GEO courseModule 2 of 10: Your Prompts and BaselineLesson 2.2
What People Actually Ask AI: Your Prompt Bank and Prompt Map
Collecting the real questions your customers ask, in their own words, and sorting them by audience and stage, so your testing and your pages are built on what people really want, not on guesses.
Why it matters
Test the wrong questions and you'll fix the wrong things. Nobody can see what people type into AI, so you have to collect it.
The evidence
Evidence: Practitioner view
Practitioner view and common sense. No platform publishes what people ask, so your own enquiries are the best evidence there is.
Do this first
Copy 30 real questions from your last month of enquiry emails and calls, word for word.
01What it is
If you want to sell lemonade, you listen to what people in the park actually ask for. You don’t invent questions and stock up for them.
Think of it like
A shopkeeper's notebook of what customers ask for at the counter. After a month, the notebook tells you more than any amount of guessing.
02Why it matters
To the people searching
People type to AI the way they talk: "should I use a letting agent or manage it myself?", not "letting agent [town]". If nothing on your site answers the question they actually asked, you can't be the answer, however good you are.
buyers
Ask about areas, costs and process in full sentences, often several in a row.
tenants
Ask what renting costs, what fees are allowed and where to live on their budget.
sellers
Ask about value, fees and which agent to choose, close to the decision to book a valuation.
landlords
Ask about rent, management options and rule changes, and which agent to trust with them.
To Google, Bing and AI
Evidence: Practitioner viewNo AI company shows site owners the prompts people use, and the "prompt volume" figures in paid tools are modelled estimates. The nearest real data: Bing's AI Performance report shows a sample of the searches Copilot ran to find your pages, and Search Console shows the long, question-style searches already bringing people to you. Google adds that AI features may run several searches per question (query fan-out), so one real question can lead to many.
An agent example
A lettings manager reads 30 landlord emails. Nine ask a version of "should I use a letting agent or manage it myself?", six ask about fees and five ask what's changed under the Renters' Rights Act. Those become three of the ten landlord prompts in the agency's standard set, written exactly as landlords wrote them.
03Your turn
Check yours, then fix it
- 1Read the last 30 enquiry emails, web forms, valuation requests and phone notes.
- 2Copy each question word for word into a sheet, with names removed.
- 3Add a column for the audience (buyer, tenant, seller, landlord).
- 1Ask every negotiator and the lettings manager for the ten questions they answer most each week.
- 2Ask for the exact wording people use, not a tidied-up version.
- 3Compare with your enquiry list. The overlap is your core.
- 1Open Performance, then Search results, then the Queries tab.
- 2Filter for queries containing "how", "what", "should" or "best".
- 3Note 20 that are questions or six words or longer.
- 1In Bing Webmaster Tools, open AI Performance.
- 2Read the sample of grounding queries: the phrases Copilot used to find your cited pages.
- 3Add any you haven't got.
Here are 30 customer questions from a UK estate and lettings agency, with names removed. Group them by audience (buyer, tenant, seller, landlord) and by stage (awareness, research, choosing an agent, ready to act). Then list the ten most common themes. Don't add questions that aren't in my list.
- 1Columns: the prompt in the customer's words, audience, stage, where you heard it, date, value (high, medium, low).
- 2Mark anything you can't trace to a real person as a guess, and keep guesses out of your standard set.
- 3Add to it every month.
- 1Make a 4 by 4 grid: audiences down the side, stages across the top.
- 2Put up to four real prompts in each cell. The starter map under Go deeper fills the gaps until your own data does.
- 3Mark value: choosing-an-agent and ready-to-act prompts from sellers and landlords lead to instructions, so they come first.
- 1Choose 10 to 20 prompts across all four audiences, weighted to sellers and landlords.
- 2Include two or three with your town or postcode, and three branded ones ("Tell me about [agency] in [town]").
- 3Freeze the wording. Lesson 2.3 explains why.
- 1For each audience, write the chain of four or five questions a customer asks in a row.
- 2Check each of your main pages answers the next question, on the page or one link away.
- 3Add a short "next questions" set of links at the end of each guide.
- 1Add "How did you hear about us, and what did you ask?" to your enquiry form and call notes.
- 2Ask your CRM or website supplier whether enquiry messages can be exported to a spreadsheet.
Then: Avoid fantasy prompts. "Which agency is the leading expert in Edwardian terraces?" is something nobody types. If you can't say where a prompt came from, it doesn't go in the standard set.
04Keep it up
Add new questions as they come in. Review the whole bank each quarter, retire prompts nobody asks, and add new ones when rules change (for example, landlord questions after a law changes).
Go deeperA starter prompt map (replace with your customers' words)
Sellers, awareness: "How much is my house worth?" "Is now a good time to sell in [town]?" Research: "How much does an estate agent charge in [town]?" "Sole agency or multi-agency?" "Online or local agent?" Choosing: "Who is the best estate agent in [town]?" "What should I ask an estate agent before I sign?" Acting: "How do I book a valuation in [postcode district]?"
Landlords, awareness: "How much rent can I charge for a 2 bed flat in [town]?" "What changes for landlords under the Renters' Rights Act?" Research: "Should I use a letting agent or manage it myself?" "Let-only or fully managed?" Choosing: "Who is the best letting agent in [town] for landlords?" "Which letting agents in [town] have client money protection?" Acting: "How do I move my property to a new letting agent?"
Buyers, awareness: "Which parts of [town] are best for families?" "What are house prices like in [area]?" Research: "How long does buying a house take in England?" Choosing: "Which estate agents in [town] sell the most homes?" Acting: "How do I book a viewing with [agency]?"
Tenants, awareness: "Where's the best place to rent in [town]?" "What fees can letting agents charge tenants?" Research: "How much deposit do I pay when renting?" Choosing: "Is [agency] a good letting agent?" Acting: "How do I apply for a rental property?"
That's lesson 2.2
Worked through it? Mark it done and we’ll keep track on this device.