AI Chatbot for Real Estate: Answer Leads and Book Viewings 24/7

A buyer finds one of your listings on a Sunday evening with three questions: is it still available, is there parking, and can they see it on Saturday? A landlord wants to know what full management costs. A homeowner wonders what their flat is worth. If the answers wait until Monday morning, some of those people will already have booked with the agency that replied first.

This guide is for independent estate agents, brokerages, letting agents and property managers who want a website chatbot that answers enquiries around the clock and passes the real opportunities to a person. It covers which questions the bot should answer, act on or hand off, what to feed it, lead capture, bookings, fair housing, metrics and setup. The worked example is PepoChat, but the checklist applies to any tool.

None of this replaces agents. In the National Association of Realtors' 2025 Profile of Home Buyers and Sellers, published in November 2025, 88% of home buyers used an agent or broker. The chatbot's job is to get the right people to an agent sooner, with the basic details already collected.

What does an AI chatbot for real estate do?

An AI chatbot for real estate does four jobs on your website: it answers questions from your own content, collects contact details when a follow-up is needed, books viewings and valuations, and hands conversations to an agent. Everything that needs a licence, a negotiation or a judgement stays with people.

A real estate chatbot is an AI agent on an agency's website that answers questions from buyers, renters, landlords and sellers using the agency's own listings and documents, and turns serious enquiries into booked appointments or agent follow-ups. The useful kind searches your content before every reply and says "I don't know, let me get an agent" when the answer is not there. The harmful kind improvises from general knowledge, which in property means invented availability, guessed fees and opinions about neighbourhoods. The difference is mostly configuration: what you let it read, do and say.

What questions do buyers, renters and sellers ask a real estate chatbot?

Most questions are about availability, viewings, costs, the buying or letting process and the area. A chatbot can answer or act on most of them; offers, negotiation, complaints and anything about who lives in an area should go to a person.

The table sorts common questions into three outcomes. Answer means the bot replies from your content. Act means it does something, such as booking a slot or filing a request. Hand off means it passes the conversation to an agent, usually after collecting contact details.

The questions a bot can finish are the ones your website already answers, or should. The riskiest rows are not the complicated ones: "Would the seller take less?" is obviously a job for an agent, while "Is it a safe area?" looks innocent and is the most likely to produce an answer you would never put in writing. The fair housing section explains why.

What should go in a real estate chatbot's knowledge base?

Put in your listing pages, area guides, fees and terms, and the step-by-step guides for buying, selling, letting and renting, plus short Q&A pairs for the questions your office answers every week. Keep anything subjective about areas or residents out.

A knowledge base is the private collection of pages, documents and snippets the chatbot searches before it replies. It answers only from what is there, so its contents decide both what the bot knows and what it might repeat.

Listings. A crawl follows links on your own site up to 200 pages and three levels deep, and include or exclude filters can limit it to your property pages. Website pages are re-fetched weekly and only changed pages are re-indexed, which is not fast enough for a flat that was let this morning. Press Sync website pages whenever a status changes, use Import again when new listings go up, and tell the bot in its instructions that an agent confirms availability before any viewing.

Two catches are common on property sites. Listings rendered by a third-party script or inside an embedded frame may import as an empty shell, because the crawler sees what an anonymous visitor's first request sees; check the imported pages and upload the details as a document if needed. And facts shown only in photos or floor plans, such as a parking space, are not in the page text, so write them into the description.

Area guides. Describe places, not people: the station and journey times, schools by name, parks, shops, the planned tram line. The chatbot repeats your wording to anyone who asks, so a guide calling a street "ideal for young professionals" puts that phrase into hundreds of conversations.

Sold and let listings. Remove them or mark them clearly, or the bot will offer viewings of homes you no longer have.

How do you capture real estate leads without a form wall?

Let the chat start without a form, and ask for contact details only when the visitor wants something that needs a follow-up, such as a viewing, a valuation or a call back. Someone asking about parking should get an answer, not a form.

A form wall is a contact form shown before any help is given, which asks for commitment before the visitor has received anything. The chatbot lead generation guide covers the trade-offs.

In PepoChat, chats start anonymous. Under Lead Capture you choose when the agent asks: before the first message, when the agent needs a follow-up, or just before a handoff. For an agency, when the agent needs a follow-up usually fits best: the agent answers simple questions straight away and asks once, in one friendly sentence, when a request needs someone to get back to the visitor. Email is always collected; tick name and phone too, because negotiators call. A booking collects a name and email anyway.

Everyone who shares an email lands on the Contacts page, with a badge showing how they arrived: chat, booking, verified email or the pre-chat form. For each new contact, PepoChat can post a Slack card or send a signed webhook to Zapier, Make or your CRM on every plan, and email your team on Starter and Growth. Each notification links to the conversation, so whoever calls back sees which property the person asked about.

How does a chatbot book viewings and valuations on a real calendar?

Connect a scheduling calendar, create a viewing type and a valuation type with a rule for when each may be offered, and let the bot offer live slots and book the one the visitor picks. The calendar re-checks the slot, assigns a negotiator and sends the confirmation.

In PepoChat, booking runs through a PepoSmart scheduling calendar connected with an API key. You switch on the appointment types the agent may offer, up to 12, and write a when to offer this note for each. Two notes that work for an agency:

• Valuation: "Offer to homeowners who ask what their property is worth or about selling or letting with us. Never offer to tenants or buyers."
• Viewing: "Offer when a visitor asks to see a property the listing shows as available. Ask which property first and repeat the address before booking."

The agent then checks availability up to 14 days ahead and offers a handful of slots in the visitor's own time zone. If a type has several locations, or your negotiator calls the visitor, it asks for their preference or phone number first. PepoSmart re-validates the slot so two people cannot book the same time, assigns a host with round-robin if you use it, and sends confirmations and reminders. A Find a time button opens the full booking page in the chat for visitors who prefer to browse. The appointment booking docs list each step.

Know what the booking carries. It records the visitor's name, email and time; which property they want to see lives in the conversation, which the lead notification links to, so check it before arranging keys. A calendar slot is also your negotiator's free time, not the property's: if a seller or tenant must agree to each viewing, the calendar cannot know that.

US agents have one more rule to plan around. According to NAR's consumer guide to written buyer agreements, since 17 August 2024 buyers working with a Realtor are asked to sign a written agreement before touring a home, in person or virtually; visiting an open house on your own does not require one. Add a scheduling instruction such as "Tell buyers that the agent will send a short buyer agreement to sign before the viewing", so nobody is surprised at the door.

For negotiators meeting strangers at empty homes, keep your usual confirmation call before a first solo viewing; PepoChat can also have visitors verify their email with a one-time code.

When should a real estate chatbot hand off to an agent?

Hand off whenever money, judgement or a protected characteristic is involved: offers, price negotiation, legal or mortgage questions, complaints, tenant emergencies and questions about who lives in an area. Also hand off whenever your content has no answer.

PepoChat's agent escalates on its own when a visitor asks for a person, when it cannot find an answer and the visitor accepts its offer of help, and when something fails. Every member of the workspace gets an email headed "A visitor needs a person", with the last three messages and a link to the conversation. In the team inbox the full transcript is waiting, Assigned to passes it to the listing's negotiator, and internal notes stay hidden from the visitor.

Two settings matter for agencies. Business hours make the bot say when the office reopens instead of promising a quick reply at 10 p.m., and ask for an email before it hands over. A no one answers live chat mode suits a sole agent: the bot never promises a live person, collects an address and passes the conversation on. Visitors are not emailed when you reply in the chat, so call or email anyone time-sensitive. The human handoff guide covers what a good escalation looks like.

Fair housing and AI disclosure: what must a real estate chatbot never do?

A real estate chatbot speaks for your agency under the same anti-discrimination rules as your agents, so it must never steer people or describe a neighbourhood by protected characteristics, and it should say plainly that it is an AI. Write these rules in before launch.

Steering is guiding people towards or away from homes or areas because of who they are. The US Department of Justice's Fair Housing Act overview gives "steering homeseekers to certain areas based on race" as an example of illegal discrimination. As of October 2026, HUD's Fair Housing Act page lists the protected characteristics as race, color, national origin, religion, sex, familial status and disability, and covers renting, buying and getting a mortgage.

The statute also covers words. 42 U.S.C. 3604(c) makes it unlawful to make, print or publish any notice, statement or advertisement about a home's sale or rental that indicates a preference, limitation or discrimination based on those characteristics. Treat every chatbot reply like advertising copy published in your name. "Perfect for a young family" states a preference about familial status; "three bedrooms, a garden and a primary school 400 metres away" states facts.

In Great Britain the Equality Act 2010 protects nine characteristics, including age, disability, race, religion or belief, sex and sexual orientation, and applies when buying or renting property. Other countries have their own lists. The practical rule is the same everywhere: describe the property and the place, never the people.

In PepoChat, hard rules go in the Never do this field of the agent instructions, up to 2,000 characters. When a visitor asks for one of these, the agent declines briefly and offers what it can help with instead. Lines worth adapting:

• "Never describe an area or its residents by race, colour, national origin, religion, sex, disability, age or family status."
• "Never say a home or an area suits or does not suit families, children, couples, older people or any other group. Describe the property's features instead."
• "Never give an opinion on whether an area is safe. Say an agent can point to official crime statistics and offer to connect them."
• "Never ask visitors about their nationality, religion, health, disability or whether they have children."
• "Never give a valuation figure or a view on what a seller will accept."

Then test it: ask the same questions as a parent, a pensioner, someone who mentions their faith and someone writing in another language, and check that the answers, listings and offered slots do not change.

Disclosure. Tell visitors they are talking to an AI: give the assistant a name that says so, and have the greeting say it. In the EU this has been a legal duty since 2 August 2026 unless it is obvious, and several US states have their own rules; the guide to chatbot disclosure laws summarises them. Keep the bot out of regulated advice as well: mortgage, legal and tax questions go to a qualified person.

Which metrics show a real estate chatbot is working?

Count outcomes, not chats: viewings and valuations booked, leads with a phone number, handoffs answered on time, and questions the bot could not answer. A busy chatbot that books nothing is a cost, not a channel.

PepoChat's Analytics shows satisfaction votes and auto-labelled topics; it does not chart volume or resolution rates. Work those out from the inbox's CSV export, up to 5,000 conversations at a time, and from the Contacts export. The chatbot metrics guide explains which numbers mislead.

How do you set up a real estate chatbot? A walkthrough

Plan an afternoon: import your site, add your documents, write the rules, switch on lead capture, connect the calendar, set your hours, install the widget and test. Here is the sequence in PepoChat.

1. Import your listings and pages

Under Knowledge Base, add your website as a sitemap import or a crawl, using an include filter for your property and area-guide paths if the site has a large blog. Check the imported rows for empty listing pages.

2. Add documents and Q&A pairs

Upload your fees schedule, buyer's, seller's and landlord's guides and the tenant application checklist as PDF or DOCX files. Add Q&A pairs for the questions your office answers every week, written the way people ask them.

3. Write the agent instructions

In How to sound and behave, introduce the assistant as an AI, set the tone and say what to focus on. In Never do this, paste your fair housing and no-valuation rules from the section above.

4. Turn on lead capture and notifications

Choose when the agent needs a follow-up, tick name and phone, and switch on a Slack card, a webhook into your CRM, or team emails on a paid plan.

5. Connect the calendar

Paste your PepoSmart API key under Appointments, switch on a valuation type and, if it fits, a viewing type, and write their notes and any scheduling instructions.

6. Set your hours

Under Team availability, choose business hours with an away message such as "The office reopens at 9:00 on Monday. Leave your number and a negotiator will call you."

7. Install the widget

Paste one script tag into your site's template, or use the WordPress plugin. The hosted chat page gives you a link for QR codes on boards, window cards and brochures.

8. Test before visitors do

Open Test the agent and work through the question table above, including the fair housing prompts. Test chats are free and stay out of Analytics and Contacts, but bookings and integration actions run for real, so cancel any test booking afterwards.

What to do next

Start with the question table: mark which questions your website already answers and write the missing pages or Q&A pairs. Then switch on one booking type, valuations, and read a week of transcripts before adding viewings.

You can create a free PepoChat workspace and try all of this before choosing a plan; every feature apart from new-lead emails to the team is on the free plan, with 200 AI replies a month, and the pricing page shows each plan's limits. Before you switch on viewings, read how to let an AI chatbot book appointments from your real calendar.