Rental Chatbots Get It Wrong: Who Pays When Renters Believe Them

Looking for a flat used to mean filters. You set a price ceiling, ticked a box for parking, drew a rough circle on a map and scrolled. Now, increasingly, you talk. On 6 October 2026, Palo Alto Online published a report by Angela Swartz on how artificial intelligence is changing the apartment search in Silicon Valley's backyard, and the picture it painted was of a market where the listing, the questions and the answers are all starting to pass through software that speaks in fluent sentences.

The paper described rental platforms that have turned search into dialogue. Zumper's assistant, Zoe, uses information from a property's Zumper listing to answer questions about an apartment, its community and the surrounding neighbourhood, and helps schedule tours. RentCafe launched its assistant, Ren, in August. Apartments.com introduced conversational search in June. On the landlord side, the report noted that Essex Property Trust uses an AI agent named Eva to greet prospective renters through the websites of complexes including 777 Hamilton in Menlo Park and Radius in Redwood City.

Then there was the renter. A 23-year-old former investment banker told the paper he had used ChatGPT and Claude to compile property management websites and a shortlist of homes matching his price and location. “It did a pretty good job of listing out apartments that fell within those parameters,” he said. “Sometimes it would include things outside of that.” In the end, the report noted, he found an apartment “the old-fashioned way: through word of mouth.”

Zumper's chief technology officer told the paper that “the way people are searching for apartments is changing”, and said the company designed Zoe to minimise hallucinations, the term for instances in which an AI system produces inaccurate or made-up information. But, as the report put it, large language models can still make mistakes. And California's Department of Real Estate has warned that licensees remain responsible for information produced by AI.

That combination of claims raises three questions. What happens to the search for a home when every layer of it is mediated by systems that can state facts about a property that are not true? Who is responsible to a renter who acts on what a system says? And what should a person be entitled to know about how the homes put in front of them were chosen?

A Small Story With a Very Large Footprint

It is worth being precise about what the Palo Alto Online report shows and what it does not. It does not document a renter who lost money because a chatbot lied. It does not show that Zoe, Ren, Eva or Apartments.com's assistant has given anyone a false answer. Its one first-hand account from a renter is modest: a general-purpose chatbot sometimes ignored his limits, and he ended up finding his home through word of mouth.

That modesty is the point. A listing over budget or outside the chosen area is easy to spot and discard. The harder problems are the errors a renter cannot see: an answer that a building allows dogs when it does not, that a unit includes parking when it does not, that a fee is included when it is not, or that a neighbourhood is “a good fit” for reasons that have more to do with who the renter appears to be than with what they asked for.

The same report also described the other side of the counter. A Palo Alto-based property manager told the paper that he had set up a database allowing an AI tool to answer questions and check property managers' schedules to arrange tours, and that he had also seen property managers use the free version of ChatGPT to ask whether an issue needed to be brought to a lawyer or whether a new policy applied to their properties. The machine is answering on both sides of the conversation.

Every Major Portal Now Wants to Talk

The shift the paper described is not a local curiosity. It is an industry-wide bet that the search box is dying.

Zumper first announced Zoe in December 2024. HousingWire reported at the time that it could answer “virtually any renter question” about a property or its neighbourhood, from pet policies to local hotspots, and could request tours and submit leads to leasing teams. In February 2026, Zumper launched a rental search app inside ChatGPT, allowing users to type requests such as “show pet-friendly rentals under $2,500 with in-unit laundry” and receive listings without leaving the chat. In announcing it, the company cited its own survey finding that the share of renters using AI tools during their home search had risen from 4.4 per cent in 2024 to 9.8 per cent in 2025.

RentCafe, owned by the property software firm Yardi, published a feature spotlight on 20 August 2026 explaining that its former AI Mode was now called Ren. “It's an ongoing, two-way conversation,” the post said. “Instead of one search and a list of results, Ren actively guides you to apartments.” It also made a striking promise: “Ren never makes up a property, a price or an amenity. Every detail is sourced directly from trustworthy, up-to-date listings.”

CoStar Group launched Apartments.com Ai on 16 June 2026, describing it as replacing traditional filters and keyword searches with real-time conversation. “Finding an apartment is one of the most important decisions people make, yet the search process has remained largely unchanged for years,” said Andy Florance, CoStar's founder and chief executive. “Apartments.com Ai represents a fundamental shift from searching listings to having a conversation with a trusted rental advisor.”

Behind the portals sit the leasing agents. EliseAI, which sells assistants that answer prospective tenants by text, email and phone and book tours, announced on 29 September 2026 that it had raised $350 million at a $4 billion valuation, and said its platform now powers roughly one in six US apartments.

A renter may now ask a general chatbot for a shortlist, question a portal's branded assistant about a listing, and then be greeted on the building's website by a leasing agent that is also software. At no point is a human obliged to be the one who answers.

Notice the industry's vocabulary: “rental advisor”, “trusted”. It describes a relationship, not a database query. A filter that returns a listing is a tool. An adviser who tells you a flat is right for you is making a representation.

Why Fluent Software Says Things That Are Not True

Why would a system confidently state something false about a property? The short answer is that large language models are built to produce plausible text, and plausibility is not the same as truth.

In September 2025, researchers at OpenAI and Georgia Tech, Adam Tauman Kalai, Ofir Nachum, Santosh Vempala and Edwin Zhang, published a paper titled “Why Language Models Hallucinate”. Their argument was blunt: language models hallucinate because their training and evaluation procedures reward guessing over acknowledging uncertainty. If a benchmark awards a point for a correct answer and nothing for “I don't know”, a model that always guesses will outscore one that admits ignorance. The authors argued that fixing this requires changing how mainstream benchmarks are scored, not merely adding more hallucination tests.

Rental platforms try to contain the problem by grounding their assistants in listing data. Zoe draws on the property's Zumper listing; Ren, according to RentCafe, on active RentCafe listings; Apartments.com Ai on CoStar's multifamily data. Grounding narrows the space in which a model can invent things. It does not eliminate it.

Grounding fails in at least three ways that matter to renters. The first is that the source can be wrong or stale. A listing that says “pets welcome” when the landlord changed the policy last month will produce a confident, grounded and incorrect answer. The second is that the model can misread or over-generalise from what it retrieves, turning “street parking available” into “parking included”. The third is the question that has no answer in the listing at all. Is this area quiet at night? Is the building safe? Will the landlord accept my housing voucher? A system tuned to be helpful may fill the gap rather than say it does not know.

General-purpose chatbots, the kind the young renter used, sit further still from the data, summarising web pages or relying on training data that may be months old. His complaint is the mildest form of the problem, and a reminder that a model can appear to follow instructions while quietly ignoring them.

California Puts the Liability on the Licensee

California's real estate regulator has been unusually direct about who owns these errors. On 17 March 2026, the Department of Real Estate published an advisory titled “Artificial Intelligence in California Real Estate: Opportunities, Risks, and Compliance Considerations for Licensees”. Among the uses it lists are AI-powered chatbots and automated lead scoring to manage enquiries. Its message about responsibility leaves little room for doubt.

“If an AI tool generates inaccurate information, misleading advertising, or improper communications with consumers,” the advisory says, “responsibility under current law rests with the licensee and their responsible broker, not the technology provider.” It adds that “AI outputs should be reviewed for accuracy and approved by the licensee before being relied upon in transactions or consumer communications”, and that licensees using AI to draft listing descriptions, emails or online advertisements “must independently verify factual claims, including claims involving property features, pricing, availability, and potential uses”. It reminds licensees that “all real estate advertising in California must be truthful, accurate, and not misleading”, and points to section 10176 of the Business and Professions Code, which allows disciplinary action for substantial misrepresentation. It even advises keeping records “showing how AI outputs were reviewed and verified for accuracy and compliance with law.”

The state has also begun legislating against one specific kind of machine-made falsehood: the doctored photograph. Assembly Bill 723, which took effect on 1 January 2026, requires real estate brokers and salespeople who use digitally altered images in advertising to say so and to provide access to the original image. Routine adjustments such as lighting and colour are excluded. In September 2026, according to the California legislature's records, Assembly Bill 2025 was chaptered, extending a similar disclosure duty to advertisements for rental property. Its stated rationale was that virtual staging and AI tools allow anyone marketing a rental to manipulate or fabricate images showing features that are misleading or non-existent.

There is an older California law that also bears on the conversation. Since 1 July 2019, section 17941 of the Business and Professions Code has made it unlawful to use a bot to communicate with a person in California online with the intent to mislead them about its artificial identity in order to incentivise a sale of goods or services, unless the bot is clearly and conspicuously disclosed. An assistant clearly labelled as software is fine. One that passes itself off as a person, in order to close a deal, is a legal problem.

Yet the advisory has an obvious limit. It binds licensees. A real estate broker who lets a chatbot speak for them owns what it says. But a renter asking a general-purpose chatbot to find flats is not dealing with a licensee at all, and a landlord renting out its own property is not necessarily a licensee either. The more the conversation moves to software that sits outside the licensing regime, the more the regulator's clear rule is left guarding a door that renters are no longer walking through.

A Chatbot Is Still Part of the Website

When a renter acts on a wrong answer, the most useful precedent is not American. It comes from a small claims tribunal in British Columbia and a grieving airline passenger.

In February 2024, British Columbia's Civil Resolution Tribunal decided Moffatt v Air Canada. Jake Moffatt had relied on the airline's website chatbot, which told him he could apply for a bereavement fare after travelling. Another page on the same site said otherwise. Air Canada argued, in effect, that it could not be held liable for what its chatbot said. Tribunal member Christopher Rivers called this “a remarkable submission” and rejected it. “While a chatbot has an interactive component, it is still just a part of Air Canada's website,” he wrote. “It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot.” He found negligent misrepresentation and ordered the airline to pay damages, interest and fees.

The ruling binds nobody outside British Columbia's small claims system, and the sum was modest. Its reasoning travels easily, though. Negligent misrepresentation does not care whether the careless sentence was typed by a leasing agent or generated by a model.

In the United States, the defence that platforms have historically relied on is Section 230 of the Communications Decency Act, which protects online services from being treated as the publisher of information provided by someone else. But a generative assistant does not merely host a landlord's listing. It composes new sentences. The closest precedent comes from housing itself. In 2008, the Ninth Circuit Court of Appeals, sitting en banc, held in Fair Housing Council of San Fernando Valley v Roommates.com that the website was not fully immune for content it helped develop, specifically mandatory questions that required users to disclose sex, family status and sexual orientation. When a service itself produces content that runs foul of the law, the court reasoned, it cannot shelter behind immunity designed for third-party speech. How far that logic extends to AI-generated answers about rental property has not been settled by the courts. It is an uncomfortable starting point for any platform arguing that its assistant's words belong to someone else.

Lay out the parties involved in a single conversational search and the problem of responsibility becomes visible.

There is the landlord or property manager, who supplies the listing and may run their own AI leasing agent. There is the listing platform, which ingests that data, wraps it in a branded assistant and decides which homes to surface. There is the company that built the underlying language model, which the platform may license. And increasingly there is a general-purpose chatbot, which may sit in front of all of them, summarising pages it found on the web.

Each can point at another. The landlord can say the platform's assistant misread the listing. The platform can say the landlord's data was wrong. The model provider's terms of service warn that outputs may be inaccurate. A general-purpose chatbot is not a licensee, not a landlord and not a listing service.

The renter, meanwhile, holds the risk: the application fee, the notice given on a current home, the signed lease. The Federal Trade Commission's December 2025 settlement with Greystar, one of the largest property managers in the United States, shows how much money can rest on what a renter is told up front. The FTC and Colorado's attorney general alleged that Greystar advertised base rents without disclosing mandatory monthly fees, so that many renters paid non-refundable application fees before learning the true price. Greystar agreed to pay $24 million, of which $23 million was earmarked for consumer refunds, and to disclose the total monthly price including mandatory fees whenever it advertises or quotes rent. FTC chairman Andrew Ferguson directed staff to begin the process of proposing a rule on unfair or deceptive fees in rental housing.

That case did not involve AI, but the principle carries: a misleading statement that a renter relies on is actionable however it was produced. The FTC said as much in September 2024, when it announced Operation AI Comply, a sweep against companies using AI in deceptive ways. “The FTC's enforcement actions make clear that there is no AI exemption from the laws on the books,” said Lina Khan, then the agency's chair.

The sensible answer to the question of responsibility is therefore not mysterious. Whoever chooses to put an AI system in front of a renter, and profits from doing so, should own what it tells that renter. That is the DRE's position for licensees, the Air Canada tribunal's position for businesses and the FTC's position for advertisers. What is missing is a rule that covers the gaps between them: the portal that is not a broker, the general chatbot that is not an advertiser, and the property manager who is neither.

Steering Without Anyone Meaning To

The most serious risk in conversational search is not a wrong fact. It is a pattern.

Under the federal Fair Housing Act, steering, guiding a homeseeker towards or away from neighbourhoods because of race or another protected characteristic, is illegal. Traditionally it required a person to do the steering. A conversational assistant can now do it at scale, without intending to, simply by reproducing patterns in the data it learned from.

Researchers have already tested this. In a paper presented at the ACM conference on Equity and Access in Algorithms, Mechanisms, and Optimization in October 2024, Eric Liu, Wonyoung So, Peko Hosoi and Catherine D'Ignazio of MIT audited GPT-4's housing recommendations using 168,000 prompts for renting and buying in the ten largest majority-minority cities in the United States, varying characteristics such as race, family status and source of income. They found evidence of “racial steering”, “default whiteness” and the steering of minority homeseekers towards neighbourhoods with lower opportunity indices, and concluded that such systems should be audited before deployment on real estate platforms.

A follow-up study, posted in June 2026 by Hana Samad, Trung Lam, Christoph Mügge-Durum and Michael Akinwumi, working with the National Fair Housing Alliance's responsible AI programme, framed the problem in exactly the terms the Palo Alto report raises. Large language models, they wrote, are “rapidly assuming an intermediary role in housing search through the integration of listing platforms within conversational interfaces”. Auditing seven open-weight and closed-source models across four US cities, they found that steering was an emergent behaviour that varied by model and by prompt. When users added lifestyle preferences, some models began steering where they had not before, suggesting the same stated preference could be interpreted differently depending on the user's perceived racial identity.

Industry players know this. In May 2024, Zillow released an open-source Fair Housing Classifier designed to detect questions in AI-powered real estate conversations that could lead to discriminatory responses and so reduce the risk of illegal steering. It was a sensible step. It is also voluntary, and renters have no way to know whether the assistant they are talking to uses anything like it.

The legal history here is not hypothetical. In June 2022, the US Department of Justice settled its first case challenging algorithmic discrimination under the Fair Housing Act, against Meta. The government alleged that Meta's advertising system used algorithms that relied in part on characteristics protected under the Act to decide who saw housing ads. Meta agreed to stop using its “Special Ad Audience” tool for housing ads and to build a Variance Reduction System to bring the audiences who actually saw housing ads closer to the eligible audiences by sex and estimated race or ethnicity, under court oversight that ran to 27 June 2026. “This groundbreaking resolution sets a new standard for addressing discrimination through machine learning,” said Damian Williams, then the US Attorney for the Southern District of New York.

Algorithms that decide who gets a home have been tested in court too. In November 2024, a federal judge approved a settlement under which SafeRent Solutions, a tenant-screening company, agreed to pay more than $2.2 million and to stop including its score on screening reports in certain cases, including where the applicant uses a housing voucher. The plaintiffs, led by Mary Louis, had alleged that the scoring algorithm disproportionately harmed Black and Hispanic applicants and voucher holders. SafeRent did not admit wrongdoing. “Everything is based on numbers,” Louis told the Associated Press. “You don't get the individual empathy from them.”

A conversational assistant is not a tenant screener, but it shares the structural risk. It decides which homes a renter sees, in what order and with what commentary. If its suggestions differ by who the user appears to be, the effect is the same as human steering, and far harder to detect.

Washington Steps Back as the Software Steps In

Just as these systems are spreading, federal oversight in the United States is receding.

In May 2024, the Department of Housing and Urban Development issued guidance explaining how the Fair Housing Act applies to tenant screening and to housing advertising delivered through digital platforms, including when AI and algorithms do the work. On 17 September 2025, HUD's Office of Fair Housing and Equal Opportunity withdrew a list of guidance documents it said should no longer be in effect, citing deregulatory executive orders. The digital platforms advertising guidance was among them.

In January 2026, HUD proposed removing its regulations on disparate impact liability, the theory that lets plaintiffs challenge practices that are neutral on their face but discriminatory in effect, arguing that such questions should be left to the courts. The proposal had not been finalised by early October 2026. Disparate impact still exists in case law; the Supreme Court recognised it under the Fair Housing Act in 2015. But it is precisely the theory most suited to challenging an algorithm, because algorithms rarely intend anything.

Elsewhere, enforcement has focused on what software does to the price of a home rather than what it says about one. In November 2025, the Justice Department reached a proposed settlement with RealPage over allegations that its pricing software used competitors' non-public data to help landlords align rents. A federal judge in North Carolina approved it in May 2026, barring RealPage from using that confidential data to help set rents. RealPage did not admit wrongdoing. The case shows regulators can reach into the code that shapes housing markets. It also shows how long it takes.

Scammers Already Speak the Language

There is one more party in the conversation that renters cannot afford to forget: the fraudster.

In December 2025, the FTC reported that consumers had filed nearly 65,000 reports of rental scams since 2020, with losses of about $65 million. About half of those who reported a rental scam in the 12 months to June 2025 said it began with a fake advert on Facebook. People aged 18 to 29 were three times more likely than other adults to report losing money. The FTC noted that fake listings “can often look very real and copy information from legitimate listings but are posted with the scammer's contact information on different sites.”

The FBI's Internet Crime Complaint Center recorded 12,368 real estate fraud complaints and $275.1 million in reported losses in 2025, a category that includes rental fraud, up sharply from the year before.

Conversational search raises the stakes. A chatbot that searches the web cannot always tell a cloned listing from a real one, and may present both in the same confident tone. And renters are being trained to trust a friendly voice that answers instantly. The young renter in Palo Alto noticed when results drifted. Not everyone will.

What Europe and Britain Already Demand

Other jurisdictions offer partial answers to the question of what a renter should be told.

The European Union's Digital Services Act requires online platforms that use recommender systems to set out, in plain and intelligible language, the “main parameters” those systems use, including the criteria most significant in deciding what is suggested to a user and the reasons for their relative importance. The largest platforms must offer at least one option not based on profiling. Applied to rental search, that principle would mean a renter could find out why these homes, in this order, rather than others.

The EU's AI Act adds a simpler duty. Since 2 August 2026, Article 50 has required that people interacting directly with an AI system be informed that they are doing so, unless it is obvious. Fines for breaching the Act's transparency obligations can reach €15 million or 3 per cent of worldwide annual turnover.

Britain has taken a different route, focusing on what listings must contain rather than how software presents them. National Trading Standards' Estate and Letting Agency Team has issued guidance on “material information” in property listings; for lettings, this includes the rent, any deposit and the council tax band. The Renters' Rights Act 2025, whose first phase came into force in England on 1 May 2026, made it illegal for landlords and agents to invite or accept bids above the advertised rent, and to discriminate against prospective tenants because they have children or receive benefits, including through advertising. Since April 2025, the Digital Markets, Competition and Consumers Act 2024 has allowed the Competition and Markets Authority to enforce consumer law directly and to fine businesses up to 10 per cent of global turnover.

None of these laws was written for a chatbot that recommends flats. But together they sketch something the US lacks: a baseline of what must be disclosed, a ban on certain kinds of exclusion at the advertising stage and a regulator with the power to act quickly.

What a Renter Should Be Entitled to Know

If the conversation is the new front door to housing, renters need rights that match. Drawing together the DRE's advisory, the Air Canada ruling, the fair housing case law and the European rules, a workable set begins to emerge.

First, renters should know they are talking to a machine and on whose behalf: a portal, a landlord or neither, and whether suggested homes include paid placements. California's bot law and the EU AI Act already point this way.

Second, renters should know where each fact came from. If an assistant says a unit allows dogs, it should show the dated listing text it relied on. Grounded systems already retrieve this; displaying it costs little.

Third, renters should be told why homes were suggested and ranked. The DSA offers a model: the main criteria, their relative importance and an option to search without profiling.

Fourth, assistants should be built to say “I don't know”. Questions about safety, schools or the “character” of a neighbourhood are where steering risk is highest and listing data thinnest. The OpenAI researchers' case for rewarding abstention applies to product design as much as to benchmarks.

Fifth, the business that deploys an assistant should be liable for what it says, as if a human employee had said it. That is already the DRE's rule for licensees and the Air Canada tribunal's reasoning for businesses. Extending it explicitly to listing platforms and property managers who are not licensees would close the gap the conversational shift has opened.

Sixth, conversational housing assistants should be audited for steering before deployment and at regular intervals, with results published. The MIT and NFHA studies show that such audits are feasible. The Meta settlement shows that algorithmic delivery can be measured against a fairness benchmark and corrected under oversight.

Finally, renters should be able to keep a transcript of the conversation. If an answer led someone to pay a fee or sign a lease, it is the evidence. The DRE already tells licensees to keep records. Renters deserve the same.

Word of Mouth Still Beats the Machine

It is easy to read the Palo Alto Online report as a story about technology that does not quite work yet. That misses the point. These assistants will improve, and the young renter's complaint may soon feel quaint.

The deeper issue is structural. Housing search used to involve a listing written by a landlord, a portal that displayed it and a renter who read it. Each statement had an author. In conversational search, the sentence a renter acts on may be composed on the fly by a system nobody fully controls, drawing on data nobody fully checked, ranked by criteria nobody disclosed. The industry describes that system as an adviser. The law has not yet decided whether it should be treated as one.

California's regulator has answered half the question: if a licensee lets AI speak, the licensee answers for it. Courts have hinted at more: a chatbot is part of the business that runs it. Europe and Britain have begun to set out what a person must be told. What nobody has yet written down is a single rule that follows the renter rather than the type of company they happen to be talking to.

Until that exists, the renter in Palo Alto may have found the most reliable method of all. He asked the machines, noticed where they drifted, and in the end found his home through word of mouth. The question for regulators and platforms is whether the next generation of renters will have to do the same.

References

  1. Swartz, Angela. “How AI is changing the apartment search: Is it working?” Palo Alto Online, 6 October 2026. https://www.paloaltoonline.com/technology/2026/10/06/how-ai-is-changing-the-apartment-search-is-it-working/
  2. Andrews, Jeff. “Zumper launches AI assistant Zoe.” HousingWire, 3 December 2024. https://www.housingwire.com/articles/zumper-ai-assistant-zoe/
  3. Inman. “Zumper introduces Zoe, latest AI for aspiring renters.” Inman, 10 December 2024. https://inman.com/2024/12/10/zumper-introduces-zoe-latest-ai-for-aspiring-renters
  4. National Mortgage Professional. “Zumper Debuts Real-Time Rental Search App Inside ChatGPT.” February 2026. https://nationalmortgageprofessional.com/news/zumper-debuts-real-time-rental-search-app-inside-chatgpt
  5. Petrut, Florin. “Meet Ren, your apartment search assistant on RentCafe.com.” RentCafe Blog, 20 August 2026. https://www.rentcafe.com/blog/renting/rentcafe-feature-spotlight-ai-mode/
  6. Mann Report. “CoStar Group Launches Apartments.com Ai.” Mann Report, 16 June 2026. https://www.mannpublications.com/mannreport/2026/06/16/costar-group-launches-apartments-com-ai/
  7. Gleeson, Cailey. “EliseAI banks $350M, hits $4B valuation for accelerated new product development.” Fierce Healthcare, 29 September 2026. https://www.fiercehealthcare.com/finance/eliseai-banks-350m-hits-4b-valuation-accelerated-new-product-development
  8. Kalai, Adam Tauman; Nachum, Ofir; Vempala, Santosh S.; Zhang, Edwin. “Why Language Models Hallucinate.” arXiv:2509.04664, 4 September 2025. https://arxiv.org/abs/2509.04664
  9. California Department of Real Estate. “Artificial Intelligence in California Real Estate: Opportunities, Risks, and Compliance Considerations for Licensees.” Advisory, 17 March 2026. https://www.dre.ca.gov/Licensees/Advisories/Advisory_2026_03_17_AI_in_California_Real_Estate.html
  10. California Legislature. AB 723 (2025), Real estate: digitally altered images: disclosure. CalMatters Digital Democracy. https://calmatters.digitaldemocracy.org/bills/ca_202520260ab723
  11. California Legislature. AB 2025 (2026), Tenancy: digitally altered images: disclosure, Chapter 572, Statutes of 2026. CalMatters Digital Democracy. https://calmatters.digitaldemocracy.org/bills/ca_202520260ab2025
  12. Perkins Coie. “I Am Robot: California's New Law Requires Disclosure of Use of Bots.” https://perkinscoie.com/insights/blog/i-am-robot-californias-new-law-requires-disclosure-use-bots
  13. Moffatt v Air Canada, 2024 BCCRT 149, Civil Resolution Tribunal of British Columbia, 14 February 2024; reported in CTV News, “Air Canada's chatbot gave a B.C. man the wrong information. Now, the airline has to pay for the mistake,” February 2024. https://bc.ctvnews.ca/air-canada-s-chatbot-gave-a-b-c-man-the-wrong-information-now-the-airline-has-to-pay-for-the-mistake-1.6769454
  14. Fair Housing Council of San Fernando Valley v. Roommates.com, LLC, 521 F.3d 1157 (9th Cir. 2008) (en banc); Congressional Research Service, “Liability of Interactive Computer Service for Violating the Fair Housing Act,” RS22668. https://www.everycrsreport.com/reports/RS22668.html
  15. Federal Trade Commission and Colorado Attorney General. “Weiser, FTC announce $24M settlement with Greystar.” 2 December 2025. https://coag.gov/press-releases/weiser-ftc-announce-24m-settlement-with-greystar/
  16. Davis Polk. “FTC announces new enforcement initiative targeting deceptive AI practices.” September 2024. https://www.davispolk.com/insights/client-update/ftc-announces-new-enforcement-initiative-targeting-deceptive-ai-practices
  17. Liu, Eric; So, Wonyoung; Hosoi, Peko; D'Ignazio, Catherine. “Racial Steering by Large Language Models: A Prospective Audit of GPT-4 on Housing Recommendations.” EAAMO '24, ACM, October 2024. MIT DSpace. https://dspace.mit.edu/handle/1721.1/157628
  18. Samad, Hana; Lam, Trung; Mügge-Durum, Christoph; Akinwumi, Michael. “The Geography of Algorithmic Judgment: LLM Intermediaries, Place Identity, and Racial Steering in Housing Search.” arXiv:2606.06694, 4 June 2026. https://arxiv.org/abs/2606.06694
  19. Zillow Group. “Zillow providing open-source technology to promote fair housing in AI-powered real estate conversations.” Press release, 21 May 2024. https://zillow.mediaroom.com/2024-05-21-Zillow-providing-open-source-technology-to-promote-fair-housing-in-AI-powered-real-estate-conversations
  20. US Department of Justice. “Justice Department Secures Groundbreaking Settlement Agreement with Meta Platforms, Formerly Known as Facebook, to Resolve Allegations of Discriminatory Advertising.” 21 June 2022. https://www.justice.gov/archives/opa/pr/justice-department-secures-groundbreaking-settlement-agreement-meta-platforms-formerly-known
  21. US Attorney's Office, Southern District of New York. “United States Attorney Implements Groundbreaking Settlement With Meta Platforms, Inc.” January 2023. https://www.justice.gov/usao-sdny/pr/united-states-attorney-implements-groundbreaking-settlement-meta-platforms-inc-formerly
  22. Bedayn, Jesse. “Class action lawsuit on AI-related discrimination reaches final settlement.” Associated Press, 20 November 2024, as published by Fortune, 21 November 2024. https://fortune.com/2024/11/21/renter-scoring-saferent-million-settle-case-algorithm-discriminating-race-income/
  23. US Department of Housing and Urban Development. “HUD Issues Fair Housing Act Guidance on Applications of Artificial Intelligence.” Press release 24-098, 2 May 2024. https://archives.hud.gov/news/2024/pr24-098.cfm
  24. Bilowich, Juliana. “HUD Withdraws Wide-Ranging Fair Housing Policies.” LeadingAge, 23 September 2025. https://leadingage.org/hud-withdraws-wide-ranging-fair-housing-policies/
  25. Ballard Spahr. “HUD proposes to remove its Fair Housing Act disparate impact rule.” Consumer Finance Monitor, 14 January 2026. https://www.consumerfinancemonitor.com/2026/01/14/hud-proposes-to-remove-its-fair-housing-act-disparate-impact-rule/
  26. Baker McKenzie. “United States: DOJ settles RealPage case.” December 2025. https://www.bakermckenzie.com/en/insight/publications/alerts/2025/12/united-states-doj-settles-realpage-case
  27. Lowe, Jonathan; Praats, Michael. “RealPage settles with DOJ over rent pricing scheme; NC case still active.” WSOC-TV, 20 May 2026. https://www.wsoctv.com/news/local/realpage-settles-with-doj-over-rent-pricing-scheme-nc-case-still-active/AKJI2ZH7R5GOHIK3JGVIPKM4HM/
  28. Federal Trade Commission. “FTC Analysis Shows Consumers Have Lost Millions to Rental Scams.” Press release, 22 December 2025. https://www.ftc.gov/news-events/news/press-releases/2025/12/ftc-analysis-shows-consumers-have-lost-millions-rental-scams
  29. Block, Eliana. “Online real estate fraud climbed to $275M in 2025, FBI says.” National Association of Realtors, 13 April 2026. https://www.nar.realtor/news/real-estate-news/online-real-estate-fraud-climbed-to-275m-in-2025-fbi-says
  30. HousingWire. “FBI cybercrime losses: real estate fraud hits $275M.” April 2026. https://www.housingwire.com/articles/fbi-cybercrime-losses-real-estate-fraud-hits-275m/
  31. European Parliament and Council. Regulation (EU) 2022/2065 (Digital Services Act), Articles 27 and 38. Official Journal of the European Union, 27 October 2022. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32022R2065
  32. European Parliament and Council. Regulation (EU) 2024/1689 (Artificial Intelligence Act), Articles 50 and 99. https://eur-lex.europa.eu/eli/reg/2024/1689/oj
  33. National Trading Standards Estate and Letting Agency Team. “Material information in property listings: Part A.” https://www.nationaltradingstandards.uk/uploads/Material%20information%20-%20Part%20A%20-%20updated%20July%202022.pdf
  34. UK Government. “Renters' Rights Act: an overview for landlords.” GOV.UK. https://www.gov.uk/guidance/renters-rights-act-an-overview-for-landlords
  35. DLA Piper. “CMA's new consumer enforcement regime comes into force in April 2025.” 7 April 2025. https://www.dlapiper.com/insights/publications/2025/04/cmas-new-consumer-enforcement-regime-comes-into-force-in-april-2025

Tim Green

Tim Green UK-based Systems Theorist & Independent Technology Writer

Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at smarterarticles.co.uk, challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship.

His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities.

ORCID: 0009-0002-0156-9795 Email: tim@smarterarticles.co.uk

Listen to the free weekly SmarterArticles Podcast

Support What We Do

Discuss...