AI agents for real estate agencies

A real estate agency's website runs on listings and on trust. Buyers ask whether a home is still available and when they can see it; sellers ask what their home is worth and what you charge; renters ask about pets and deposits. An AI agent can answer from your listings and service pages at any hour and put a showing on the calendar. It also speaks for your brokerage under fair housing law, which shapes what it may say about neighborhoods and who it says a home suits.

Who visits a real estate agency's website

• Buyers. Availability, price history, HOA fees and taxes, showing times, whether they need to sign anything before a tour, and which agent knows the area.
• Sellers. What their home might sell for, what you charge, how you market a listing and how long homes take to sell.
• Renters and landlords, if you lease or manage property: pets, deposits, application steps, management fees.
• Relocating buyers and investors. Buying remotely, video tours, neighborhoods that fit a commute or a budget.
• Other agents. Showing instructions and listing details for their clients.

Which ASDK agents fit a real estate agency

ASDK is in early access. All agents answer from one knowledge base built from your website: listings, agent bios, service pages and policies.

Sales: from listing question to booked showing

Example flow: "Can I see 42 Oak Lane on Saturday?" Sales checks that the listing page still shows the home as active, then asks a few questions: is the visitor already working with an agent, when do they want to move, have they spoken with a lender. If they are not represented, it explains that your team will send a short written buyer agreement before the tour. Since August 17, 2024, NAR's rules require an MLS participant working with a buyer to have one before touring a home, in person or by live video. Sales then offers Saturday slots with the listing team through Cal.com, Calendly or Google Calendar and sends the lead to your team with a summary; native CRM sync is coming soon. Sales is included from the Pro plan.

Chat: listings, process and policies

Example flow: "What are the HOA fees for unit 4B, and are dogs allowed?" Chat answers from the listing and building pages, links them, and notes that fees and rules are confirmed in the association's documents. Asked "how does your commission work?", it repeats your published explanation. NAR's rules require written buyer agreements to state that broker commissions are not set by law and are fully negotiable, so make sure your own pages say the same.

Forms: seller valuation requests (coming soon)

When Forms ships, a seller will be able to answer a few structured questions (address, property type, size, condition, timeline) and request a market analysis, which your agent prepares. The agent never gives a number itself.

Agent-Ready: readable for AI assistants

Agent-Ready hosts an llms.txt file that points assistants to your service areas, agents and listings, and lets browser agents request a showing or a call.

Fair housing and other guardrails

What the agent says counts as your brokerage's statement. HUD's advertising regulation makes it unlawful to make any notice, statement or advertisement about the sale or rental of a dwelling that indicates a preference, limitation or discrimination based on race, color, religion, sex, handicap, familial status or national origin, and it covers oral as well as written statements by people engaged in selling or renting. Instruct the agent to:

• Describe homes, not the people who should live in them. No "perfect for a young couple" or "ideal for retirees".
• Leave neighborhood character alone. NAR's Standard of Practice 10-1 says Realtors shall not volunteer information about the racial, religious or ethnic composition of a neighborhood and shall not engage in any activity that may result in panic selling or steering. For "are the schools good?" or "is it safe?", the agent links official sources, such as the school district's site and the city's public data, and lets the visitor decide.
• Treat every visitor the same. The same question gets the same information and the same offer of a showing.

Availability is confirmed by a person. Listings change by the hour, and the agent answers from your pages. Keep status current on each listing and have the agent say that availability is confirmed by the listing agent.

No pricing or offer advice. "Should I offer under asking?" and "What's my house worth?" go to a licensed agent. So does anything that sounds like legal or tax advice about a transaction.

Referrals as published. If you recommend lenders, inspectors or attorneys, the agent repeats your published list and any disclosure you attach to it, and nothing more.

Say it is AI. ASDK always tells visitors they are talking to an AI, and a handoff to an agent on your team is one step away.

How AI assistants research real estate agencies

A buyer can ask an assistant for "a buyer's agent in Austin who speaks Spanish and works with first-time buyers". It can only match you on what your site states plainly. Make these explicit, as text:

• Service areas: the towns, neighborhoods and price ranges you work in.
• Agent bios with license numbers, languages, specialties and contact details.
• How you work with buyers and sellers: agreements, what is included, how you are paid.
• Listing pages with written descriptions, not only photos: price, size, fees, taxes and status.

RealEstateAgent structured data can describe the brokerage and its offices. The free Agent Readiness Checker shows what AI agents read on your site today, and Agent-Ready keeps llms.txt current as listings change.

Where to start

Write the agent's instructions with your managing broker: the fair housing rules above, the buyer agreement wording and what it never answers. Start with Sales on listing pages, where showing requests come from, and read the first week of conversations. Pricing shows which plan includes each agent.