How AI and a Major Settlement Are Shaking Up Real Estate in the USA

A legal change and AI are changing the rules for real estate in the USA

The real estate USA market is being rewired by two forces: a legal settlement that altered how buyer-agent commissions are disclosed and collected, and fast-moving artificial intelligence tools that are inserting automation into every step of a transaction. These shifts are not theoretical. They are already changing how people buy, sell, and finance homes — and they can cut tens of thousands of dollars from the typical cost of a residential deal.

I have tracked technology shifts in property markets for years, and this one is significant because it combines regulatory pressure with commercially viable AI products. The result is a concrete set of options for buyers and sellers, and new operational risks for lenders and brokers.

What changed in August 2024 and why it matters

In August 2024 a settlement stemming from antitrust litigation involving the National Association of REALTORS (NAR) changed long-standing industry customs. One practical outcome is that a buyer must often sign a contract with a buyer's agent before touring a house, and agents commonly stipulate the full 3% buyer-agent commission when they sign those agreements.

Why that matters:

• For decades, the U.S. real estate market used a seller-paid commission model that often aggregated to 6% of the sale price (typically 3% to the seller's agent and 3% to the buyer's agent).
• The settlement removes a degree of opacity and puts pressure on how commissions are negotiated and disclosed.
• That pressure creates an opening for lower-cost, technology-enabled models that split the buyer-agent fee differently or refund parts of it to buyers.

The settlement did not eliminate agents. It did, however, make the buyer-agent relationship more explicit and provided the legal conditions that startups and established firms are now exploiting with AI.

How AI is being used to buy houses: Homa and the 1% model

One of the earliest and most discussed entrants is Homa, an AI-led buyer's agent that is live in Florida and preparing to expand to Texas within weeks, with California and Georgia expected by the end of the year. Homa combines automated analysis with a network of local agents who conduct in-person showings.

Key points about the Homa model:

• Homa positions itself as an AI-powered buyer's agent that charges 1% of the buyer's commission and refunds the remainder to the buyer when the seller pays the commission.
• Example from Homa's public explanation: on a $500,000 home with a 3% buyer-agent commission ($15,000), Homa keeps $5,000 and refunds $10,000 to the buyer. That refunded amount can be used to lower closing costs or buy down mortgage interest.
• The company uses an "Uber-style" network of local showing agents for walkthroughs, which buyers have told Homa they like because it avoids putting pressure on a single agent's time.
• Homa's in-house negotiators are paid a flat fee, which the company says reduces the incentive to push for a higher sale price.

From an investor or buyer perspective, that structure translates to immediate, visible savings on transaction costs. It also shifts the buyer's relationship with their agent: less personal, more transactional, and more reliant on platform-provided negotiation and documentation support.

Sellers and DIY AI tools: how to sell with AI support

On the selling side, AI already has practical use cases that ordinary homeowners can access. One high-profile example is a Florida seller who used ChatGPT to plan repairs, write marketing materials, and assemble an MLS listing. He reported several offers within 72 hours and a signed contract in five days.

Low-cost services and what they offer:

• Companies like Beycome allow sellers to list on MLS for a flat fee. The service advertises a basic MLS listing for $99, with title and escrow services available for an additional $99. Beycome claims sellers can save up to 6% by avoiding traditional commissions.
• ChatGPT and similar large language models are used to draft listing copy, prepare seller marketing, and suggest cost-effective improvements that can raise saleability.
• AI-driven virtual staging and photo enhancement tools can present a more polished listing quickly; buyers should remember that some enhancements can be misleading and may create expectancy mismatches at inspection.

For sellers who are comfortable managing details and legal paperwork, these tools can cut the commission bill sharply. The trade-off is more personal time spent handling escrow, title, and negotiation — or paying fixed-fee professionals for those functions.

AI in mortgage lending: Tinman and why accuracy matters

AI is not just changing how homes are bought and sold; it is altering how mortgages are underwritten. Better, a well-known digital mortgage lender, has for years used a machine-learning engine called Tinman to automate underwriting and process steps. According to Better's CEO Vishal Garg, the company layered large language model orchestration on top of Tinman to manage process flow and communications.

Important data points:

• Better claims its cost to produce a mortgage is close to $2,000, compared with an industry average effort closer to $10,000.
• Garg says conventional large language models make calculation errors at a high rate if used naively (he cites roughly 50% error rates for raw LLM arithmetic and even higher error rates for complex, recursive calculations). For mortgages, a near-zero error rate is required.
• Better says its Tinman engine does the math and rule application, while LLMs are used for orchestration, and that this approach yields "zero errors" in the math and rules application.

What this means for buyers and investors:

• Faster, cheaper mortgages can lower closing costs and reduce the time risk between offer and closing.
• Automated underwriting can increase throughput for lenders, but it introduces operational risk if models are not correctly validated and audited.
• Borrowers should continue to get written, lender-certified calculations for debt-to-income (DTI) and other key ratios instead of relying solely on an app-generated preapproval screenshot.

Who wins and who loses: incumbents, agents, buyers, and platforms

This is not a tidy disruption where one side wins and another disappears. I expect a prolonged, uneven transition.

Winners likely include:

• Consumers who are comfortable with guided technology and willing to replace some personalized services with platform workflows; they can capture direct refunds or lower fees.
• Tech-first brokerages and mortgage lenders that invest in accurate machine learning and process automation.

Potential losers or losers-at-risk include:

• Traditional agents who resist fee compression and remain paid only on percentage-based commissions. The industry-wide inertia is strong because many agents are independent contractors (1099) and can choose to work where commission structures favor them.
• Buyers or sellers who assume AI eliminates professional oversight; inadequate legal or title work can produce costly mistakes.

Agents will not disappear. Licensed agents still must perform negotiation and execute contracts. But their role is likely to shift from a commission-maximizing salesperson to a negotiator and compliance guardian. Flat fees and salaried models for certain duties are becoming tangible options.

Practical advice for buyers and investors — how to use AI without getting burned

Here is how to approach AI-enabled property services in the current U.S. market.

• Compare total costs. Look beyond headline percentages. If a platform refunds part of the commission, confirm whether the refund requires concessions on price or specific contract terms.
• Verify calculations. For mortgage offers, insist on lender-signed calculations for DTI, loan-to-value (LTV), and closing costs. AI-generated numbers should be treated as provisional.
• Check licensing and local rules. Real estate is state-regulated.