Big Real Estate’s AI Chatbots and Steering

Large real estate consumer search engine platforms have been curating, and using their own personal algorithms to monetize agent data, and listings for consumers for years.
Consumer data is now fiercely competitive and coveted by all parties, and larger conversations are happening about how algorithms impact exactly what homes consumers shown online. Is this steering? Is this a biased search are agents liable for it? In this article I will break down how algorithmic flows, and now A.I. chatbots may not be sharing the whole story with consumers.
Developing A.I. laws in states like California bring up a good point for brokers who are using A.I. to outsmart their competition: A.I. isn’t licensed.
Portals, searches, and bots
The portal era promised you the whole market. Zillow, Redfin, Realtor.com all pull from the same MLS feed, and for a while that empowered buyers; it was curation. The platforms decided which homes you saw, in what order, based on what generated clicks and leads, and the buyer never saw the listings that fell off the page. 
Curation looked neutral because everyone got the same algorithm. But the ranking was always about monetization, not about the buyer.
The largest brokerages added their own layer, steering sellers off the open market into private networks before the public ever saw the home. The portals steered from the other side, deciding which homes got seen and which got buried.
Then the platforms create chatbots to further “refine” results, which per HUD crosses the line into steering. And since departments of real estate (California specifically) have now determined that AI chatbots are “unlicensed assistant” guess who is on the hook for compliance?
Are the platforms handing over too much to A.I.?
Then the platforms discovered AI. Curating begat chatbots because a bot is curation you can talk to. Zillow launched AI Mode in March 2026, describing it as a shift from query-based search to guided intelligence, and Redfin shipped an app inside ChatGPT in February. 
The same curated feed, now delivered as a personal conversation that feels like advice. It feels helpful. That’s the problem.
Does A.I. know not to steer?
Are chatbots “steering” because the machine learned from training data full of historical bias and repeated it in the polite language of recommendations? Chatbots have access to publicly available demographics and are good at making assumptions.
A buyer mentions young children and the system serves only suburban houses. A renter asks about safe areas and gets a list that shifts with who is asking. No slurs. No red flags. Just a quiet nudge that sends people in different directions based on who they are. That is steering, and it is still steering when a machine does it. 
The Compliance nightmare for brokers
This is where it lands in the brokerage’s/broker’s lap. Steering begat broker nightmares, because the Fair Housing Act does not care who generated the sentence. It cares that it went out under your license.
The state made it explicit. The California Department of Real Estate classified AI as an unlicensed assistant and put its supervision squarely on the managing broker, stating that responsibility for AI-generated errors rests with the licensee and the broker, not the technology provider. You cannot outsource the liability to the vendor. You can only manage it.
Just to be clear, the broker/compliance officer is responsible when agents use AI. And not just in California.
States like Utah have explicitly removed the ‘the machine did it’ defense under SB 149, making brokerages strictly liable for any AI-generated misrepresentations. Colorado’s SB 24-205 classifies housing as a high-risk AI sector, legally forcing firms to run assessments and assume full compliance liability for their agents’ AI tools.
The HUD test
And HUD just sharpened the test. On April 24, 2026, the agency stated that sharing crime and school information is not, by itself, a violation when it is shared consistently and without discriminatory intent. HUD news release. The clarification removes any sort of vagueness around what constitutes intent. And intent under the Fair Housing Act can be proven by pattern.
The pattern
A family with kids gets the suburban list and school stats while a single renter with the same budget gets urban condos and a crime warning. The government already won that argument against Meta’s ad algorithms in 2022.
The machine makes the pattern at a scale no human could match. A chatbot does not steer one buyer a month. It steers thousands, in an afternoon, with every conversation logged and stored. That is the difference between an agent who made a mistake and a system built to sort people. Regulators do not need to read the code to see it. They just need to run the same question twice.
Broker exposure
The exposure to the broker costs real money. HUD penalties can reach $26,262 for a first violation and climb, before private lawsuits and license discipline.  Errors and Omissions companies are starting to pay attention to AI use in brokerages.
Some carriers are already excluding coverage for losses tied to AI output. If your E&O form does not ask about AI today, it will at renewal, and the answer may decide whether a claim is covered at all.
The real cost is not the penalty. It is the license. HUD can fine you; a plaintiff can sue you, and a carrier can drop you, but a few states can also take your livelihood over a pattern the chatbot created.
For Now
So, what does a broker do?
Go to the NAR AI policy kit, read it, and make a plan to implement an AI policy. NAR has already built the template; our memberships are still worth something! Put something together because something is better than nothing, and the policy can evolve.
Be proactive and call your E&O carrier before they call you. Tell them you have an AI policy in place. Ask whether your policy covers AI-generated content. Ask if there are any AI exclusions. Ask if their renewal requirements include an AI policy. 
Put the policy in writing this week: the tools agents may use, the human review before anything ships, the test every bot has to pass, and the logs you keep when it fails. A.I. is everywhere, and it is something that brokers cannot turn a blind eye too. Make sure your bases are covered, and seek out what specifically is happening with developing laws in your state.