10 questions your customers ask ChatGPT — and how to find your own

Your customers are already asking the AI questions about your industry — almost never mentioning your company name. Here’s a list of 10 questions divided into the four categories, a half-hour exercise to help you derive your own from real customer enquiries, and the method for prioritising them down to the five that are most likely to generate revenue.

Your customers are already asking the AI questions about your industry — but almost never mention your company name. They ask about problems, prices, options and “who’s nearby” — and the answers determine who makes the shortlist. Here’s a collection of 10 phrases you can customise in five minutes, an exercise to help you derive your own from real customer enquiries, and a simple method for prioritising down to the five that are closest to generating revenue — the ones your monitoring strategy should be built around.

Why are the questions more important than the keywords?

Because the battle has shifted from the search results to the answer itself. Ahrefs’ analysis of 300,000 keywords found a 34.5 per cent lower click-through rate for the top position when an AI answer appears at the top — and 58 per cent in the follow-up measurement in December 2025. When the customer receives the answer directly, it is the wording of the question that determines which companies and sources appear at all.

And questions behave differently from search terms. In our own analysis of 576 AI responses, four unbranded customer questions were enough to paint a picture of three entire sectors: lawyers were recommended by name in 57 per cent of the responses, plumbers in 38 per cent and accountants in 32 per cent (the methodology behind our analysis here). A search term is a fragment — “accountant Aalborg price” — but an AI question is a whole situation: “how much does an accountant cost for a small online shop, and what determines the price?”. It is the situation you need to target, not the fragment.

What are customers actually asking? The 10 questions

Customer questions fall into four types — recommendation, comparison, problem and local — and this guide covers all four. Fill in the fields in square brackets with your own industry, town and your own tasks, and you’ll have a rough list of ten in five minutes:

Notice what’s missing from all of them: your company name. Customers who know your name don’t need to ask — it’s the anonymous queries that determine whether you’re even considered by everyone else.

How do you find your own questions?

The bank mentioned above is a starting point, but your best questions are already in your inbox and your phone notes — phrased by real customers in their own words. The exercise takes half an hour and gives you phrasing that no generic list can match:

1. Gather 20 real enquiries: quote requests, initial emails, contact form messages, and the questions you get asked time and time again on the phone or at initial meetings.
2. Write down the customer’s exact wording — not your interpretation of it. The customer writes “my decking has turned green and slippery”, not “algae treatment for wooden decking”. The AI uses the customer’s words, so that’s what you’ll be judged on.
3. Rewrite each enquiry in the form of a question — without your company name, and retaining the town/situation.
4. Sort them into the four types and merge nearly identical phrasing.
5. Supplement this with your search data: if you have Google Search Console, look for queries phrased as questions — these are gold, because they prove that people actually phrase things that way. No search data? Then your customers’ emails are your best source.

The result is a rough list of about ten that are yours — rooted in what your customers are actually asking for, not in what a marketing department thinks they’re asking for.

How do you prioritise from 10 down to 5?

You need to narrow it down to five, because five is enough to spot the pattern – and because that’s the number you can monitor for free. Prioritise proximity to purchase over volume: a phrase used by a thousand people in the research phase is worth less than one used by ten people with their hand on their wallet. Each of the four types has its own place in the journey:

A balanced set of five for most people: two ‘recommendation’ types, one local, one comparison and one problem. If you’re unsure why the two results in the right-hand column are so different — and why they each require different approaches — read ‘Cited or recommended? Short version: ‘Recommendation’ and ‘Local’ show whether you’re on the shortlist; ‘Comparison’ and ‘Problem’ show whether your content is being used as a source.

So what do you do with the five?

Two things, in that order. First: test them manually today — the 5-question test will show you in a quarter of an hour whether you’re mentioned, who’s mentioned instead, and what sources the answers are based on. That’s your sample, and it costs nothing.

Next: turn the sample into a monitoring system. The results change from week to week, and a single manual test cannot distinguish a real change from random variation. Create a free account with Geoa and enter those five terms — they’ll then be monitored automatically, and you’ll see how they’re performing rather than having to guess. The free five-point check is exactly the kind of spot check you could do yourself manually — only measured correctly, every week, with a notification when something changes.

Frequently asked questions

How many variables should I monitor?

Start with five — that’s enough to spot the pattern, and that’s the number included for free. Only expand once those five have been running for a few weeks and you know which types matter to your business: paid plans monitor up to 25, 100 or 300, depending on the tier.

Should I ask the questions in my own language or in English?

In the language your customers use. The AI responds in the language the question is asked in and often draws on other sources — so a test in English doesn’t tell you much about what your local customers will see. If you sell to several countries, test the most important questions in each market’s language.

What if my customers aren’t using AI yet?

You can rarely tell straight away — but the shift is measurable in click-through rates, and it’s a one-way street. The advantage of starting now is that it takes just half an hour’s work, and you’ll establish a baseline: if your customers start asking the AI questions next year, you’ll already know where you stand, whilst your competitors will only be able to guess.

Can I use my Google keywords as questions?

Not directly — keywords are fragments, and AI questions are whole situations. But they’re good raw material: take your most important keyword and write out the situation it represents. ‘Leaky gutter price’ becomes ‘how much does it cost to get a leaky gutter repaired, and is there anything I can do myself first?’. Question-form queries from Search Console, on the other hand, can be used almost verbatim.