The Real Portals vs LLMs Battle Isn't For Search, It's For Consideration

The pace at which topics flutter into the industry's collective gaze, become the focus and then dance out again has never been faster. This discourse turnover, though, is still slightly slower than the rate of the sector's technological development and the theory behind its deployment.

The billion-dollar question in the industry still seems to be: will portals lose search to LLMs? Most people we've spoken to, and the public indications from portals, suggest this question already has an answer.

The matter has, of course, not been settled conclusively. Using ChatGPT to search for property may just be taking a while to truly enter the zeitgeist, and we may end up with a long interregnum period where consumers use both LLMs and portals in tandem.

But we can at least say that the growth of ChatGPT and Claude on portal traffic pie charts has been far slower than expected (and hoped for in some quarters) and that the current user experience while searching on LLMs still leaves a lot to be desired.

There are plenty of theories we can point to, from LLMs being uninterested in the relatively small returns they'd get from going directly into real estate search to the fact that 25 years of training consumers to interact with uniform search boxes, results pages and contact forms will take a while to shift.

While Zumper Co-founder Anthemos Georgiades told us that ChatGPT accounted for between 4 and 4.5 per cent of the US rental player's organic traffic, other portals are reporting smaller numbers.

• Earlier this month, Baltic Classifieds Group's FY26 statement to the market noted that "visits originating from AI search remain negligible"
• REA Group's Q2 2026 earnings call, CEO Cameron McIntyre said that LLM traffic was "sub-1% and declining"
• In a February call with investors, Rightmove's CEO Johan Svanstrom said that traffic from the big AI players represented less than 0.5% of the portal's traffic.

Svanstrom's choice of the word "meaningless" to describe the trickle of visitors coming to Rightmove from LLMs felt pointed. But it was the rest of his answer to the question put to him about Rightmove's recently released ChatGPT plugin that is perhaps more interesting.

"Look, it is a tool that lots of people use for, you know, different things. For us, this is a test and learn, right? We wanna be where some consumers are, see what we can learn from that. Very importantly, of course, it’s an app that we created. It basically displays listings and consumers then go back and do much more to the experience, where they have all of that experience, and again, all the data and tools and their own history and so forth, on Rightmove, and that’s what we expect going forward as well."

Read that quote carefully. It sounds like Rightmove's CEO is saying something along the lines of...

"Yes, we built something for someone else's platform — we're experimenting and hedging our bets when it comes to where people may choose to search in the future. But it's fine because the user will ultimately come back to the portal where all the data is for the next phase of their journey."

Rightmove don't think they'll lose the search phase any time soon, and I think they're right. But I'm not so sure they and other portals can take the "go back and do much more to the experience" phase for granted. We'll call that one the consideration phase, and it's very important because it's the point in the home hunting journey at which leads usually get sent.

From search to consideration

Svanstrom is right to reiterate to investors that, in the context of a home hunter's journey, Rightmove is the place where a lot of the data, tools and history live.

Portals have become quite good at inferring users' intentions, and they love building tools to help them take the next step. Home hunters can get a valuation of their current home, find an agent to help sell it, and Rightmove's users can even get a mortgage decision in principle through the platform.

Another class of tool has begun to proliferate across the real estate portal industry, and the speed at which these tools have been released suggests that portals know very well that the battle has shifted from search to consideration.

Portals around the world (including Rightmove) have been scrambling to consolidate their tools and data into one interface that can answer natural language questions about listings users may be weighing up as well as, occasionally, questions about financing, agent selection and other considerations.

Above: I went through all the product updates we've covered since the beginning of AI's blossoming in 2023 to find portal tools that allow users to interrogate listings data

Some of these are quite good. Homes.com's Homes AI, for example, can remember the details of a property you previously clicked on and answer questions about trade-offs in commute times and ceiling heights between that listing and the one you're currently viewing. And it can do it all in Russian if you like.

All of these portal tools and interfaces whose job it is to keep users on site for the consideration phase have at least one fundamental flaw, however.

At their core, they want the best outcome for the portal and its agent customers, not the buyer. A portal-built tool is unlikely to tell a prospective buyer what it really thinks a house is worth, and it will pair them with an agent who pays for leads rather than the one who's objectively best suited to sell their home.

There's also the trust issue. Portal brands have been trusted places to search for property listings for decades, but can we say that users around the world have trusted them when they've branched out, for example, into mortgages?

Will consumers trust Rightmove to weigh up questions about trade-offs between homes they've found as well as financing and timelines? Or will they turn to their trusted, artificially intelligent sidekick who has already helped them with their taxes, booked their dentist appointment and told them not to worry about their ingrown toenail?

The consideration phase, I would argue, sits in a different trust category from the search phase, where the dopamine from scrolling through photos has been replaced by the cortisol associated with decision-making. It requires a helper whose loyalties lie with the buyer, not the seller.

An LLM does have the barrier of having to explicitly ask the user about their intentions, budget, timeline, etc and it may not have all of the proprietary data a portal has. But its loyalty is unquestionably to that user looking for a home, not to people working to maximise the seller's price.

Next week I'll look at whether portals can overcome this loyalty problem and win the battle for the consideration phase.