How can real estate businesses track their visibility in ChatGPT and AI search?
How real estate businesses can track their visibility in ChatGPT and AI search, from our agency work: prompt design, local checks, GA4 and Promptwatch workflows.
Property is a local, high-trust purchase. People ask an AI engine which agent to call, which developer to trust, which neighbourhood to move to. The answers shape the shortlist before anyone visits a website. When a real estate client asks us how to track that, we give them the same sequence every time.
Start with the prompts, not the tool
The prompt set is where most of the value sits. We build it from the questions the client's staff hear on the phone, not from a keyword export. A typical estate agency set has four parts:
- Intent prompts: "who is the best agent to sell a flat in [area]", "agents that specialise in new builds".
- Comparison prompts: "agency A versus agency B", where the client is one of them.
- Neighbourhood prompts: schools, commute, prices, safety. These decide whether the client's local guides get cited.
- Brand prompts: the agency name plus reviews.
We organise them by city and by persona, such as first-time buyer, landlord or investor. The answers differ a great deal by persona.
Two weeks of manual checks first
Before buying anything, we run the prompts by hand in ChatGPT, Google AI Overviews and Google AI Mode and log who is named and what is cited. It teaches the client how unstable answers are, which makes the case for daily tracking better than any slide.
Add GA4 for the traffic you can see
In Google Analytics 4, we segment referral traffic from AI engines such as chatgpt.com. It undercounts, because many readers never click, but it is the cheapest sign that something is working.
Why we move clients to Promptwatch
For ongoing tracking we run client programs on Promptwatch, because each step of the workflow maps to a feature rather than to another spreadsheet.
Tracking. Daily prompt panels per model, country and language, with Visibility Score, Share of Voice and Position in Answer. Prompts can be tagged by topic and prompt type, and personas let us test the same question from a buyer's and a landlord's angle.
Local. State and city tracking sits on the Business plan ($579 a month) and on every agency plan. Local GEO/SEO targeting, on Business and Agency Scale, pins a monitor to exact coordinates for ChatGPT, Google AI Overviews and Google AI Mode answers.
Citations. Page and domain citation analytics, with Reddit and YouTube reports. For property, portals and directories tend to dominate, so we see where the client is missing.
Crawler logs. Agent Analytics shows IP-verified hits from GPTBot, OAI-SearchBot and ChatGPT-User, plus errors. Listing pages that load through heavy scripts or sit behind bot protection are common culprits. Crawler logs start on Professional ($245 a month).
Traffic and enquiries. Visitor analytics attributes AI-referred visits and conversions, so we can report enquiry-form submissions, not only mentions.
Action. The Action Board converts the data into prioritised tasks, and the Content Agent (Business and Enterprise) can draft neighbourhood and buyer-guide pages against the gaps, with a review inbox for the client.
There is a free Explore tier (10 prompts, ChatGPT only) for a first look, and an AI Visibility Report that needs no login.
What we are careful to say
We never tell a real estate client that tracking will get them recommended. It will not. It shows which sources the engines trust, whether the site is readable to bots, and whether AI visitors turn into enquiries. The gains come from better local content, consistent listings, genuine reviews and a technically healthy site.
We also avoid putting any client or buyer personal data into prompts. The tools track public answers, and prompts describe generic buyers.
A simple monthly rhythm
- Review Share of Voice by city against the two or three nearest competitors.
- Read the new citations. Which portals, which guides, which forum threads?
- Check the crawler error report for property pages.
- Pick two fixes. One content, one technical.
- Compare the same prompts next month.
It is not glamorous. It works because the loop is short and the evidence is specific. If you want to start from your own numbers, the free Promptwatch tier is enough to see a first picture for ChatGPT.