Last week Drew Myers casted some doubt on the latest challengers in the portal wars. In the invite-only GEM forums, he asks whether an AI-enabled real estate portal or AI broker can really topple the incumbents such as Zillow, Realtor, and Homes. The article then gives many examples of portals who were reasonably well-resourced and found early traction only to ultimately fold or reach a soft landing. Getting bought by a mortgage company seems to be the latest exit of choice.
Anyway, it’s a great read and sure hits a little too close to home. Currently the “transmission” is accessible only to the GEM community, and instead of rehashing it, at AnyProp we see yet another angle on what is hurting many upstart AI-powered innovators and broker-adjacent models: the data hurdle necessary to compete with the incumbents is increasingly difficult and expensive.
The NYC Case Study: Deed Data and Listings Joined
Once upon a time, StreetEasy in NYC dominated because they were the first to clean and join NYC listings data with the local equivalent of assessor and deed data in ACRIS (initially via scraping). It sounds basic today, but back in 2007 it was completely unheard of for consumers to find both the owner name, closing price, and financing information alongside the listing broker, original asking price, and listing photos. There was even a paid “Insider Mode” that allowed users to profile a particular listing agent and see all of their closed deals — primarily so you could use the data to negotiate against that agent on your own deal.
Nowadays, assessor plus listing data is just the ante to achieve parity with a slew of struggling niche portals. Just about every new site needs to build their brand and accumulate trust. A fancy Ux or chat bot isn’t going to help with that at all if you don’t have the underlying data. For example, we know everyone searches for their own house and lose interest in a platform if they can’t find it. Even when testing out our AnyProp APIs, we often need to let clients know that they couldn’t find their own house during our API trial because they are using sample data, not real listings.
There are several knowns paths to gain access to such data, but the underlying sources are tightening up. While the NAR hasn’t recommended anti-AI language yet across all of their associations, some MLSs have proactively included a “no machine learning” clause for the vendors and downstream customers who access their data. The most notable is REcore which powers California Regional MLS and their various data shares covering southern California. How they enforce it is an open question, but it’s a legal landmine for startups to note. In the Bay Area, MLSs there have seen so many startups come and go that they even ask for a revenue share from IDX/portal companies.
No one really knows how the MLS landscape will look in 5 years, but it seems like a safe bet that potential innovators and disruptors will increasingly see more resistance to accessing the data. At AnyProp, our goal is not simply to normalize and unify real estate data into a single API. We also specialize in compliance and regulation differences mandates by each unique state, brokerage association, and MLS organization.
