ChatGPT for real estate: what buyers ask it, and what an agent should do

Buyers and sellers now put the questions they used to ask an agent to an assistant first: what is happening to prices in this suburb, what will my flat sell for, what does selling cost, which agent should I use, how long does the process take. ChatGPT answers those from public pages, portal listings, local news, government and council data, review sites and forums. It does not have your CRM or your pipeline. So the work for an agent or a brokerage is to become the public source for those questions, in your own patch, with dated numbers and a profile that says the same thing everywhere.

The four questions, and where the answer comes from

What they ask	Where an assistant finds it	What you can change
What is the market like in this area	Portal market pages, national indices, local news	Publish a dated local page with your own figures and method
What is my property worth, what will it cost to sell	Portal valuation pages, fee comparison sites, whatever page has a number on it	Publish your fee structure and a worked example, with the date
How does buying or selling work here	Government and regulator pages, law firm explainers	Publish the process for your jurisdiction, step by step, with timescales
Which agent should I use in this area	Directories, review platforms, portal agent profiles, your own site	Make your profile consistent and your reviews real and plentiful

The pattern across all four rows is the same. An assistant uses the most specific public page it can find with a number and a date on it. In most suburbs no such page exists, which is why national averages end up in the answer your seller quotes back at you. Publishing the local figure is the whole opportunity.

The nine things to fix, in order

1. Check your listings are readable at all. Many agency sites load property data through an IDX or portal widget inside an iframe or after JavaScript runs. Treat that content as invisible to an AI fetcher. Publish readable, server rendered pages for areas, property types and price bands, and keep the widget for the interactive part.

2. Let the named crawlers in. Check your robots file and your access log for GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot and ClaudeBot. Agency sites on template platforms are often blocking several of them without anybody deciding to.

3. Write one page per neighbourhood, with real numbers. Median or average sold price for the last quarter, time on market, number of sales, the school and transport facts that matter, and the date you compiled it. This is the single highest value page type in the category, because almost nobody publishes it properly.

4. Publish the cost of selling. Your commission structure, what is included, what is extra, and a worked example on a typical local price. Fee questions are asked constantly, and the pages that answer them with figures are the ones that get quoted.

5. Publish the process for your jurisdiction. Steps, who does what, and realistic timescales. Buyers ask assistants about the process more than about any individual property.

6. Make every agent a clean entity. One page per agent: full name, licence or registration number where you have one, years in the area, specialisms, languages, and the areas covered. Identical wording on your site, your portal profiles and your directory listings. Variation splits one agent into several weak records.

7. Keep name, address and phone identical everywhere. Google Business Profile, the portals, local directories, your own footer. This is unglamorous and it decides whether you are one business or three.

8. Get reviews where the answers come from. Ask which platforms were cited when you test, then concentrate on those rather than spreading effort evenly.

9. Add correct structured data. Organisation and local business markup, and listing markup where you have real listing pages, matching the visible text on the page.

Testing it in your own market

Run these monthly, logged out or in a temporary chat, with your location set, and without naming your agency:

• Best estate agents in [your suburb]
• What is the average house price in [your suburb] right now
• What does it cost to sell a house in [your city]
• Which agent should I use to sell a two bedroom flat in [your suburb]
• Is [your suburb] a good place to buy a family home
• How long does it take to sell a property in [your city]

Record which agencies come back, which domains were cited, and whether anything said about your area is wrong. The cited domains are your next quarter of work. The wrong facts are a correction job at the source.

An agent is chosen locally, and local is exactly where an assistant improvises. Ask it for a good agent in your suburb and it returns two or three names assembled from listings, reviews and directory pages, none of which you control. The useful first step is to run those questions for every area you work and see whether you appear. AI Knows Us does that area by area and records what each answer cited.

Using assistants inside the business

Two uses hold up well, and one needs a rule.

• Drafting and editing. Listing descriptions, area guides and follow up emails, drafted fast and then checked. Give the model the property facts; do not let it supply them. A model filling a gap will produce a plausible room count.
• Research and summarising. Pulling the key points out of a long lease, a survey or a planning document, with a human reading the original before anything goes to a client.
• Compliance is a human job. Advertising and discrimination rules in property are strict and differ by country. Every generated description gets read by a person against your local rules and the property record before it is published.

This page is about assistants and property search. For the search side, neighbourhood pages, listings and IDX, see real estate SEO. For the general method, see how to get recommended by ChatGPT.

Asked and answered: ChatGPT for real estate

Can ChatGPT value a property?

It can produce an estimate from public asking prices and index data, which is why sellers arrive with a number already in their head. Your advantage is the sold data, the condition of the actual property and the current buyer demand, so publish your local figures and a dated method, then the conversation starts from your page.

Do assistants have live listings?

They read listing pages when they search, so what they return is whatever the portals and agency sites had published at that moment. Keep your own listing and area pages server rendered so they are part of that picture.

Will assistants replace agents?

They are replacing the first conversation, which used to be an enquiry. The practical response is to be the business named in that conversation, because the answer now shapes the shortlist before anybody calls.

Which matters more, my website or the portals?

Both, for different reasons. Portals and directories give an assistant independent corroboration that you exist and operate in that area. Your own pages give it the specifics nobody else publishes. Neither substitutes for the other.

What if an assistant says something wrong about my area or my agency?

Fix the source. Find which public pages carry the wrong figure or the old address, correct them at origin, and publish a dated page stating the correct version plainly. The sequence is in what to do when ChatGPT gets a fact about you wrong.

What to do first

• Run the six prompts above for your own patch today and save the answers and the cited domains.
• Open one of your listing pages, view source, and search for the property description. If it is not there, that is your first fix.
• Write one neighbourhood page with real sold figures and the date you compiled them.
• Publish your fee structure with a worked example.
• Make every agent profile identical across your site, the portals and your directory listings.