what buyers asked in their own words right before or after they chose you. Most of those people asked an AI tool something similar first.

Look for three things:

1. The opening question. "We're trying to figure out how to..." is often a prompt in disguise.

2. Competitors named. "We're also looking at Gusto and Rippling" becomes "Gusto vs Rippling for a restaurant" and "best Gusto alternatives."

3. Deal-breaker constraints. Team size, budget, location, integrations. These are the details buyers add to AI prompts that they rarely type into Google.

If you record calls, search transcripts for phrases like "looking for," "compared," "alternative to," and "how do I." If you do not, add one question to your demo form: "What did you search for or ask before finding us?" The answers are free prompt research.

How do you find prompt ideas in Google Search Console?🔗

Google Search Console shows the long, question-style queries that already bring people to your site, and those are the closest Google data gets to AI prompts. Use a regex filter on the Queries tab to pull them out.

In the Performance report, add a Query filter, choose "Custom (regex)," and try these two patterns:

^(how|what|which|best|is|are|can|should|why|who)\b

^(\S+\s+){7,}\S+$

The first finds queries that start like a question. The second finds queries of eight words or more. Search Console uses RE2 syntax for these filters, and Google notes that filtering by query can change totals because of "the omission of anonymized queries" (Google Search Console Help). In practice, many of the longest and rarest queries are hidden, so what you see is a sample.

One thing Search Console will not give you: AI prompts themselves. Its Generative AI performance report, rolled out to all sites as of August 31, 2026, shows impressions from AI Overviews and AI Mode by page, country, date and device, but has no query dimension and no clicks (Google Search Console Help). Use it to see which pages appear in AI features, then use the regex filters to guess which questions led there.

How do People Also Ask, Reddit and competitor searches help?🔗

People Also Ask, Reddit and competitor searches show how buyers phrase questions when they are comparing options, which is when AI recommendations matter most.

People Also Ask. Search your main category terms on Google and expand the "People also ask" box. Each answer opens more questions. Copy the ones with buying intent ("Which is better, X or Y?", "What is the best X for small businesses?"). This is not a fringe source: Ahrefs says it builds the prompts in its Brand Radar tool from its keyword database plus "People Also Ask and semantic fanout" (Ahrefs).

Reddit. Search your category on Reddit with words like "recommend," "alternative," "switching from" and "worth it." Read the thread titles, not just the replies. A title like "Best invoicing app for a solo contractor? QuickBooks feels like overkill" is almost a ready-made prompt. It also tells you which competitor buyers are leaving. Our post on Reddit citations in AI answers explains why these threads also shape what AI tools say.

Competitor comparisons. Type a competitor's name into Google and note the autocomplete suggestions: "[competitor] vs," "[competitor] alternatives," "[competitor] pricing." Then look at which brands appear on the comparison pages that rank. Each pairing is a prompt worth tracking. Our competitor AI visibility analysis guide shows how to turn this into a full gap review.

Which prompt volume tools exist, and can you trust them?🔗

Three well-known vendors sell prompt volume data as of October 2026, and each builds it differently. None of them has direct access to OpenAI's or Google's logs, so treat every number as an estimate.

How they differ matters. Profound says its data comes from real conversations, but its announcement does not say where those conversations come from. Ahrefs starts from Google search demand and turns it into prompts, so its numbers reflect Google searches, not chatbot use. Semrush calls its prompts real but its feature page does not explain the collection method. When two tools disagree, different sources are the likely reason.

Use these tools to rank a list you already built, and to spot topics you missed. Do not use them as your only source. A prompt that a tool says is popular across the US may never come up in your sales calls, and the reverse is common for niche B2B products.

We build TrendlyAI, so weigh this: TrendlyAI has no prompt volume data. It tracks the prompts you choose, which is why the free sources above matter so much for our own customers.

How do you build a 25-prompt tracking set?🔗

Build a 25-prompt tracking set by spreading prompts across five types, writing each one the way a buyer would say it, and freezing the wording so results stay comparable week to week. Twenty-five is enough to cover the main buying paths without drowning in data.

Here is a split that works for most B2B and ecommerce brands:

A few rules make the set more useful:

• Use real constraints. Include the team size, budget, location or integration your buyers mention. "Best CRM" tells you little. "Best CRM for a real estate team using Gmail" tells you a lot.

• Write for US buyers if they are your market. Use dollar budgets and US terms.

• Keep competitors in the set. At least half the prompts should be ones where a competitor could be named, so you can see who wins.

• Do not change wording once tracking starts. Small edits change the answers. If a prompt stops being useful, retire it and add a new one, and note the date.

• Review every quarter. Swap in prompts from new sales calls and drop ones nobody asks.

TrendlyAI tracks 25 prompts per brand on every plan, which is why we use 25 here. The same set works in a spreadsheet if you check by hand. For how to run the checks and read the results, see our guide to tracking brand mentions in ChatGPT, Perplexity and Google, and for the bigger picture of each engine, the GEO hub.

What do you do once you have the list?🔗

Once you have the list, run every prompt in the AI tools your buyers use, record which brands are named and which pages are cited, and start with the prompts where a competitor is named and you are not. Those gaps are your to-do list.

For a quick first pass, the free TrendlyAI check runs 5 prompts on ChatGPT and Perplexity in about 30 seconds with no signup, and shows whether your brand or a competitor gets named. If ChatGPT is the gap, our guide on how to get mentioned in ChatGPT covers the fixes.

Frequently asked questions🔗

Can I see what people ask ChatGPT about my brand?🔗

Not directly. OpenAI does not share user prompts with brands. You can estimate demand with third-party prompt volume tools, and you can see which questions mention you by tracking a fixed set of prompts yourself.

Is there a Google Keyword Planner for AI prompts?🔗

No official one. Profound, Semrush and Ahrefs sell prompt data, but each builds it from different sources, such as anonymized conversations, clickstream data or Google search demand. Use them to rank ideas, not as exact counts.

How many prompts should I track?🔗

About 25 is a practical number for one brand. It covers category, comparison, alternative, problem and brand questions without too much noise. Larger teams and agencies often track more per client.

Should my prompts include my brand name?🔗

Only a few. Most prompts should be unbranded, like "best payroll software for restaurants," because that is where buyers discover options. Keep two or three branded prompts to check what AI tools say about you.

How often should I update my prompt list?🔗

Review it every quarter. Keep the wording of active prompts fixed so trends are comparable, and swap in new prompts from recent sales calls, support tickets and Reddit threads.

Related reading🔗

• How to track brand mentions in ChatGPT, Perplexity and Google

• Competitor AI visibility analysis

• How to get mentioned in ChatGPT