Salesman Hal

Getting found by ChatGPT: a guide for real estate agents and teams

Buyers and sellers now ask ChatGPT, Claude, Perplexity and Google’s AI Mode to explain a neighborhood, guess what a house is worth, or find them an agent. The big portals are already inside those assistants. Here’s where that leaves an individual agent or a small team, what to fix on your site and profiles, the compliance lines to keep in mind, and how to see what AI says about you today.

Updated

How buyers and sellers use assistants now

People bring the whole question to an assistant now, not a few keywords. Three kinds come up again and again in real estate:

  • Searching for homes. “Three-bed houses in Round Rock under $450,000 with a fenced yard and a short drive to the Domain.” The assistant narrows things down in conversation instead of through filters.
  • Learning about a place or a price. “What’s it like living in Mueller?” “What’s my condo worth?” “Should we rent or buy in Austin right now?” Zillow says early testers of its own conversational search mostly asked about affordability, neighborhoods and comparing homes.
  • Finding an agent. “Find me a listing agent in Austin who knows condos.” “Who are the best-reviewed buyer’s agents for first-time buyers in Pflugerville?” Here the assistant builds a shortlist of people, and that shortlist is where you want to be.

The first kind is mostly the portals’ game now. The second and third are where an agent’s own site and reputation matter most.

The portals are already inside the assistants

As of November 2026, this is what the big portals and Google have shipped, by their own announcements and trade reports:

  • Zillow in ChatGPT (October 2025). Signed-in U.S. users on Free, Plus and Pro can start a prompt with “Zillow” and get for-sale and rental listings with photos, maps and prices. To request a tour, contact an agent or look at financing, the user leaves ChatGPT for Zillow.com. Zillow says the listings carry broker and MLS attribution.
  • Redfin in ChatGPT (February 2026). Listings that Redfin says are updated from the MLS, refined by conversation. Tours, agent contact and financing are handled through Redfin.
  • Realtor.com in ChatGPT (March 2026). Aimed at the stage before a search: budgets, comparing neighborhoods by commute and school boundaries, renting versus buying. It shows limited listing previews and sends serious buyers back to Realtor.com to see listings and reach an agent.
  • Zillow in Claude (October 2026). A connector with the same pattern: search in Claude, then tours, agents and loan officers through Zillow. Zillow says it doesn’t send listing data to Anthropic.
  • Google has been testing for-sale listings in search results through HouseCanary’s ComeHome, and RISMedia reported in July 2026 that the pilot is expanding to all 50 states, with a direct line to the listing agent. That’s classic Search, not a feature of AI Mode; we couldn’t find a Google announcement of listings in AI Mode itself.

Two things follow for a small team. First, when a buyer browses listings inside an assistant, the “contact an agent” step usually goes to the portal’s own flow, not to the listing agent’s website. Second, NAR has left it to each MLS to decide whether these displays comply with IDX rules, and critics argue they don’t. That debate is between the portals and the MLSs, but it’s a reason not to build your own AI plans on other brokers’ listing data (more below).

Where you can still win directly is the question no portal app answers for you: which agent should I talk to?

What assistants look at when picking an agent

No assistant publishes a formula for recommending agents, so be wary of anyone selling one. Google says its local results weigh relevance, distance and prominence, including reviews and links from other sites. The other assistants search the web and read pages live. In practice, they’re trying to confirm a short list of facts:

The client wants to know Where the assistant looks
Do you work where I’m buying or selling? Neighborhood and town pages on your site, service areas on your Google Business Profile, areaServed in your markup
Do you know my kind of property or situation? Specialty pages (condos, new construction, relocation, first-time buyers), your bio, recent sales you’ve written up
Are you any good? Review text on Google, Zillow and Realtor.com profiles; press, association and community mentions
Who exactly are you? Your name, team, brokerage, license and contact details, and whether they match across your site and profiles
What’s on the market? Portal apps first; your own IDX pages if they’re readable
How do I get started? A consultation booking link or valuation request form that works, a phone number that gets answered

Every fix below fills in one of those rows.

Fix 1: profiles and reviews that agree

  • Claim and complete your Google Business Profile. Use the Real estate agent category, list the areas you serve, and keep hours and phone current. Google says detailed, accurate profiles help relevance.
  • Make your portal profiles match. Name, team name, brokerage, phone, photo, specialties and areas should read the same on Google, Zillow, Realtor.com, your brokerage’s roster and your own site. An assistant comparing sources trusts the facts that agree.
  • Ask for specific reviews. “Ada sold our Travis Heights bungalow in nine days and handled two inspections” tells a reader, human or AI, what you do and where. “Great agent!” doesn’t. Reply to reviews in the same concrete way.
  • Earn a few independent mentions: a local news quote, an association page, a community sponsorship. Assistants read the wider web, not just profiles.

Fix 2: neighborhood and specialty pages

An assistant can only say “she knows downtown condos” if a page says so, with detail that proves it.

  • One page per area you really work, with things only someone local knows: housing stock and age, typical HOA arrangements for condos, flood zones and insurance questions, commute routes, which school district boundaries apply and where to check them, recent sales you handled. A page that’s the same text with the suburb name swapped helps nobody.
  • One page per specialty: condos and high-rises, first-time buyers, relocation, new construction, investment property, downsizing. Say how you work, what it costs the client, and what you don’t do.
  • A seller’s page explaining how you price a home and what a valuation from you involves. This is the page an assistant reads when someone asks “what’s my house worth?”
  • An FAQ with the questions your inbox gets, in the client’s words: “Do you work with buyers in Cedar Park?”, “How do you charge?”, “Can you help if we’re moving from out of state?” Short, direct answers are what assistants quote.
  • Plain text, not graphics for your name, license number, brokerage, phone and areas. Many AI fetchers read only the raw HTML.

The compliance section below covers what to keep out of these pages.

Fix 3: a team page with RealEstateAgent markup

Give every agent a page, and give the team one page that lists them all. Then add structured data, a block of code that states your facts in a format machines read without guessing. schema.org has a RealEstateAgent type, a kind of LocalBusiness, and Google’s guidelines say to use the most specific type you can. Google requires name and address and recommends telephone, url and opening hours. areaServed takes place names, parentOrganization can name your brokerage, and knowsAbout, which schema.org intends mainly for people, suits each agent’s specialties. Here’s an example for a team:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "RealEstateAgent",
  "name": "The Okafor Team at Example Realty",
  "url": "https://okaforteam.example/",
  "telephone": "+1-512-555-0186",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "800 W 5th St, Suite 300",
    "addressLocality": "Austin",
    "addressRegion": "TX",
    "postalCode": "78703",
    "addressCountry": "US"
  },
  "areaServed": ["Austin", "Round Rock", "Pflugerville", "Cedar Park"],
  "parentOrganization": {
    "@type": "RealEstateAgent",
    "name": "Example Realty",
    "url": "https://examplerealty.example/"
  },
  "employee": [{
    "@type": "Person",
    "name": "Ada Okafor",
    "jobTitle": "Team lead, licensed sales agent",
    "url": "https://okaforteam.example/team/ada-okafor",
    "knowsAbout": ["Condominiums", "First-time buyers", "Downtown Austin high-rises"]
  }],
  "openingHoursSpecification": [{
    "@type": "OpeningHoursSpecification",
    "dayOfWeek": ["Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday"],
    "opens": "09:00:00",
    "closes": "18:00:00"
  }]
}
</script>
  • Most real estate site platforms and SEO plugins can add this. Ask your web person to check the type isn’t just “Organization”.
  • Don’t mark up your own star rating from testimonials. Google recommends aggregateRating and review only for sites that review other businesses.
  • Keep it in step with your profiles and pages. Conflicting phone numbers or brokerages are worse than none.

Fix 4: IDX listings on your own pages

Many agent sites show listings through an IDX widget inside an iframe. To a visitor it looks like your page. To many AI fetchers, which read raw HTML and don’t load frames or run scripts, your page is close to empty, and the listings belong to your IDX vendor’s domain.

  • Ask your IDX provider whether listing and search pages render on your own domain as HTML (often called “organic” or “SEO-friendly” IDX), rather than in an iframe.
  • Build your own pages around your own listings and sales: a page per active listing with the full write-up, and short case stories for sold homes. Those are yours to publish however you like, within your MLS’s rules and your sellers’ permission.
  • Don’t expect to win the open-ended search. The portal apps hold that ground inside the assistants. Readable IDX pages help an assistant confirm that you’re active in an area; they won’t make you Zillow.
  • Keep IDX display rules intact on every page: broker attribution, MLS disclaimers and refresh requirements.

Fix 5: a consultation or valuation path, and fast follow-up

An assistant that has picked you still has to make contact for the client. Give it two clear doors.

  • Book a consultation. A buyer or listing consultation on a scheduling tool (Calendly, Cal.com and similar) with a few open slots a week. Name each meeting plainly: “30-minute listing consultation, by phone or at your home”.
  • Request a home valuation. A short form: address, property type, rough condition, timeline, contact details. Say who prepares it and how fast: “Ada sends a written market analysis within two business days.” An instant automated estimate is a different thing; NAR’s AI policy template says automated responses shouldn’t give consumers specific property valuations and should hand pricing questions to a licensed agent.
  • Make the forms agent-friendly. Clear labels, only the fields you need, a free-text box, and a clear confirmation. Puzzle CAPTCHAs can stop a browser agent acting for a real client; server-side spam checks do less collateral damage. Our WebMCP guide explains how a form can describe itself to agents.
  • Answer fast. An inquiry an assistant sent is a real person, often comparing you with the two other agents the assistant also suggested. Route forms to a phone, not just an inbox. If your team can’t always pick up, a speed-to-lead tool such as our sibling company Callingly calls new web leads back automatically.
  • Let the assistants in. Check that robots.txt and your security settings allow OAI-SearchBot and ChatGPT-User and their equivalents. OpenAI says sites that block OAI-SearchBot aren’t shown in ChatGPT search answers. Our robots.txt and Cloudflare guides cover the settings.

Compliance: fair housing, MLS rules and license disclosure

Anything you write for assistants to read is advertising, and an assistant may quote it word for word. The usual rules apply, and a few deserve extra care. This isn’t legal advice; check with your broker and your state’s rules.

Fair Housing Act and steering

  • No statements of preference. The Act bars any notice, statement or ad about selling or renting a dwelling that indicates a preference, limitation or discrimination based on race, color, religion, sex, handicap, familial status or national origin. NAR’s AI policy template gives examples to avoid: “perfect for a young professional couple”, “great neighborhood for families with children”, “walking distance to” a named house of worship, “quiet, mature community”, and any description of a neighborhood’s racial, ethnic or religious makeup.
  • Describe places, not people. HUD’s regulations define steering as restricting someone’s housing choices by word or conduct because of a protected characteristic, including by exaggerating drawbacks or telling someone they wouldn’t be “comfortable” somewhere. Neighborhood pages and FAQs should stick to housing stock, prices, commutes, amenities and where to find official data.
  • Crime and schools: objective, sourced, consistent. In an April 24, 2026 letter, HUD’s Office of Fair Housing said agents don’t violate the Act merely by discussing neighborhood crime or school quality. NAR’s guidance, updated after the letter, says to share the same kind of information with every client, stay objective, and avoid subjective commentary, opinion and hearsay. On your site, that means linking named third-party sources rather than writing “safe” or “good schools”. The letter is about how HUD reads the federal Act; state and local laws often protect more groups, and NAR’s template also lists sexual orientation and gender identity.
  • Don’t answer protected-class questions. If a client, or an assistant on their behalf, asks “which neighborhoods have lots of [religion] families?”, the answer is the facts you’d give anyone, not a demographic steer. If you run a chatbot or give an AI tool approved answers to repeat, write that rule into them and test it.

MLS and listing data

  • NAR’s position is that new technology is fine if it follows MLS rules and data license agreements, and that each MLS decides for itself. Its AI policy template says AI tools must not be used to scrape, redistribute or repurpose MLS data against those rules, and that any AI tool touching MLS data needs brokerage approval.
  • In practice: don’t paste other brokers’ listings into an AI tool or feed your IDX data to a third-party assistant without checking your MLS and IDX agreements. Your own listings, written in your own words with your sellers’ permission, are the safe material to publish. Some MLS data vendors now offer approved routes for AI apps, so ask yours.
  • Have a person review AI-written listing descriptions before they go out. NAR’s template makes that mandatory for its model brokerage.

License and brokerage disclosure

  • Put your license details on the site in text, where an assistant will read them along with everything else. Rules vary by state; two examples:
  • Texas: TREC requires a link to the Information About Brokerage Services form on your business website’s homepage, in a readily noticeable place, and a Consumer Protection Notice link as well.
  • California: the DRE requires your name, 8-digit license number and your responsible broker’s identity on first-point-of-contact materials, including internet ads, with the license number no smaller than the smallest type used. Team names must appear with a licensed member’s name and license number, and the broker’s identity as prominently as the team name.
  • Teams: make the brokerage obvious. A team page that reads like an independent company invites the wrong description from an assistant, and many states restrict that in advertising.

How to check what AI says about you

Set aside half an hour. Use ChatGPT, Google’s AI Mode, Gemini, Perplexity and Claude, signed out or in a private window where you can, and ask the way a client would:

Find me a listing agent in [your city] who knows condos. Who has the best reviews?
Who are the top buyer's agents for first-time buyers in [your area]?
What's my house at [an address you sold recently] worth, and who could do a proper valuation?
Tell me about [your name or team]. What do they specialise in, and where do they work?
I want to sell in [neighborhood] this spring. Book me a listing consultation with [your team].

For each answer, write down:

  1. Were you mentioned? If not, who was, and what do their sites and profiles have that yours don’t?
  2. Were the facts right? Brokerage, areas, specialties, phone, license. Click through to the sources the assistant shows and fix wrong facts where they came from.
  3. Where did the action go? Ask it to book a consultation or request a valuation. Did it reach you, or send the client to a portal’s contact form?
  4. What did it say about your reviews? Assistants summarise review themes. If “slow to respond” comes up, that’s an operations fix.
  5. Did anything cross a fair housing line? If an assistant describes a neighborhood you serve in terms of who lives there, check that none of that wording came from your own pages.

Answers vary from run to run, so repeat the same prompts monthly and look for trends. To see which assistants actually visit your site, read our guide to measuring AI agent traffic. For the bigger picture on giving your site a direct connection to assistants, start with MCP for business websites.

How Salesman Hal handles real estate

Hal gives your site its own MCP server, a connection AI assistants can use directly instead of guessing their way through your pages. Assistants get tools to read your business info, search and read your pages, and ask your own assistant a question, so they describe your areas and specialties from what you’ve written. You choose how that assistant answers: from your pages with your approved answers taking precedence, from your approved answers only (word for word, with anything else referred to you), or not at all. Hal doesn’t screen answers for fair housing, so the approved-answers mode is worth a look for neighborhood questions. When a client wants a consultation or a valuation, Hal sends the request through your existing contact or quote form (Gravity Forms, Contact Form 7 and WPForms through the WordPress plugin, or a HubSpot or plain HTML form Hal finds on your site), or emails it to you if that isn’t possible. If you take consultations in Calendly, Cal.com, SavvyCal, Acuity or HubSpot Meetings, assistants can see open times and book them, and every booking is tagged as AI-sourced. Assistants must confirm the client agreed before Hal sends anything, and requests are rate-limited. Hal has no IDX or MLS connection and doesn’t search listings. Your dashboard shows which AI agents visited and what they sent you.

See how Hal works

FAQ

Will ChatGPT send my listing’s buyers to Zillow instead of me?

If a buyer finds your listing through the Zillow, Redfin or Realtor.com app in ChatGPT, the next step (a tour request or agent contact) runs through that portal’s own flow, as it does on the portal’s website. Strong listing pages on your own site, with a clear way to reach you, give assistants a direct route too.

Can an AI assistant recommend me by name?

Yes, when someone asks for an agent rather than a house. Assistants build that shortlist from what they can read: your site, your profiles and your reviews. No assistant publishes how it ranks agents, so the dependable levers are clear pages, consistent profiles and specific reviews.

Should my website give instant home valuations?

You can link a third-party estimate, but be clear about what it is. NAR’s AI policy template says automated responses shouldn’t give consumers specific property valuations and should pass pricing questions to a licensed agent. A valuation request that you answer personally is both safer and a better lead.

Can I mention schools and crime on neighborhood pages?

HUD said in April 2026 that discussing crime or school quality isn’t by itself a federal Fair Housing violation. Stick to objective data from named third-party sources, give everyone the same information, and avoid opinions such as “safe” or “good schools”. State and local laws can be stricter.

Can I put MLS listings into ChatGPT or a custom GPT?

Check your MLS and IDX agreements first. NAR leaves the call to each MLS, and its AI policy template says AI tools must not redistribute or repurpose MLS data against MLS rules. Your own listings, in your own words, are the safe starting point.

How do I know if AI is sending me clients?

Ask new clients how they found you and add “ChatGPT or another AI assistant” to the options. Check your analytics for visits from chatgpt.com, perplexity.ai and similar, and log assistant visits to your site.

Sources

Further reading

Be the agent the AI can actually reach.

Hal answers assistants’ questions from your own site, sends consultation and valuation requests into your forms and books on your calendar. Free during early access.

Or run the free AI-readiness scan to see what assistants can find on your site today.