Prompt research: how to find out which questions your B2B customers ask ChatGPT

You know how you rank in Google. But you have no idea what a potential customer sees when they type your category into ChatGPT. And that is exactly the problem: more and more B2B buyers start their research with an AI assistant, not a search engine. If you do not know which questions they ask, you cannot know whether you appear in the answer either.

In this article, we show how to do prompt research: how to find out which questions your customers ask ChatGPT about your sector, and how to build a concrete monitoring set from that which you can track month after month.

Why prompt research is the new keyword research

In classic SEO, everything revolves around isolated search terms: short, typed words that someone throws into a search bar. With AI search engines, it works differently. People ask full questions, often in spoken language, often with context attached.

Someone no longer searches for “crm software sme”. They ask: “Which affordable CRM do you recommend for a Belgian SME with a small sales team?” That is a prompt, not a keyword.

The difference is fundamental:

• A keyword is a fragment, a prompt is a complete intention.
• A keyword gives you ten blue links, a prompt gives one summarized answer with a few named brands.
• With keywords you count positions, with prompts you count whether you are mentioned at all.

Want to understand how this shift rewrites your entire findability? Then read our ultimate GEO guide. And if you doubt whether classic SEO still counts: it does, but the rules are changing. We explained it in GEO vs SEO.

How to discover the prompts buyers use to search for suppliers like you

You do not have to guess here. There are a few concrete sources that together give you a surprisingly sharp picture.

Start with your sales and support team

Your colleagues hear every week which questions prospects ask. “What is the difference between you and X?” “Who is this suitable for?” “What does something like this cost roughly?” Those are literally the prompts that buyers also put to an AI assistant. Collect them.

Ask ChatGPT itself

Open ChatGPT and pretend you are your own customer. Type questions a buyer would ask and see which follow-up questions the model suggests. Also ask explicitly: “Which questions do people usually ask before they choose a supplier in this category?” You get a usable starting list back.

Mine the places where AI sources its answers

AI assistants lean heavily on forums, review sites and community content. Reddit, industry forums and comparison sites help shape what ChatGPT says about your category. The questions that recur there are often exactly the prompts you want to cover. We wrote a full Reddit playbook for visibility in ChatGPT that fits this perfectly.

Look at your existing search data

Your Google Search Console and keyword data remain worth their weight in gold. Long-tail question phrases that people already type translate almost one to one into prompts. The difference is in the form, not in the underlying intent.

How to do keyword research for AI search engines instead of Google

The nice thing is: you are not starting from scratch. You just flip your lens.

1. Collect intentions, not isolated words. Group everything you find into buying stages: orientation, comparison, decision.
2. Phrase each intention as a natural question. Write them the way a human speaks, not the way a marketer types.
3. Add variations. The same question exists in ten forms: “best”, “reliable”, “affordable”, “for SMEs”, “in Belgium”. Each variant is a separate prompt.
4. Mark buying intent. Not every question is equally valuable. A question with a buying signal weighs more heavily than a general definition.

For a deeper framework on how AI search engines work and what GEO exactly involves, read what is GEO (Generative Engine Optimization). Do you work in B2B? Then GEO for B2B is your most practical next step.

Building a monitoring set of 20 to 50 prompts per category

This is where the real work happens. A loose list of questions is nice, but you want a fixed set of prompts that you test and track repeatedly. That way you see movement instead of a snapshot.

Determine your categories

Split your offer into recognizable categories or problem areas. A marketing partner would, for example, have categories like “lead generation”, “SEO”, “advertising” and “GEO”. Each category gets its own set.

Fill each category with 20 to 50 prompts

Aim for a mix:

• Orientation prompts: “What is a good approach to lead generation in B2B?”
• Comparison prompts: “Which agencies are strong in GEO in Belgium?”
• Decision prompts: “Which partner do I choose for SEO and advertising together?”
• Problem prompts: “My website gets no leads, what now?”

Twenty to fifty sounds like a lot, but you have them faster than you think once you include the variants.

Measure, and measure again

Run each prompt through ChatGPT, Perplexity and Google AI Overviews. Each time, note: are you mentioned, in what context, and who stands next to you. This is your AI share of voice, and you want to know whether it is rising or falling. We explain the measurement in measuring AI share of voice and give an overview of handy AI visibility tools.

A growing part of the B2B buying journey now plays out inside those AI answers, often before a buyer ever lands on your website. Whoever is invisible there drops off without noticing.

Which prompts should you really dominate

You do not need to appear everywhere. That is wasted energy. Focus on the intersection of two things: high buying intent and a direct link with your offer.

• A definition question that everyone appears on? Low priority.
• A comparison question where buyers weigh up suppliers? High priority.
• A problem question that describes exactly your solution? That is your gold mine.

Besides that, always look at the competition. Which prompts does another brand consistently take in your place? Those are your catch-up opportunities. We have a dedicated guide for that: benchmarking competitors in AI search results.

What you can and cannot expect

Let us be honest: prompt research is not a button you flip and then lead flow rolls out. It is a compass, not an engine. What you can expect:

• A clear picture of where you stand today in AI answers.
• A priority list of prompts that truly matter for your revenue.
• A measurable baseline against which to set your progress.

What you should not expect: dominating everything in a week. Visibility in ChatGPT is built with content, mentions and patience. Anyone who tells you it can go faster is selling hot air. Want to know whether classic SEO becomes redundant in the process? The short answer is in is SEO dead because of AI.

Frequently asked questions

Which questions do my customers ask ChatGPT about my sector?

It varies per category, but they are almost always full questions in spoken language: comparisons between suppliers, price indications, “who is this suitable for” and problem-oriented questions. You find them out via your sales team, via ChatGPT itself and via the forums where AI sources its answers.

How do I discover the prompts buyers use to search for suppliers like me?

Start with the questions your sales and support team hear daily, supplement with the suggestions ChatGPT gives itself, and check which questions recur on Reddit and industry forums. Translate those intentions into natural question phrases and group them by buying stage.

How do I do keyword research for AI search engines instead of Google?

You reuse your existing search data, but you think in full questions instead of isolated search terms. Group by intent, phrase each intention as a natural prompt, add variants and mark buying intent. Then you test each prompt in ChatGPT, Perplexity and Google AI Overviews.

How many prompts do I need per category?

Aim for 20 to 50 prompts per category. Enough to cover orientation, comparison, decision and problem questions, but not so many that you lose the overview and the follow-up.

How often should I remeasure my monitoring set?

Monthly is a good rhythm. AI answers change constantly, so a snapshot says little. Only by measuring repeatedly do you see whether your share of voice rises or whether a competitor is gaining ground.

Ready to see what ChatGPT says about you?

Prompt research is the first step. The next is making sure you appear in those answers, and that you keep appearing. We are happy to help you with that, from strategy and monitoring to the content and mentions that truly make you visible in AI search engines. See how we approach it via our GEO agency, and if you want to understand how customers find you via ChatGPT, the full strategy is ready.

We are a small team, so we move fast and often do more than you expect. Schedule your free intake and we will show you exactly which prompts you are missing today.