Real Estate: brokerages open their data and start metering AI

In real estate, AI stopped being a separate tool an agent buys and became something that arrives through the brokerage, the listing service and the home-search portal. Between September and October 2026, one large US brokerage let agents plug the AI assistant of their choice into their own business data, another began billing its own AI per use, and listing services started bundling AI into the membership agents already pay for. For a small firm, the question now is what it already has, and what one finished job costs.

The brokerage opens its data, and starts metering its AI

When Era Haus last looked, in Real Estate: the AI assistant is reaching the brokerage's back office, a software vendor was testing a connection that let an agent's own assistant read brokerage files. In October eXp Realty, one of the largest US brokerages, made that its policy.

On 8 October, at its annual conference, eXp launched a way for each agent to connect Claude, ChatGPT or another assistant to their own records: leads, deals in progress, sales totals and contacts (Inman and Real Estate News). It runs on MCP, a shared standard that lets an assistant read and act inside another company's software, and each agent switches it on personally. eXp calls it the first piece of a larger system due in early 2027 and has not said what it will cost.

On 20 October The Real Brokerage, a fast-growing North American brokerage with more than 36,000 agents, took the other road and launched its own AI. Leo Pro is a set of more than three dozen ready-made helpers for listing videos, follow-up with past inquiries that went quiet, transaction steps and listing presentations, and it asks the agent before any important action (Inman and HousingWire). The bigger change is the bill. Real charges by usage, counted in tokens (the units AI companies use to meter work), where most agent software charges a flat fee per person. Real has not said what a token costs.

AI inside what you already pay for

The second shift is quieter. AI is arriving bundled into the listing service: in the US, the MLS (multiple listing service), the shared database of homes for sale that local agent associations run and agents pay dues to.

On 6 October Realtor.com added an assistant to Realtor.com+, its client-collaboration platform, which agents get as a member benefit through their MLS in 21 US markets, covering more than 192,000 agents (company release and Inman). Each morning it sends the agent a summary of what every client searched, saved and asked, with suggested next steps, and drafts messages the agent approves before they go out.

On 15 September Hive MLS, which serves North Carolina, South Carolina and parts of Georgia, opened an AI property-research tool to nearly 20,000 members (Inman, 18 September). Basic reports are free; a fuller version with AI price comparisons and document summaries costs $20 a report, according to Hive's member page.

Latin America is moving the same way through the portal. QuintoAndar, Brazil's largest housing platform, took Domi, its assistant that answers partner agencies' WhatsApp inquiries and books viewings, from São Paulo to Curitiba on 8 October (Canaltech).

Most US brokerages now use some AI; most agents still do not use it every day. The National Association of Realtors' 2026 technology survey of US agents, published 22 September, found 23% use AI daily and another 25% weekly; among users, 75% use it for property descriptions. Cost and the learning curve were the two most-named obstacles.

The buyer starts the search inside the assistant

On 19 October Zillow, the largest US home-search site, launched a connector that brings its homes for sale, its automated home-value estimates and its affordability tools into Claude (Zillow release; Real Estate News, 20 October). Zillow already had an app in ChatGPT (October 2025) and a rentals service in Gemini. Each home keeps the broker's name, and the buyer is routed to Zillow's own tour requests and agent contacts. When a buyer's first search starts in an assistant, the agent who gets the call is whoever the platform behind that assistant sends it to.

The rules moved at the edges

Two enforcement steps landed in the window, both US. Since 1 September, Metro MLS in Wisconsin can fine an agent $10 a day for a digitally altered main photo until it is replaced, according to the trade site The Close. And on 15 October a US federal judge gave final approval to $359.9 million in settlements by 41 landlords and managers accused of using RealPage, a rent-pricing software company, to share competitors' private rent data (MLex). The US Justice Department's proposed settlement with RealPage limits such tools to private data at least a year old. For a landlord, "what data does my pricing tool use, and how recent is it?" is now a question to put to the vendor in writing.

What it means for you

The cost of AI in a small agency is moving from a subscription line to a usage meter, and from a separate purchase to something folded into dues and platforms you already pay. That makes the real price harder to see. It also makes the brokerage's data policy a recruiting question: an agent choosing between firms will ask whether they can bring their own assistant.

What to do about it

Before buying any AI add-on, check what your MLS, portal and transaction software already include; some agents will find they are paying twice for client summaries and property reports. When a vendor quotes usage pricing, ask for the cost of one completed job, such as a full set of listing materials or a follow-up sequence, and compare it with the seat fee you pay now. If you use AI-staged or edited photos, keep the original and label the edit on every listing. Usage pricing itself is new and unpublished: watch it, and wait before rebuilding your budget around it.