Here’s the received wisdom: 'search' is the ultimate moat. Own the place people go to find things and you can charge whatever you like for the privilege of being found. Google proved it for twenty-odd years. The UK’s Rightmove and Sweden’s Hemnet, amongst others, proved it in property portals. The toll goes up, the traffic keeps coming, and the shareholders enjoy their bumper dividends.
On Thursday that moat sprang a $500bn leak.
Alphabet, parent company of Google, saw its shares fell 12.6% on 29 October - the worst day in the company’s history, wiping just over $500bn (£374bn) off its value.
So what spooked the market? Two things - and it’s worth being precise, because the lazy read is “AI bad”.
The first is cash. Capital expenditure in the quarter was $51.4bn, more than double the $24.0bn of a year earlier. Free cash flow went negative, at minus $9.07bn. That’s the bigger number - and the one most of the coverage led on, because Google's always been a cash machine and it's now it's run dry.
The second is the toll - and it’s the bigger signal. Search & other revenue rose 12% to $63.35bn, but was virtually flat on the previous quarter and missed forecasts, having beaten them in each of the previous four. Meanwhile Sundar Pichai, CEO, told analysts query growth was the highest in five years.
More searching than ever - but, for the first time, the money didn’t follow.
Philipp Schindler, Google’s chief business officer, explained why: “the strong growth in AI Mode queries had an impact because of the lower monetisation we were seeing in AI Mode.” Translation: when Google answers the question itself, there are fewer blue links to sell to advertisers.
Pichai’s answer: user experience first, monetisation later. That’s the hope, anyway.
The market will forgive a company for spending if it believes the toll holds.
On Thursday it priced in the possibility that it doesn’t.
That’s the deeper story underneath the surface numbers. Search dominance was never really about search. It was about being the unavoidable step between wanting something and finding it - with dominant portals and platforms being able to charge for the click across. When discovery gets easier elsewhere, the step becomes optional and the toll booth gets bypassed.
Sometimes by a rival. Sometimes by the incumbent itself.
Which brings us to Sweden.
Hemnet was the 'poster child' toll booth in European property portals. Its average revenue per listing (per LISTING, not per branch) went from about £37 in 2015 to £621 in 2025: nearly 17 times over. Its fee climbed from 0.02% to 0.25% of the average Swedish sale price. In 2023 the company told investors its Average Revenue Per Listing (ARPL) “doesn’t have a ceiling”.
Then Booli, an aggregator created by the state-owned bank SBAB, did something terribly unglamorous. It indexed listings from agents’ own websites and showed them for free.
In February, Fastighetsbyrån, a huge agent and former Hemnet shareholder -think Connells but blonder - agreed to upload its listings on Booli.
By Hemnet’s own September presentation, Booli carried about 89,000 listings against Hemnet’s 55,000.
And the response? On 20 October Hemnet launched free listings for agents.
Free. Imagine another company of similar stature in the UK being forced to do that ...
In short, Hemnet is turning itself into Booli. The incumbent has copied the challenger and cut its own price to zero - which is precisely what Google is doing to its own ad inventory with AI Modem under pressure from upstarts ChatGPT and Claude. Both incumbents chose to cannibalise themselves rather than be cannibalised.
For UK agents, be careful which lesson you take from Sweden, because the lazy one is “portals are finished”.
They aren’t. Booli is a portal. What’s dying isn’t the portal. It’s the toll.
That distinction decides where your listings should go. AI assistants don’t read the web evenly. They lean on large, well-structured, authoritative sources, and in property that means national sites carrying complete stock, not thousands of separate agency websites. It’s very likely part of what lifted Booli: an aggregator with every listing in the country is exactly what a machine wants to read.
And a portal whose value is completeness can’t charge you to list. Every fee creates an agent that won't pay, which creates a gap - and gaps are precisely how Hemnet lost Sweden.
For that kind of portal, 'free' isn’t a promotion. It’s the business model – and yes, full disclosure, of course that’s what we’re doing at MyPorta. It's why we’ll never charge any agent a penny.
So the answer isn’t “ditch the portals and go it alone”. On its own, your website is one small site among thousands, and an AI will likely still prioritise listings from a portal it can read easily and knows many more customers hang around on.
Do both.
Make your own site the most complete, current and machine-readable source of your stock: price, status, floorplan, EPC, the lot - updated before anyone else gets it.
Then put that stock on at least one free-to-list national portal built for machines to read. Look for four things.
National coverage, so it’s the site an AI assistant reaches for.
Enriched listings: your data plus the local context buyers ask about.
AI that scans your photos for the features nobody typed in, like the wood-burner, the island, the south-facing garden.
And listings packaged in the formats AI tools actually use, such as MCP, with apps inside ChatGPT and Claude, so a buyer’s assistant can query them directly.
Know your direct-lead share. If you can’t say how many applicants found you without a paid portal, you can’t negotiate; you can only renew.
Add the free one first. Then look hard at which paid one you’d miss least.
Hemnet thought its ARPL had ‘no ceiling’. Certainly before it found one, it fell through the floor - and the floor was Booli's, and 'free'.