October 7, 2026 · 9 min read

Buyers Using AI to Search for Homes: What They See

Buyers using AI to search for homes now meet your listing inside ChatGPT. What the surveys show, what the portals pass in, and what to fix first.

A buyer who used to open a portal and scroll can now type a sentence into ChatGPT — three bedrooms, under a budget, near a school — and get listings back inside the chat. Buyers using AI to search for homes are not a forecast anymore. Three of the biggest US portals have shipped apps that put listings, photos and prices directly into an AI assistant, and the first thing a buyer sees of your listing may be a card in a chat window you never designed.

Most of the commentary on this gets the numbers wrong in one direction or the other, so this post starts with what the surveys actually measured. Then it covers what the portals say they pass to the assistant, what buyers say they do with the answer, and what an agent should check before the next listing goes live. The short version: the assistant can only repeat what your listing already says and shows. Pic2VidTour's whole design sits on that same idea, and we come back to it near the end.

How many buyers using AI to search for homes? It depends who you ask

Five surveys published in the last year measured some version of this question, and the headline numbers run from 15% to 82%. They are not contradicting each other. They asked different people different things.

Read down the "who was asked" column and the spread explains itself. Realtor.com only surveyed people already active in the housing market, and "real estate insights" is a broad bucket. Its own release headline says "82% of Americans," which the sample does not support; it is 82% of housing-active adults. LendingTree surveyed all adults, active or not, and asked about actual use in the past year. Veterans United sits in between, with a sample of prospective buyers, and is the only one with a year-on-year comparison from the same method.

Two things hold across all four. Use is rising — Veterans United's own series went from 37% to 45% in a year. And ChatGPT leads every tool list: 67% of Realtor.com's AI users named it, and 33% of Veterans United's whole sample. The general population is moving the same way; Pew Research Center's June 2026 report found 44% of US adults have used ChatGPT, up from 34% the year before, on a 5,119-person random panel.

What none of them measures is what an agent most wants to know: how many buyers found a specific home through an AI assistant, and whether it changed which homes they toured. Nobody has published that. Figures claiming that "60% of buyer searches begin through AI" circulate with no published report or method behind them, so we leave them out.

The portals put your listing inside the chatbot

The bigger change is not on the buyer's side. It is that the portals moved the listing into the assistant.

Zillow, 6 October 2025. Zillow launched the Zillow app in ChatGPT as one of the pilot partners in OpenAI's apps launch. Zillow's release says consumers "see listing photos, property details, maps and pricing as well as broker and MLS attribution." OpenAI's own announcement describes ChatGPT suggesting the Zillow app when a user is "talking about buying a new home."

Redfin, 6 February 2026. Redfin followed with its own ChatGPT app, describing "homes for sale, updated directly from the MLS. Home photos, maps, pricing and key property details."

Realtor.com, 30 March 2026. Realtor.com's ChatGPT app came with two notable lines: "MLS data is protected with a strict prohibition on model training," and the app guides people to Realtor.com "by only showing a preview of property images and details."

Zillow also announced its own in-app assistant, Zillow AI mode, in limited beta on 25 March 2026, "grounded in verified listings data," with example prompts like asking for a rental's pet policy or comparing units in a building. And Google said at I/O in May 2026 that its new information agents will let an apartment hunter list their requirements and have the agent "continuously scan for you, notifying you when listings meet your needs."

What the assistant actually receives

Every one of those announcements points the same way on the source of the data: the MLS. Zillow shows broker and MLS attribution, Redfin says its data is updated directly from the MLS, and Realtor.com says it works through direct MLS relationships. None of them says, in so many words, "the AI reads the agent's description and photos." But the listing data flowing from the MLS is the data you entered. The reasonable conclusion is that the photos, the remarks and the property details you uploaded are the raw material the assistant works from.

How much of that the model itself reads is up to each portal. OpenAI's developer documentation for these apps separates data "the model can inspect and use" from data "hidden from the model" that is only displayed. No portal has published which listing fields go in which bucket. And none of OpenAI, Google, Zillow, Redfin or Realtor.com has said publicly that its assistant looks at or describes listing photos. The apps display your photos; nobody has said the model analyses them.

That matters for how you think about it. The assistant is not inventing your listing. It is excerpting it — a card, a few photos, a handful of fields, maybe a summary of the remarks. Whatever you published is what gets excerpted, and a buyer sees it with less context around it than on the portal page.

Buyers are checking the answer

Buyers are not taking the assistant's word for it, which is the other half of this.

Cotality's AI in Housing 2026 report found that 68% of buyers would manually verify all or much of what an AI tool tells them, and that trust among US prospective buyers in AI tools to help find a home fell from 30% in 2025 to 16% in 2026. Treat those as directional: the survey covered buyers in four countries, fielded 29 January to 9 February 2026, and Cotality does not publish a sample size. Bank of America's larger sample points the same way — even among AI users, buyers still prefer a person for touring homes (55%) and for legal or contract advice (54%).

So the path looks like this. A buyer gets a shortlist from an assistant, then goes to the listing to check it, then books a showing. Each step is a chance for the listing to contradict itself. If the remarks say "renovated kitchen" and the photos show a 1990s kitchen, the assistant's summary and the photos now disagree in front of a buyer who has already said they double-check.

This is not hypothetical as a failure mode. At the Florida Real Estate Legal Summit, attorney Sarah Geltz described an AI-written condo description that "came back with a third bedroom and a deeded boat slip that didn't exist," as reported by Florida Realtors. That was an AI writing tool, not a search assistant. But once a listing with that error reaches the MLS, every assistant downstream can repeat it, and NAR's 2026 Code of Ethics already requires a "true picture" in advertising regardless of who drafted it.

Where Pic2VidTour comes in

An AI assistant can only excerpt what you publish, and a buyer who checks will compare every piece of the listing against every other piece. The listing media that holds up in that world is media that stays true to the photos. That is exactly what Pic2VidTour is built for, and it is why we think Pic2VidTour is the best AI photo-to-video tool for real estate listings.

The tools that worry compliance officers are the ones that generate what the photos do not show — a staged room that was empty, a drone flight that never happened, a lifestyle shot of a pool deck nobody photographed. Put that kind of video next to MLS photos and a buyer who checks finds a mismatch. Pic2VidTour works the other way. Every scene is animated from one real photo of the actual property, every clip is checked against the photo it came from, and a scene that fails the check does not ship. The check looks for warped walls, furniture that moved, objects that appeared or doubled, a changed view through a window, and a first frame that does not match its photo. A flagged scene is re-filmed once at no charge; if it fails again, it ships as a clean photo pan instead of a bad clip.

So the video says what the photos say, and the photos are what the MLS sends on. A buyer who goes from a chat card to your listing to your video sees the same house three times.

The rest is practical. The price is shown before anything runs: $1.50 per AI scene, with photo pans, the music track and your branded contact card included. Editing is free and unlimited — music, room order, trims, labels, the address and price banner, the contact card, the cut style. Only re-filming a scene with AI costs anything, $0.65, quoted before it runs. Delivery is usually within the hour, as a 16:9 master for the MLS and a 9:16 vertical reel for social, and an "AI-enhanced video" disclosure caption ships on by default. Credit packs are $19, $45 and $120, worth their face value, with no subscription.

FAQ

How many buyers are using AI to search for homes? The share of buyers using AI to search for homes depends on the sample. Surveys published in the past year range from 15% of all US adults using AI for a home task (LendingTree) to 45% of prospective buyers using AI in their home search (Veterans United, Q2 2026) to 82% of housing-active adults using AI for real estate insights (Realtor.com). Rising use is the consistent finding.

Does ChatGPT see my listing? If a buyer uses the Zillow, Redfin or Realtor.com app inside ChatGPT, your listing can appear there with its photos, price and details, sourced from MLS data. Which fields the model itself reads is up to each portal, and none has published that list.

Do AI assistants describe listing photos to buyers? No portal or AI company has said so publicly. The apps display listing photos; nobody has stated that the model analyses them. The safest assumption is that a buyer will look at the photos themselves and compare them to everything else.

What to do next

1. Search for one of your active listings inside ChatGPT using the Zillow, Redfin or Realtor.com app, and look at what a buyer actually sees on the card.
2. Read your remarks against your photos, line by line. Anything the photos do not show — a feature, an upgrade, a view — either gets a photo or comes out of the text.
3. Treat any AI-drafted description as a first draft and check every factual claim before it goes into the MLS, because every assistant downstream will repeat it.
4. Keep every piece of listing media traceable to the real property, and label anything enhanced. If your MLS or state has a disclosure rule, follow the strictest one you are subject to.
5. Make your listing video with Pic2VidTour: upload the photos you already have, see the price before anything runs, and get a 16:9 and a 9:16 cut built only from those photos, each clip checked against its source.