Launching a property portal with no traffic: where AI actually helps

There’s a run of articles saying AI is coming for the property portals. The Negotiator ran one in February 2025 with the headline “Property portals face existential threat from AI says tech pioneer”. The picture is an AI agent that goes off and searches across all the portals for the buyer, and nobody needs to visit a portal again.

I started at realestate.com.au in 2010, and I’ve done pretty much everything in real estate platforms since. The disruption pieces skip two things. One is who actually has the rights to the listings. The other is how anyone starts a portal from nothing, and that’s the one I care about, because I’ve built a portal and I want to launch it. I’ve been building it with coding agents since January (how that works is in How I let coding agents merge to production), and this is the plan for getting it traffic.

The chicken-and-egg problem

Bringing something to market is a chicken-and-egg problem. In a brand new market you’ve got no traffic. You can’t get traffic without listings, you can’t get listings without the agents, and you can’t get the agents without some sort of traffic.

None of this is new. Every two or three years in Australia some agent group got together and wrote a portal. Being able to write a portal is nothing new. Then they discovered that not having visits was the biggest problem. It’s a distribution problem, and it’s always been a distribution problem. AI has made the building a lot cheaper. It hasn’t changed that.

Getting the listings is also harder than it sounds, even when an agency is on your side. The agency says yes, you can have my listings, but they often don’t control them as much as all that. The listings sit in a CRM you have to integrate with, and often the CRM is owned by the major portal and won’t add you.

Getting listings from agents, not portals

The part I’ve put the most work into is getting listings before you have any agent relationship. The system goes to an agent’s website and does an AI blueprint of it. It works out where all the listings are and how the site is laid out, encodes that, then scrapes it and creates all the listings. It doesn’t matter what the website looks like, as long as the listings on it are reasonably complete. All of that works. All I need is a list of agents and I have every listing for an entire country. It refreshes too: you hit each agent website every week to keep it fresh.

What matters is where the listings come from. You’re not scraping the big portal, you’re getting them directly from the real estate agents. That’s harder, but it’s also safer from a legal perspective, because it’s basically what Google does when it indexes the web. And you can always take it down if an agent asks you to, but an agent has less incentive to ask than a competitor does.

You also get things the portals don’t have. Agent sites carry stuff the agent is only pushing to their own website, so you pick up listings that never make it to the portals, off-market listings essentially. I had the idea back in 2018. I drew out a thing that went to the agents directly without needing the CRMs. Back then it was just a drawing, and now it actually works.

Give agents value before you can send leads

Once you’ve got an agent’s listings, you can go to the agent and actually make a deal. But you can’t offer them leads yet, because there’s no traffic to send. The theory is that if I can give agents value outside of leads, I can get the relationship started. I’ve built a bunch of agent products that don’t need traffic volume to matter:

• An AI assistant they can embed on their existing website, which answers questions about their listings and about the agency.
• Agent websites.
• AI social generation off the listings, so each listing turns into posts the agent can put out.

None of those needs a single visitor to my portal to be worth something to the agent. The relationship starts small and low commitment, and it’s built on something the agent can see working on their own site.

Agents are already doing some of this by hand. The National Association of Realtors’ 2025 technology survey found 46% of Realtors use AI-generated content like listing descriptions, and ChatGPT was the most common tool at 58%. That’s an American survey, but it’s a fair signal that agents will take the work if you hand it to them already done.

Traffic: suburb content, search and ChatGPT

As soon as you’ve got the listings you’ll get a certain amount of search traffic. On top of that I generate a whole bunch of content around suburb profiles and other relevant content for SEO, so there’s at least some substance there to bring traffic to. Then you make it accessible within ChatGPT, do all the search optimisation, get some level of traffic, and do some advertising as well. Once you’ve got traffic you can start delivering leads to agents, and it becomes a bit of a flywheel: traffic, leads, agents, more traffic.

ChatGPT is where the disruption pieces put all their weight, and I’d be careful with it. Zillow launched an app inside ChatGPT in October 2025, which OnlineMarketplaces reported as the only real estate app there, and Zillow’s earlier ChatGPT plugin from 2023 had been pulled over fair-housing concerns. And in September this year the OnlineMarketplaces editor, Edmund Keith, wrote that any business getting more than 1% of its traffic direct from the likes of ChatGPT or Gemini is an outlier. I’ll still make the listings reachable in ChatGPT, but it’s one channel alongside search, suburb content and advertising, right? It’s not the whole plan.

Picking a market

Which market you go into matters as much as any of the technology. Some of my ex-REA colleagues suggested the one I’m going into. It’s a paid vendor-advertising market, so there’s a lot of money there, and the agents are a bit unhappy because they’ve been charged a lot and there isn’t any innovation going on. The more I looked at it, the more I thought, yeah, this is a good market to go into. It’s fairly wealthy and it’s small, which makes it more achievable for a small player.

Coming from Australia helps here. Because of Domain and REA, property tech from Australia goes out to the rest of the world, and if you can do it in Australia, you can do it anywhere. The rest of the world is a long way behind on market behaviour. Not the portals specifically, but the market itself: the auction-based competitiveness doesn’t exist in most places.

What’s still unproven

I’ve designed the platform to be extremely cheap to run, but you still have to pay the traffic bill, and the bigger the market, the more traffic you have to serve. That’s a big part of why a small market makes sense for a first launch.

The listing ingestion works and the agent tools are built. The rest is theory, and it’s untested for the most part: whether agents take up the tools, whether the suburb content ranks, and whether the flywheel turns before the traffic bill does. I’ll find out when I launch. We’ll see.