Portals know what buyers click. Nobody knows why they walk away.

Last week I read a piece in Elite Agent that I haven’t stopped thinking about.

It summarised new analysis in Harvard Business Review by Sangeet Paul Choudary, an AI and platform strategy adviser and senior fellow at UC Berkeley. His argument is simple and uncomfortable for anyone who runs a marketplace: you can keep attracting customers and keep processing transactions while quietly losing control of the decisions that bring those customers to you.

Real estate is about to test that idea harder than almost any industry.

Search is changing first

For twenty years, buying a home online meant the same routine. Open realestate.com.au or Domain. Pick a suburb, a property type, a price range. Scroll.

That routine is already breaking. Zillow became the first real estate app inside ChatGPT in October 2025. REA launched a realestate.com.au app inside ChatGPT in February this year. Buyers are starting to describe the home they want in plain English and let an assistant do the filtering.

When the assistant does the searching, the portal becomes a source of listings rather than the place where the decision starts. That is a structural shift in how property gets discovered, and it is happening now, not in some five-year forecast.

But that isn’t the part of the article that stuck with me.

The blind spot

Choudary’s sharper point is about what happens after discovery.

The right-hand column is where the real advantage sits. Which enquiries turn into offers. Which offers turn into contracts. Which contracts actually settle.

Today, most of that knowledge lives in two places: an agent’s head and an agent’s notebook. Sometimes a CRM. Rarely anywhere a seller can see it, and almost never in a form anyone can learn from across thousands of sales.

So we have a strange situation. The biggest platforms in the industry hold the widest view of buyer behaviour online, and almost no view of the moments that decide the price.

Why this one hit home

This is exactly the problem we’re building Unreserved to solve.

When we started, the pitch was about commission: a flat fee instead of 2% or more of your home’s value. That still matters. But the longer we run campaigns, the clearer it gets that the bigger opportunity is the data the old model throws away.

Every Unreserved listing already gets a weekly report from our AI Recommendation Engine, built from portal data pulled directly from realestate.com.au and Domain:

• Ten measures, from impressions and views through to enquiries, inspections, offers and days on market.
• Benchmarked against Unreserved campaigns that actually sold, and against the competing listings the same buyers are looking at.
• Two or three recommended actions, each carrying the number that triggered it. If the engine can’t back a recommendation with evidence from your own campaign, it doesn’t make one.

The next layer is the one Choudary is describing: what buyers say, what they like, what they don’t like, what stops them from offering. Captured at the transaction level, across every campaign, so the next recommendation is better than the last one.

That is a very different job from the one a portal does, and a very different job from the one a traditional agent does.

The incentive problem nobody mentions

There’s a reason this data has never been organised for sellers.

In a commission model, buyer feedback is a negotiating tool. It gets filtered, softened or sharpened depending on what the agent needs the vendor to hear that week. "Buyers love it but think it’s a bit high" is doing a lot of work in a lot of campaigns.

What this means

I don’t think portals disappear. I don’t think AI replaces the people who negotiate, explain competing interest and manage the messy parts of a transaction. The Elite Agent piece makes the same point, and I agree with it.

What changes is who holds the most valuable information.

For portals, that means discovery alone won’t be enough.

For agents, it means a notebook full of buyer feedback that the seller never sees is no longer an advantage. It’s a liability.

For sellers, it means you should start asking a better question than "what do you think it’s worth?" Ask: what are buyers actually doing, and why aren’t they offering?

Reading the HBR analysis gave me a lot of confidence that we’re building in the right direction. Residential sales are going to look very different in a few years, and the companies that understand what happens after the enquiry are the ones that will shape it.

Frequently asked questions

Will AI replace real estate portals?

Not in the near term. The Harvard Business Review analysis by Sangeet Paul Choudary does not predict that portals disappear. It argues that AI assistants will do more of the searching, so portals may lose influence over where a buyer’s decision starts.

What data can a property portal not see?

A portal can see searches, saved listings, page views and enquiries. It cannot see why a buyer inspected one home and rejected another, why an offer was or wasn’t made, or why a contract fell over. That information is gathered at inspections and during negotiation.

Is realestate.com.au available in ChatGPT?

Yes. REA Group launched a realestate.com.au app inside ChatGPT in February 2026, after Zillow became the first real estate app in ChatGPT in October 2025.