AI Property Search: The Future of Finding Real Estate

AI property search lets buyers and renters describe the home they want in plain language and turns that request into a search of live listings. Most big portals now offer it, and several work inside ChatGPT. It copes with wishes that filters can't, but it can also miss better options, so treat it as a first pass.

This guide explains how AI property search works, which portals offer it in the US, UK and UAE, what the evidence says about results and adoption, where it goes wrong, how Dubai's advertising rules apply, and what buyers, brokers and developers should do about it.

Key takeaways

• It's now standard on the big portals. Zillow, Redfin, Realtor.com, Homes.com, Rightmove, Zoopla and Bayut all offer natural-language or conversational search, and five portals run apps inside ChatGPT.
• It translates; it doesn't know. The AI turns your words into filters and tags drawn from listing text and photos, then searches live listings. It doesn't hold the listings in its memory.
• Early results are promising but self-reported. Redfin says users of its conversational search are 47% more likely to request tours; Rightmove reports close to double the propensity to send leads. None of these figures comes from a controlled test.
• Adoption is real but modest. US surveys put buyers' AI use anywhere from 20% to 82% depending on the question, and Rightmove measured under 0.5% of its traffic arriving from AI chatbots.
• The risk is omission, not invention. A September 2026 study found models invented no listings when given real ones, but 39% of their picks were beaten on price, commute and size by another listing in the same set.
• In Dubai, the permit is the check. Every property ad needs a Dubai Land Department permit and a Madmoun QR code, whatever tool surfaced it.

What is AI property search?

AI property search is any home search in which a language model interprets what you type or say instead of you setting filters one by one. It has developed in four steps:

Stage	What it does	Example
Natural-language search	Turns one typed sentence into filters	Zillow's AI search (2023), which later added commute, school and affordability searches (2024)
Conversational search	Refines results over several turns	Redfin (web November 2025, iPhone January 2026); Bayut's AI Search (announced August 2026)
AI mode	Conversation plus actions such as booking tours, with preferences remembered across sessions	Zillow AI mode (beta from March 2026); Realtor.com RealAssist AI (beta from June 2026)
Portal apps in assistants	A portal's live listings shown inside ChatGPT, with next steps handed back to the portal	Zillow, Redfin, Realtor.com, Zoopla and Bayut apps in ChatGPT

Definition

AI property search — a home search in which a language model reads a plain-language request, such as a three-bedroom villa near a good school with a garden under AED 4 million, converts it into structured filters and descriptive tags, and retrieves matching live listings, often refining them through follow-up questions.

Zillow says it was the first major residential marketplace to launch natural-language search, in January 2023. Since then the leading US portals have moved from single queries to conversations, and in 2026 to AI "modes" that remember what you like and can book a viewing.

How does AI property search work?

The model sits between the buyer and the listings database and translates one into the other. CoStar, which runs Homes.com, says its model acts "as a translator, not a student": it sees a short snippet of listing text to answer a question, doesn't learn the listing data and forgets it afterwards.

From a sentence to a shortlist

1. 01Read the request
   • Budget and location
   • Must-haves
   • Soft wishes

   Plain words, any order

2. 02Translate
   • Standard filters
   • Descriptive tags
   • Commute or school limits

   What the portal can search

3. 03Search live listings
   • That portal's inventory
   • Current prices
   • Listing text and photos

   Not the model's memory

4. 04Rank and show
   • Matching homes
   • Filters applied
   • Reasons to compare

   Check what was applied

5. 05Refine or hand off
   • Follow-up questions
   • Tour or agent request
   • Saved preferences

   Leads go to the portal

The descriptive tags come from the listings themselves. Realtor.com says its AI search reads listing photos and descriptions, can filter on more than 300 terms and tags, and shows users in real time what the site is actually searching for. Redfin says half of its conversational searches include some request that a standard filter can't capture.

That design has two consequences. The AI can only return homes the portal holds, so each tool sees one portal's inventory. And the quality of the match depends on the listing: a sea view, a maid's room or a service charge that isn't in the text or visible in the photos is hard for any AI to find.

Which portals offer AI property search?

Most of the large portals in the US and UK now have it, and in the UAE Bayut has moved first.

Portal	Market	AI search on the portal
Zillow	US	Natural-language search since January 2023; AI mode in beta since March 2026, available to a growing number of users by June
Redfin	US	Conversational search since November 2025 (web) and January 2026 (iPhone), built with Sierra
Realtor.com	US	AI search since October 2025; RealAssist AI, built with Google's Gemini, in beta since June 2026
Homes.com	US	Smart Search (October 2025) and Homes AI (February 2026), by voice or text
Rightmove	UK	"Ask Rightmove" conversational search, rolling out in 2026
Zoopla	UK	AI-powered search tested in its app, backed by an enterprise agreement with OpenAI (April 2026)
Bayut	UAE	BayutGPT assistant (2023); AI Search, a conversational search alongside its filters (announced August 2026)
Property Finder	UAE	No consumer conversational search announced as of September 2026; its AI announcements focus on agents

Five portals have also launched apps inside ChatGPT that show their own live listings in the chat: Zillow (October 2025), Redfin (February 2026), Realtor.com (March 2026), Zoopla (April 2026) and Bayut (June 2026), which calls itself the first UAE-based property platform to do so. In each case, viewings, agent contact and financing are handed back to the portal. What this means for property marketing is covered in AI search for real estate companies.

Does AI property search work better than filters?

The portals say it does. Their data can't prove it yet.

What the data shows

What the portals report

• Redfin: in early testing, conversational search users viewed nearly twice as many listings and were 47% more likely to request tours or other services; on iPhone, users favourite, tour and share homes about 60% more often than non-users.
• Rightmove: early indications from Ask Rightmove suggest about a 40% uplift in average time on site and close to double the propensity to send leads (first half of 2026).
• Zoopla: app users who tried AI-powered search showed 80% more listing views and 150% more leads in early tests.
• Property Finder (UAE): time spent on a listing has fallen from about 60 seconds to around 40, while searches by specific attributes have risen sharply.

Each comparison sets people who chose to use the new feature against people who didn't. Keen, active searchers are more likely to try it, so the gap may say as much about them as about the tool. None of the portals has published a controlled test.

General chatbots are not yet a big source of visits either. Rightmove measured less than 0.5% of its traffic in the first half of 2026 arriving from large language models, while more than 85% was direct or organic. For now, the conversational layer that matters most sits inside the portals.

How many buyers use AI to search for homes?

It depends entirely on how the question is asked.

Survey	Who was asked	Finding
Bank of America (April–May 2026)	2,000 US adults (1,000 homeowners, 1,000 renters) who own, have owned or plan to own a home	20% used AI tools or chatbots for homebuying research in the past year (Gen Z 32%, Millennials 28%)
Veterans United (June 2026)	859 veterans, service members and civilians	45% of prospective buyers use AI tools in the homebuying process, up 8 points in a year
NerdWallet and The Harris Poll (November 2025)	296 adults planning to buy within a year	48% have used or will use AI
Realtor.com (August 2025)	1,000 US adults buying, selling or recently done so	82% use AI for housing-market information

Agents remain central. In NAR's 2025 Profile of Home Buyers and Sellers, 88% of US buyers used an agent or broker, and in Zillow's 2025 survey, half of buyers named their agent as their most helpful resource, against 36% for a real estate app or website. The first step is still usually online: among buyers who purchased between mid-2023 and mid-2024, 43% started by looking at properties online and 21% by contacting an agent.

Trust is moving the other way. Cotality found US prospective buyers' trust in AI tools to help find a home fell from 30% in 2025 to 16% in early 2026, and 68% of buyers said they would manually verify every detail, or a significant amount, of what an AI tool tells them. Buyers also punish AI mistakes harder: 70% would lose trust after a significant AI error, against 60% after a comparable human one. No UAE survey of AI use in property search has been published.

What can go wrong with AI property search?

The main risks are the results it doesn't show, bias in what it recommends, and data that is wrong before the AI ever reads it.

• It misses better options. A September 2026 preprint tested AI models from OpenAI and Anthropic on 150 New York rental scenarios, each with a pool of 120 real listings — 9,945 model calls in all. The models broke a stated hard requirement in only 1.8% of recommendations and invented no listings, but 39% of their recommendations were beaten by another listing in the same pool that was a median $900 a month cheaper and 3.5 minutes closer. On rent, they did worse than picking at random. It's a single-author study that hasn't been peer reviewed, but it points at the right risk.
• It can steer. A June 2026 audit of seven AI models across four US cities found racial steering in location recommendations, though it varied in direction and size by model and city. Zillow released an open-source Fair Housing Classifier in 2024, warning that many AI tools ignore fair-housing requirements, and its AI mode checks both questions and answers with it.
• US rules have loosened. HUD withdrew its 2024 guidance on advertising housing through digital platforms, effective 17 September 2025. Conduct that breaks the Fair Housing Act itself remains subject to enforcement.
• No tool sees every home. Each AI searches one portal's inventory, and Redfin itself tells buyers not to assume every site shows the same homes. Pre-market programmes such as Redfin Early Access and Zillow Preview split inventory further.
• It remembers you. Zillow AI mode and RealAssist remember preferences across sessions. The no-training promises from Realtor.com and Homes.com concern MLS listing data, so check each portal's privacy terms for what happens to your conversations.
• The listings can be wrong. An AI is only as accurate as the ads it reads. In Dubai, the Land Department's AI monitoring had automatically modified 29% of the more than 279,000 portal ads it had checked by April 2025.

Common misconception

"The danger is that AI search invents listings." When it's grounded in a portal's live data, it rarely does. The measured problems are omission and bias: good homes it doesn't show and patterns in what it recommends. Ask what it left out, and search more than one portal.

How does AI property search work in Dubai?

Dubai has one portal with conversational search, one without, and a regulator that polices the listings both of them show.

• Bayut became the first UAE-based platform with an app in ChatGPT in June 2026, letting users search by location, budget, size, bedrooms and property type. In August 2026 it announced AI Search, a conversational search that works alongside its existing filters and that Bayut says no other UAE portal offers. Its BayutGPT assistant dates back to 2023.
• Property Finder hadn't announced a consumer conversational search as of September 2026. It says it uses AI-driven nightly sweeps to remove thousands of duplicate or misleading listings a day, and its research describes buyers arriving with clearer criteria, higher expectations and less patience.
• Every ad needs a permit. Dubai Land Department requires an advertising permit, issued through Trakheesi, for every channel including electronic ads.
• Every ad needs a Madmoun QR code. Since 24 April 2023 the code has linked each ad to its permit. DLD's inspections check that the ad's details match the permit, that the permit is valid and that it was issued to the same company, with fines of AED 50,000 per violation in its 2023 statement. DLD advises buyers to deal only with ads that carry the code.
• AI already polices the portals. DLD's AI advertising platform had monitored more than 279,000 ads on Property Finder, Dubizzle and Bayut by April 2025 and automatically modified 29% of them.

No official or portal statement confirms whether AI search results or ChatGPT app cards show the permit or QR code, so open the full listing before you act on it. Valuation tools are covered in AI property valuations, and the wider local market in Dubai proptech.

What this means

For Dubai buyers, AI search changes how you find a listing, not how you verify it. Whatever surfaced a home, a genuine ad carries a Madmoun QR code that opens its Land Department permit.

How should buyers use AI property search?

Use it to cast a wider net, then check its work.

1. Describe everything at once, including soft wishes such as light, layout or a short school run, then check which filters the tool actually applied.
2. Search more than one portal. Each AI sees only its own inventory, including in ChatGPT.
3. Push back. Ask whether anything cheaper, closer or larger also fits; the evidence suggests a first answer often isn't the best available.
4. Verify every listing. Open the full ad, scan the Madmoun QR code in Dubai, and confirm price and availability with the agent.
5. Keep documents out of the chat. Don't paste passport or Emirates ID details into a search assistant, and review what it remembers about you.
6. Use an agent for what AI can't do: viewings, negotiation, and off-plan or off-market stock the portals don't show.

What does AI property search mean for brokers and developers?

It turns listing quality into search visibility, and it sends better-informed buyers to whoever answers first.

• The listing is the matching surface. Complete every structured field — view, floor, parking, maid's room, service charge, handover date, furnishing — write specific descriptions and use accurate photos, because the AI reads both.
• Keep permit data exact. A mismatch between an ad and its permit is an explicit DLD inspection point, and DLD's AI already checks the portals.
• Expect sharper leads. Buyers arrive with detailed requirements and affordability questions. Fast lead qualification and routing matter more, not less.
• Show what the AI missed. Agents who search the whole market, including stock on other portals and off-plan launches, can add exactly what a single-portal AI leaves out.
• Give developers' data structure. Payment plans, handover dates, escrow registration and floor plans give AI tools accurate facts to summarise.
• Measure it. Ask portals for lead attribution from AI search and ChatGPT apps, and test how your own listings appear in them.

For the marketing side of the same shift, see AI real estate marketing.

Final takeaway

AI property search has moved from novelty to default on the big portals in two years. It's good at turning messy human wishes into a search, and the portals report more engagement from the people who use it. But it only sees one portal at a time, it can overlook better homes, and buyers trust it less the more they use it. For Dubai, the rule is simple: let AI widen the search, then verify every listing against its permit and let a good agent close the gaps.