- Data source
- Answers and cited sources from ChatGPT, Gemini, Claude and Perplexity, with web search switched on, for common home-buying questions
- Collected
- November 11, 2026
- Sample
- 150 questions (15 common buyer questions, each asked plainly and in 9 situations), asked once of 4 AI assistants for 600 answers and 5,058 citations from 883 different websites
Key findings
- 1Agent, broker and other local real estate business sites receive 11.9% of the 5,058 citations AI assistants gave for home-buying questions.
- 2Lenders and credit bureaus are the largest named group in AI search for home buying, with 23.7% of citations, followed by government sites at 16.6%.
- 3ChatGPT sends 83.9% of its citations to government sites, while Gemini, Claude and Perplexity cite lenders and a long tail of other sites far more often.
- 4Gemini cites 14.6 sources per answer, almost five times as many as ChatGPT, which cites 3.
- 524.9% of citations go to sites outside every defined group, such as independent blogs and local or niche sites, and the 10 most-cited sites take only 19.9% of all citations.
- 6Zillow is the most-cited listing site at 2.6% of citations, ahead of Redfin at 2.0% and Realtor.com at 0.9%.
Summary
A first-time buyer asks an AI assistant how much she needs for a down payment. The answer comes back with a short explanation and a row of numbered sources under it: a lender's guide to loan types, a government page on first-time buyer programs, a personal finance article, perhaps a local agent's blog post about closing costs in her state. She clicks one or two of them. Those links are the new front page for her question.
Agents and brokers spend years building websites full of buyer guides, and AI search for home buying now decides whether those guides get read. Assistants that search the web choose a handful of pages to cite, and the choice is not the same as a page of Google results. Knowing which kinds of sites the assistants trust for buyer questions tells an agent where their own content stands and which sources it competes with.
Agent, broker and local real estate business sites earn 11.9% of citations across ChatGPT, Gemini, Claude and Perplexity. Lenders and credit bureaus earn twice that, 23.7%, and government sites 16.6%. The assistants differ more than they agree: ChatGPT cites government pages 83.9% of the time, while Claude gives 16.9% of its citations to agent and broker sites, the highest share of the four.
What we measured
On November 11, 2026 we asked four AI assistants 150 home-buying questions with web search switched on: ChatGPT running GPT 5.4 mini, Gemini 3.5 Flash, Claude Haiku 5.5 and Perplexity Sonar. The questions started from 15 that buyers commonly ask, such as "How much do I need for a down payment?", "Do I need a buyer's agent to buy a house?" and "What are first-time home buyer programs?". Each was asked plainly and in 9 situations, such as "I live in Texas.", "I'm self-employed." or "I want to buy a condo.". Each question went once to each assistant with no earlier conversation, giving 600 answers, 150 per assistant.
We recorded every source an answer cited and reduced it to its website. 70.5% of answers cited at least one source, and the median answer cited 8. That produced 5,058 citations from 883 different websites. Each website was placed in one group by fixed rules based on its name: government sites, which include .gov and state .us domains and housing finance agencies; lenders and credit bureaus; news, finance and home media; listing and neighborhood sites; agent directories and review sites; forums, video and social sites; legal, tenant and university resources, which include .edu pages; national brokerage brands; and agent, broker and local business sites, meaning sites whose names point to a realtor, broker, property or other local real estate business. Every other site falls into Other.
Shares are the percentage of all citations, or of one assistant's citations, that point to each group. Because assistants cite different numbers of sources, Gemini's 2,196 citations weigh more in the overall figures than ChatGPT's 447. Individual sites are named only when they are widely known platforms; agent and business sites are never named.
Limits: each question was asked once, and AI answers vary from run to run and change when the models are updated. The group rules read a site's name, not its pages, so a lender's blog post counts as a lender and a local news site with an unusual name can land in Other. Perplexity returned 10 citations for every answer, which may reflect a cap on how many sources it lists. This edition does not measure the age of the cited pages or which page on a site was cited. The study describes what the answers cite, not why the assistants chose those sources.
Lenders and government sites supply two in five citations
of AI citations for home-buying questions point to lenders, credit bureaus or government sites
Buying a home is a money question first, and the sources show it. Lenders and credit bureaus receive 23.7% of all citations, the largest of the named groups, and government sites 16.6%. Together they supply 40.3% of the sources behind AI answers to buyers. News, finance and home media add 9.2%, so close to half of all citations point to a finance, official or media source.
Finance and official sources lead; agent and broker sites hold about one citation in eight.
The single most-cited website is a federal consumer finance agency's site, with 3.8% of all citations, and the federal housing department's site has 1.5%. Many of the 15 questions turn on rules and numbers, such as credit scores, down payments, mortgage rates and first-time buyer programs, where an official page or a lender's rate table is an obvious source to cite.
Agent and broker sites earn 11.9% of citations
of citations go to agent, broker and local real estate business sites
Sites run by agents, brokers and other local real estate businesses receive 11.9% of citations, about one in eight. In AI search for home buying they are the fourth-largest group, ahead of the big listing sites at 8.6%. National brokerage brands, the franchise sites of the largest companies, receive just 0.1%, and agent directories and review sites 1.9%. When an assistant cites a real estate business, it is almost always a local one.
Claude and Gemini cite local real estate businesses most; for Perplexity the group is outside its five largest, below 13.2%.
Claude gives 16.9% of its citations to agent and broker sites and Gemini 15.7%. ChatGPT gives 9.6%, but of only 447 citations, most of them government pages. For Perplexity, agent and broker sites are not among its five largest groups, which means they make up less than 13.2% of its citations.
ChatGPT cites government sites 83.9% of the time
of ChatGPT's citations for home-buying questions point to government sites
The four assistants build their source lists in very different ways. ChatGPT cites 3 sources per answer, and 83.9% of them are government sites. Gemini cites 14.6 per answer and spreads them widely, with 28.6% in Other and 22.7% at lenders. Claude cites 6.1 per answer, with the highest lender share, 31%. Perplexity cites 10 per answer and has the most even mix, led by lenders at 27.7% with news and media at 14.5% and listing sites at 14.1%.
| Assistant | Citations | Per answer | Largest group | Second group |
|---|---|---|---|---|
| ChatGPT | 447 | 3.0 | Government, 83.9% | Agent and broker sites, 9.6% |
| Gemini | 2,196 | 14.6 | Other sites, 28.6% | Lenders and credit bureaus, 22.7% |
| Claude | 915 | 6.1 | Other sites, 34.8% | Lenders and credit bureaus, 31% |
| Perplexity | 1,500 | 10.0 | Lenders and credit bureaus, 27.7% | Other sites, 20.3% |
Source: Stagingmate analysis of 5,058 citations in 600 answers from ChatGPT, Gemini, Claude and Perplexity · November 2026
The difference matters for anyone hoping to be cited. A page that ChatGPT would pass over in favor of an official source can still be one of Gemini's 14 or Perplexity's 10. Lenders appear in the top two groups for three of the four assistants. Listing and neighborhood sites rank fifth for Gemini at 8.5% and Claude at 3.9%, and only 0.7% of ChatGPT's citations go to them.
One citation in four goes to sites outside every group
of citations go to sites that fit none of the defined groups
The largest single share, 24.9%, is Other: sites that fit none of the defined groups. They are mostly independent blogs, local and niche sites, small publishers and single-topic guides whose names do not mark them as a lender, a government office, a media outlet or a real estate business. Claude sends 34.8% of its citations there and Gemini 28.6%, while ChatGPT sends just 2.2%.
The spread is wide. The 5,058 citations come from 883 different websites, and the 10 most-cited sites together account for only 19.9% of them. No single source dominates buyer answers. For a smaller site that is good news: a clear, specific page has room to be cited, because the assistants are drawing on hundreds of sites rather than a fixed short list.
Zillow is the most-cited listing site, at 2.6%
of all citations point to Zillow, the most-cited listing site
Listing and neighborhood sites receive 8.6% of citations in all. Zillow is the most-cited of them, with 2.6% of all citations, followed by Redfin at 2.0% and Realtor.com at 0.9%. Perplexity cites listing sites the most, 14.1% of its citations. Forums, video and social sites are a small group for buyer questions, at 1.9%, with YouTube and Reddit at 0.9% each.
Even the biggest listing sites hold a few percent of citations each.
What this means for agents and real estate marketers
Write the page an assistant would want to cite. In AI search for home buying, local real estate business sites already earn 11.9% of citations, more than the big listing sites, while national brokerage brands earn 0.1%. Aim for pages that answer one buyer question clearly, for a named place: closing costs in your state, how long a purchase takes in your market, what an inspection covers. Generic pages compete with lenders and government sites and usually lose.
Match the sources the assistants already trust. Where a question turns on a rule or a number, link to the official source and explain what it means locally, because 40.3% of citations go to lenders, credit bureaus and government pages. Test your own questions in more than one assistant: ChatGPT cites 3 sources per answer, mostly government, while Gemini cites 14.6 and Claude gives agent and broker sites 16.9% of its citations.
Buyers who click through from an answer land on your listings as well as your guides, and empty rooms are what they see first. For listings that need it done for you, compare what virtual staging services include and how fast they turn photos around.
If you would rather hand the work off entirely, our guide to the best virtual staging companies compares the main services on price and turnaround.
Cite and download
Stagingmate Research. (2026). AI Search for Home Buying: Agent Sites Get 12% of Citations. Stagingmate. https://stagingmate.com/research/which-websites-ai-cites-for-home-buying-questions
Free to use and republish, as long as you credit Stagingmate Research with a link to this page. Need a local cut or a quote? Email support@stagingmate.com.
Frequently asked questions
Lenders and credit bureaus are the largest named group, with 23.7% of 5,058 citations, followed by government sites at 16.6% and agent, broker and local business sites at 11.9%. Another 24.9% go to sites outside every defined group, such as independent blogs and niche sites.
Yes. Agent, broker and local real estate business sites receive 11.9% of citations for buyer questions. Claude gives them 16.9% of its citations, Gemini 15.7% and ChatGPT 9.6%. National brokerage brand sites receive 0.1%.
Gemini, with 14.6 citations per answer, followed by Perplexity at 10, Claude at 6.1 and ChatGPT at 3. 70.5% of all answers cited at least one source.
Very different ones. 83.9% of ChatGPT's citations for home-buying questions point to government sites. The other three assistants cite lenders, media and a long tail of independent sites far more often.
Zillow receives 2.6% of all citations, the most of any listing site, ahead of Redfin at 2.0% and Realtor.com at 0.9%. Listing and neighborhood sites as a group receive 8.6%.



