AI in Real Estate: Where It's Actually Working and Where It's Not (Yet)
TL;DR: AI is reshaping real estate, but not evenly. The biggest changes today are happening in structured, behind-the-scenes work like underwriting, title, and document processing, while the highest-stakes client decisions still rely heavily on judgment, relationships, and local context. We tested the tools, read the research, and spent time in the industry where much of this technology is being built. Here is what is actually worth paying attention to.
Using AI to Sell Your Home?
Last month, a Florida homeowner used ChatGPT to sell his house without an agent, got five offers in 72 hours, and the story spread quickly. In practical terms, it was a modern FSBO (for sale by owner), but with AI making the experience meaningfully more capable. ChatGPT helped him research pricing, write his listing, navigate forms, and manage a process that previously required significantly more effort or outside expertise. What the headlines missed is what actually drove the result: when Dennis Norman of MORE, REALTORS looked closely, the five offers did not originate from AI. They came from the MLS. The seller still needed a flat-fee brokerage service to get listed. The property was likely underpriced relative to comps. And by the final walkthrough, the seller said he would probably use an agent next time. AI made the FSBO easier. The marketplace and a human agent on the other side did the rest.
The story also points to something worth watching more broadly. AI tools are widely used by buyers and sellers today, and while they are not yet a reliable substitute for experienced judgment, they do something else: they introduce more variables, surface more doubt, and make it easier to second-guess a decision mid-stream. As Inman contributor Deb Siefkin observed this week, AI is not just making it easier to attempt a transaction; it is making it easier to question one already in progress. That is worth understanding before you rely on it as your primary advisor.
The Right Use Cases for AI
Having spent time at ServiceNow building industry-specific products and workflows, a pattern becomes familiar: AI tends to deliver earliest and most reliably on work that is high-volume, tedious, and structured. The harder and slower gains come in areas that require contextual judgment, incomplete information, and reading people or situations in real time. That pattern does not mean AI stays in its lane forever. The ceiling is rising quickly. But it is a useful guide for evaluating what is genuinely changing now versus what is still aspirational.
Real estate has both kinds of work in abundance. Which is why the honest answer to "will AI replace agents" is neither yes nor no. It depends which part of the transaction you are talking about, and when.
Search Tools: A Useful Orientation, Not a Decision-Making Tool
Major portals including Zillow, Realtor.com, Redfin, and Homes.com are rapidly rolling out AI-powered search and conversational features. We tested them all with a common set of prompts in our market. What we found is that most are conversational interfaces layered on top of search infrastructure that has not fundamentally changed. They can infer filters well, translating a natural language request into bedrooms, price, and location parameters. Where they struggle is reasoning about what those results actually mean. It is also worth noting that searching for homes is not inherently a high-volume, tedious, or structured task for the individual buyer. It is personal, contextual, and already fairly simple with existing tools. AI-powered search may be solving less of a problem than it appears.
The exception worth noting is the Realtor.com integration inside ChatGPT, which lives outside the portal itself. It gave us genuine analysis: which sitting listings were worth investigating and why, and a clear take on what a $2.5 million approval actually buys in today's Redwood City market. That is closer to advisory behavior. The catch is that using it requires you to already be working in ChatGPT, which most buyers are not. For now, these tools are most useful for early orientation (understanding neighborhoods, building a budget, thinking through tradeoffs) before the real work begins.
Where AI Is Actually Delivering
A real estate transaction involves a surprising amount of work that fits exactly the profile where AI performs well: high-volume, tedious, and structured. Processing a mortgage application means reviewing dozens of pages of financial records, tax filings, and legal documents. Title clearance requires searching property history across multiple jurisdictions to identify liens, unpaid taxes, recording errors, and ownership gaps. Underwriting requires matching borrower data against a complex matrix of guidelines and risk thresholds. These are not simple tasks, but they are structured ones, and they have historically consumed enormous amounts of time and human labor. That is changing.
- Rocket Close partnered with Amazon Web Services to deploy a generative AI system that processes mortgage document packages in under two minutes. Previously, each package averaging roughly 75 pages of deeds, liens, and legal records took up to ten hours of manual work. The system now handles approximately 2,000 packages per day with around 90% accuracy, with human specialists reviewing exceptions.
- Better.com embedded its Tinman AI underwriting platform directly into ChatGPT, allowing loan officers to run borrower eligibility checks conversationally against Fannie Mae and Freddie Mac guidelines in real time. One early client reduced origination costs by 30% and doubled its business. Better's CEO has described the vision as moving mortgage decision-making from siloed internal systems into an intelligence layer accessible anywhere.
- Title automation is also advancing. Fannie Mae's Title Acceptance Program, which uses algorithmic risk assessment in place of traditional manual title searches on eligible refinance loans, has been applied to a significant share of qualifying transactions. Opendoor's acquisition of Doma is one example of how companies are positioning to build on that infrastructure, pairing closing and escrow operations with automated title risk evaluation to reduce timelines and costs.
For buyers and sellers, these gains are largely invisible but directly felt. Faster document processing means faster closings. More accurate underwriting means fewer last-minute surprises. These are not future promises; they are happening in transactions today.
What AI Does Not Yet Have
Real estate transactions regularly surface novel situations that no dataset fully anticipates, and navigating them requires creative problem-solving, local relationships, and judgment that is genuinely hard to replicate today.
Sometimes it is about helping a buyer see the potential in a property that does not immediately fit their needs. We regularly bring contractors into the process early, walking properties with buyers so they can understand what a renovation would realistically cost and what the home could become. That vision, and the trusted relationships that make it possible to assemble quickly, is often what moves a decision forward and leads to a better outcome than buying a fully updated property at a premium with multiple competing offers.
Sometimes the situation is more specific. A few years ago, we worked with buyers on a San Mateo property that included a small bridge on the land. The question was whether it could support the weight of a fire truck in an emergency, a real concern for insurability and safety. Answering it required finding a structural engineer who specializes in residential bridges, which required knowing that person existed and having the relationship to get a timely response. An unstable bridge could mean expensive repairs and an uninsurable property. That is not a question you want to get wrong.
In both cases, the limiting factor was not information. It was knowing who to call, how to read the moment, and what to prioritize, along with the kind of ongoing relationship where a client trusts you enough to act on advice that does not obviously serve your immediate interests. Whether AI will eventually close this gap is an open question and the technology is moving fast. What we can say with confidence is that, as of today, this dimension of the work remains firmly human.
What This Means for You
We work in Silicon Valley, and we take seriously that many of our clients are building the technology reshaping this industry. We have embraced AI in our own work and believe the agents who do not will be outpaced by those who do. We are closely monitoring what is evolving, what is creating real efficiency gains, and where it makes sense to integrate new tools into our process.
What has not changed is the nature of the decision itself. At this price point, the professionals who will serve you best are the ones using AI to remove friction behind the scenes while still bringing judgment, relationships, and local knowledge to the decisions that matter most.
Sources
HousingWire: Realtor.com launches ChatGPT app, Opendoor moves into closing and escrow with Doma deal, Fannie Mae partnership, Rocket Close, AWS collaborate on AI system to automate mortgage document workflows, Better bets on ChatGPT as the new front door for origination.
Inman: The real lesson from the Florida ChatGPT home sale story, The truth about AI in real estate according to agents, ChatGPT isn't sabotaging real estate deals. It's exposing how they're made.
