How is AI being used in real estate? Real examples, benefits and risks in 2026
AI in real estate mostly writes listings, powers chat-style home search and sets prices. Here's where it helps, where it fails and what to watch.

In March, Zillow started letting a small group of users ask its site questions like "Can I afford this apartment if I move in June?" and get an answer built from live listings. That's a fair snapshot of how AI is being used in real estate right now: mostly as a writing assistant for agents, a conversational layer on top of property search, and a pricing and screening engine running in the back office. It saves people time. It also brings real risks around accuracy, doctored photos and fair housing, and I think those risks get less attention than they deserve.
The agent side is the most measurable. In NAR's 2026 technology survey of 1,200 agents, nearly half said they use AI daily or weekly, and only 12% said they weren't using it and had no plans to.
The background: from price estimates to chat search
Property has been a data business for a long time. Every sale leaves a record (price, date, square footage, bedrooms), and that's exactly the kind of structured history that statistical models like.
Automated valuation models, the software behind Zillow's Zestimate and similar home-value estimates, have been around since the 2000s. They look at comparable sales and property features and spit out a number.
What's changed in the last three years is the other half of the job: words. Listings, emails, disclosures, leases and neighbourhood guides are all text, and large language models are good at producing and summarising text. So the industry now runs on two kinds of AI at once. There's the older predictive kind that estimates prices, rents and risk, and the newer generative kind that writes, chats and edits images.
The predictive side can go badly wrong, though. Zillow shut down its home-flipping business, Zillow Offers, in November 2021 after its buying algorithm overpaid for homes in a shifting market. I'd treat that as the industry's most expensive lesson in trusting a model's price over a local expert's judgement, and it still shapes how cautious some brokers are.
Real examples of AI being used in real estate
Most of the day-to-day use is unglamorous. Among agents who use AI, the same NAR report found writing listing descriptions was the top job at 75%, followed by social media posts (56%) and emails and follow-ups (52%). ChatGPT dominates: 91% of AI-using agents use it, with Gemini at 42%, Copilot at 27% and Claude at 25%. Pricing help and comps support came in at just 23%.
Here's how the main uses break down by who's doing the work and where a human still has to check it:
| Job | What AI does | Who uses it | The human check |
|---|---|---|---|
| Listing copy and emails | Drafts descriptions, posts and follow-ups | Agents | Facts, tone and fair housing wording |
| Home search | Answers natural-language questions about listings and affordability | Buyers and renters | Tours, inspections, loan terms |
| Valuation | Estimates price or rent from comparable sales | Portals, lenders, investors | Appraisals and local knowledge |
| Photos and staging | Adds furniture, removes clutter, changes finishes | Listing agents, photographers | Disclosure of what's been altered |
| Leasing and management | Answers renter enquiries, books viewings, sets rents | Landlords and property managers | Screening fairness, pricing rules |
Search you can talk to
Zillow is the biggest consumer example. Its app inside ChatGPT launched in October 2025 for logged-in US users, so you can type "Zillow, show me homes for sale in Kansas City" and get real listings with photos and maps. Then in March it began testing Zillow AI mode on its own site, a beta that can compare homes, estimate renovation costs from local data, explain price cuts and book a tour. Zillow says it plans to expand that experience over the year. I'm curious whether people trust it with offer strategy; asking a portal that also sells agent leads how hard to negotiate feels a little odd to me.
Leasing assistants
Rentals have moved faster than sales, probably because renters send so many repetitive questions. EliseAI sells an assistant that answers enquiries by text, email, website chat and phone, with voice support in seven languages and written replies in 51. For a big apartment operator, that means a pet-policy question at 11pm gets an answer instead of a voicemail.
Virtual staging and photo edits
AI staging tools can drop a sofa into an empty living room or repaint a kitchen in seconds. It's cheap and, for a vacant flat, often helpful. It's also easy to abuse, which is why California's AB 723 took effect on January 1. California's largest MLS, CRMLS, explains in its guide to the law that a listing with an altered image now has to show the original photo right before or after it, labelled "virtually staged" or "digitally altered". Basic edits like lighting, white balance and cropping are still fine. (If you're curious about the design side, our roundup of AI apps for interior design covers tools that do similar room makeovers.)
Pricing algorithms in rentals
The most contested example is rent-setting software. The US Department of Justice sued RealPage, and in November 2025 it filed a proposed settlement that would stop its software using competitors' nonpublic data to set rents in real time. Model training would be limited to nonpublic data aged at least 12 months, and a court-appointed monitor would check compliance. That's a pretty specific picture of where regulators think a pricing model crosses the line.
Benefits and limitations
Time is the clear win. A separate survey of 225 agents by Realtors Property Resource, an NAR subsidiary, found 68% save at least an hour a week with AI and 71% named saving time as its top value. An hour a week isn't dramatic. Still, for a solo agent juggling a dozen clients, it adds up to a couple of working weeks a year.
The limits show up in the same survey. Accuracy was the top concern for 63% of those agents, followed by compliance or legal issues (49%) and misreading market data (47%); only 28% flagged fair housing, which seems low to me given how much of real estate law is about who gets offered what. A chatbot that invents a square footage, or a listing description that hints at the "ideal family" for a home, can land an agent in real trouble.
A rough analogy helps here. An AI valuation is like a weather forecast for a single street: useful for planning, often close, and occasionally wrong in ways that matter a lot to the person standing in the rain. You wouldn't skip the umbrella because the app said 10%, and you shouldn't skip the appraisal because the estimate looked right.
Fair housing is where we know the least about how things will play out. In May 2024, HUD issued guidance on AI in tenant screening and housing ads, making the point that the Fair Housing Act applies even when an algorithm makes the call. HUD has since withdrawn the advertising guidance as part of a deregulatory review. The law itself hasn't changed, though, so I'd guess the legal risk has moved from regulators to private lawsuits rather than disappeared.
Practical takeaways for buyers, sellers and agents
If you're buying or renting, use AI search to narrow things down, not to decide. Ask it the dull questions (pet rules, commute times, what similar homes sold for) and then check anything that affects money with a person: the lender, the inspector, the agent. If a listing photo looks too perfect, look for a "virtually staged" label and the original next to it.
Sellers should ask their agent how they'll use AI on the listing. Staged photos are fine with clear labels. Edits that hide a crack or a power line aren't, and in California they're now a legal problem as well as an ethical one.
For agents, the safest pattern I can see is to let AI write first drafts and never let it publish. Read every listing description for wrong facts and wording that steers buyers by family status, religion or race. Our guide to writing better AI prompts will get you better drafts, and if you're weighing whether any of this threatens the job itself, we've looked at whether AI will replace real estate agents.
The next few years
What we know: adoption is rising fast, money keeps flowing in (NAR counted 231 proptech funding rounds worth $4.53 billion in the first half of this year alone), and the big portals are betting on conversational search. What we're guessing at is how much of the transaction AI will actually run. Right now agents barely use it for contracts and disclosures; the NAR survey puts summarising those, auto-filling forms and recommending next steps at 10% or less.
My guess is that changes over the next couple of years as AI agents built for real estate start handling scheduling, paperwork and follow-ups end to end. I'm less sure how regulators will respond. California's photo law and the RealPage settlement suggest the rules will arrive piece by piece, one bad practice at a time, rather than as a single AI law for housing. Honestly, I think that's probably the right way to do it, even if it leaves buyers with a patchwork for a while.
FAQ
How is AI being used in real estate?
Mostly for writing and search. Agents use tools like ChatGPT to draft listing descriptions, social posts and emails. Portals such as Zillow let buyers search by asking questions in everyday language. Behind the scenes, AI also estimates home values, sets rents, screens tenants and creates virtually staged photos.
How many real estate agents use AI?
Nearly half of agents use AI daily or weekly, based on NAR's survey of 1,200 agents. Only 12% said they don't use it and don't plan to, down from 32% who hadn't tried it the year before.
Is virtual staging with AI legal?
Generally yes, as long as it isn't misleading. California's AB 723, in force since January 1, requires listings with digitally altered images to show the unaltered original alongside them with a clear label. Many MLSs also ban edits that hide defects or add features that aren't part of the sale.
Can AI tell me what a home is worth?
It can give you a reasonable starting point. Automated estimates like the Zestimate use comparable sales and property data, but they can't see a new roof, a damp basement or a noisy neighbour. Zillow's AI mode will even let you question the Zestimate and ask how it's changed over time, which is handy for context. For a purchase or a loan, a licensed appraisal and a local agent's view still matter more.
Does AI in real estate raise fair housing concerns?
Yes. The Fair Housing Act applies whether a person or an algorithm makes a housing decision. Tenant screening models, targeted ads and AI-written listings can all discriminate if they're built or used carelessly, so agents and landlords should review outputs for wording or patterns that exclude protected groups.



