If you’ve ever used an online platform— Zillow, Realtor.com, Compass.com, Redfin—you’re already aware that those platforms are steering you towards listings that they monetize; but at least you still get to choose your neighborhood.
AI chatbots have taken steering to an entirely new level, and it’s because they’re trying to help.
Steering is the illegal practice of directing buyers or renters toward or away from certain neighborhoods based on who they are, not what they want.
Federal regulations define it as acting to exclude or limit persons from neighborhoods because of race, color, religion, sex, familial status, disability, or national origin. The Fair Housing Act has prohibited it since 1968.
Use a Chatbot? You Are Being “Guided”
You are not building these tools. You are using them, and the whole point of the machine is to steer you somewhere. AI chatbots want to make you happy so you keep giving them your information.
The largest brokerages have spent years steering sellers off the open market entirely, marketing homes through private networks before the public ever sees them. The portals steer from the other side, curating which homes you see based on what generates clicks and leads.
Steering
Chatbots rarely say anything openly discriminatory—Grok, “hold my beer”. They learn from training data full of historical bias and repeat it in the polite language of recommendations.
If a buyer mentions young children and the system serves only suburban houses, or 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 trying to be helpful.
The Platforms Build These Bots
The platforms are building these bots themselves. 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.
Washington has a way of catching up to the industry's new tricks, and this time it did not take long. The agency that polices housing discrimination just fired back, and the answer was not what the platforms expected.
HUD Claps Back
ON April 24, 2026, HUD issued a Dear Colleague letter stating that sharing crime and school information with clients is not, by itself, a violation when it is provided consistently to everyone. That clarity comes with a trap.
The trap is that the letter removed the ambiguity that used to protect everyone. If a chatbot serves school data only to families with children, or uses "safe neighborhood" as code to route people toward wealthier, less diverse communities, it has crossed into selective delivery.
The letter's "intentional discrimination" standard sounds like a shield but is not, because intent under the Fair Housing Act can be proven by pattern and courts treat racially coded comments about crime and schools as evidence of intent.
The pattern is the intent, 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, and HUD just wrote the test for spotting it. The letter is only guidance, not law, so it can be withdrawn, courts are not bound by it, and stricter state and local fair housing rules remain in force.
Practical Examples
Wondering how this affects you and your clients?
Source of income is not one of the seven federally protected classes, but many states, including California, prohibit discrimination against voucher holders, so a bot that rejects Housing Choice Voucher applicants can create liability.
Disability is simpler. A prospect who asks about a service animal or a ramp is entitled to equal access and a reasonable accommodation. A bot that books tours instantly for everyone but tells a disabled prospect a human will call back is unequal treatment with a timestamp.
The Price of Steering
Steering carries a price even before the lawyers arrive. Homes sold off market typically fetch about $4,975 less than comparable homes on the MLS, and in California the gap widens to roughly $51,000. HUD penalties can reach $26,262 for a first violation and climb from there, before private lawsuits, damages, and license discipline.
For Now
AI chatbot steering is real and it is happening and there’s not much you can do to prevent it; you can mitigate it. The best way to see everything is to call an agent or go to your local MLS and access their public portal directly.
Treat every chatbot answer like a recommendation and just assume that, like with the public portals, your search results are being at best nudged and at worst steered.
