Use case P44 · Human
House hunter gives a custom ChatGPT his requirements to shortlist properties, then catches listings on the wrong continent
- Home, garden and repairs
- Intermediate
- ChatGPT, Word
- Source checked · blog post
The short version
What they did
Ben M Schorr made a custom GPT called Steven for his family’s house hunt and uploaded a Word document of areas, must-haves, preferences and dealbreakers. He grouped neighborhoods into zones so he could request new listings without repeating the full geography. He reports useful matching and corrections, but also an episode when a request for one zone returned London properties on the wrong continent; his theory of a service problem is not a confirmed explanation.
Result (as claimed)
“For the most part it’s identifying properties that suit our needs, and when it doesn’t I’m able to gently correct it (and update the document) to refine the results.”
“It proceeded to list a few properties in London (wrong continent) that didn’t really seem to align to any of our requirements.”
Prompt
I’m house hunting in [country and named areas]. My requirements document is [attach document with zones, must-haves, preferences and dealbreakers]. These are the listings I want screened: [paste listing text or accessible listings].
1. Confirm each property’s country and area, then exclude any outside the specified places.
2. Compare each remaining listing with every requirement, showing the listing text that supports the match.
3. Give me a shortlist table with price, location, confirmed matches, unknown details and questions for a viewing.Use plain words and keep explanations brief. Don’t treat missing information as a yes. Don’t invent listings, prices, school ratings or legal rules. If you can’t open a listing, ask me to paste it. This is a shortlist for me to check, not a recommendation to buy.
Do it yourself
These steps are my suggested way in, not Ben’s setup.
What you need
- A chat assistant that can keep a file and instructions across chats, such as a custom GPT or a project in ChatGPT, Claude or Gemini.
- A requirements document: areas, must-haves, nice-to-haves and dealbreakers.
- Listing links or text from the sites you already use.
Start simple
- Write your requirements in one document. Name your areas or group them into zones.
- Create a project or custom GPT, upload the document and paste the prompt above as its instructions.
- Paste in five listings you already know. Check that it sorts them the way you would.
- Then paste new listings and ask for a shortlist with evidence for each requirement.
- When it gets one wrong, correct it and update the document.
Take it further
- Ask for a comparison table of your shortlist: price, commute, size and open questions.
- Ask for a list of questions for the agent or for a viewing.
- Keep a “rejected and why” list so it learns your dealbreakers.
Check your work
- Open every listing yourself. Confirm the address, price and status. Ben got listings on the wrong continent.
- Treat missing details as unknown, not as a yes.
- Check school ratings, taxes, zoning and legal questions with official sources and your agent or attorney before you make an offer.
Before you rely on it
- This is general information, not financial, tax or accounting advice for your situation.
- AI tools make arithmetic mistakes and misread statements, and they sound just as sure when they do. Check every number against your own records before you act on it.
- For decisions with tax, loan, insurance or retirement consequences, talk to a qualified professional who knows your situation.
Source checked. The link, author, date and quote were confirmed against the source. The result itself has not been reproduced. Prompt adapted for this library, inspired by the case. Business size: Individual. Results are quoted as the source claimed them and never improved. Spot an error or a broken link? Email charlie@cbarmorecpa.com.