A chatbot for real estate can help visitors understand a property, compare the services an agency offers and find the right route for a viewing or valuation enquiry. People often arrive with short questions that assume context: “Is there parking?”, “Can I bring a dog?” or “How do I book?” A useful assistant needs to establish which property or service they mean before applying an answer.
A public knowledge chatbot is most dependable when it explains published information and captures clear enquiries. It should not claim that a property is still available, that a viewing has been reserved or that an applicant has been approved unless a verified system supplies that result. This guide focuses on building that useful, bounded experience for an agency website.
Distinguish the agency’s audiences
Buyers, tenants, sellers and landlords come to the same website with different goals. A tenant asks about a rental process, while a landlord asks about management services. The assistant should recognise the difference and use the relevant material rather than sending everyone the same generic sales response.
Start with the audience that generates the most repetitive public questions. For a lettings agency, that may be viewing guidance and application steps. For a sales-focused office, it may be valuation enquiries, marketing services and local office information. A defined starting scope makes it easier to prepare reliable knowledge.
Use a greeting that explains the available help without overpromising: “I can help with our property information, services and enquiry process. Which property or service are you interested in?” Let visitors ask naturally, and clarify only when the missing detail changes the answer.
Give every property a stable identity
Property information should include a clear reference, location and the date of the material where appropriate. Street names can repeat, and several flats may share a building name. A reference helps the bot connect a follow-up question to the correct listing rather than guessing from a partial address.
Keep descriptions factual. Include the published room details, parking information, outdoor space and viewing route. If a feature is absent from the source, the assistant should not infer it from a photograph or a nearby property. “The listing does not specify that; the team can confirm” is more reliable than a plausible guess.
Avoid mixing several listings into a paragraph with unclear boundaries. Each property should have its own heading and associated facts. This helps both human maintenance and retrieval, particularly when questions mention a reference rather than the full listing title.
Treat availability as changing information
A listing imported into knowledge is a snapshot of published content. It does not automatically become a live property feed. A home can be reserved, withdrawn or updated after the source was added. Make the current listing page or agency contact the route for confirming availability.
Set an operational process for removing or updating sources when properties change. If the agency cannot maintain individual listings frequently, begin with stable service information and use links to the current property search. That narrower approach can be more useful than a large collection of stale property descriptions.
Test time-sensitive questions directly. “Is it still available?” should not receive a confident yes simply because the listing exists in the knowledge. “Can I view tomorrow?” should lead to the approved enquiry or booking process, not a made-up appointment confirmation.
Explain processes without giving individual decisions
Public guidance can describe the usual steps for arranging a viewing, requesting a valuation, submitting an enquiry or contacting a property manager. Organise these instructions in order, and include what information the agency asks for through its normal forms.
Do not use the public chatbot to decide whether someone will be accepted for a tenancy, mortgage or other regulated arrangement. A general process description is not an individual assessment. Where a question depends on personal circumstances or professional advice, direct it to the appropriate qualified team or established channel.
Avoid filling knowledge with unsupported legal summaries. Use the agency’s approved public policies and ask the responsible staff to review the wording. The bot should explain the service it has evidence for, rather than improvising obligations, rights or guarantees based on a visitor’s brief description.
Make agency services easier to compare
Sellers and landlords often need to understand what a service includes before they enquire. Prepare clear descriptions of valuation appointments, marketing options, management services and any published service tiers. Explain exclusions as well as included tasks.
If pricing depends on a discussion, say that plainly in the source material. The assistant can explain the factors the team considers and show the enquiry route. It should not invent a fee from a general description or imply that every customer receives the same arrangement.
Use examples carefully. A published case study can show how a service works, but a past sale or letting does not guarantee the same outcome for another property. Label examples as examples and avoid turning an illustration into a forecast about price, demand or timing.
Set up the assistant using approved knowledge
In Chaat, create the bot through the guided assistant or manual settings. Describe the agency, the audience and the distinction between public information and confirmation by staff. Ask the bot to identify the property before answering listing-specific questions and to keep replies concise.
Add public URLs, approved PDFs or pasted text. Review extracted website knowledge for missing references, old listings and content from unrelated areas of the site. Remove sources that are no longer useful. If information lives in a private property system, do not assume a public URL import gives the bot access to it.
Set suggested questions around practical actions: “How do I arrange a viewing?”, “What does property management include?” or “How can I request a valuation?” Choose the name, colours and launcher style to fit the agency’s website while keeping links and controls easy to read.
Design enquiries that staff can act on
If contact requests are enabled, make the purpose clear. A viewing enquiry should identify the property and the visitor’s question. A valuation enquiry may need a different follow-up. The assistant can guide people toward the correct route without claiming that the enquiry has already become a confirmed appointment.
Decide who monitors requests and whether email notifications are useful. Test the complete path from the public chat to staff review. The team should be able to identify the topic and read the relevant conversation before responding through its normal process.
Keep private documents in the appropriate systems. A public chat should not encourage visitors to upload identity documents, financial records or application evidence as general knowledge. Where the agency requires those materials, direct the visitor to the approved submission process.
Test ambiguity and changes of property
Create a test conversation in which the visitor starts with one listing and later asks about another. Check that the bot updates the context and does not carry over parking, room or pricing details. Also test abbreviated addresses and a reference number without a full property name.
Ask about a feature that the listing does not specify. Ask for a viewing outside the published process. Ask whether a property can be held. The assistant should distinguish what the source establishes from what requires confirmation, while still giving a useful next step.
Use Chaat’s Test source option to inspect answers involving important facts. Check that any link points to the correct listing or service page. A working link to the wrong property can be more confusing than no link, so include destination checking in the review.
Introduce the widget without blocking enquiries
Test the embedded chat alongside existing property controls. On a phone, the launcher should not cover “Arrange a viewing”, the map controls or a cookie banner. Choose a position and size that work with the site rather than relying only on how the preview looks inside the dashboard.
Keep existing enquiry forms and office contact details available. The chat should add a conversational route, not force visitors through an assistant before they can reach the agency. A shareable link can help colleagues review the experience before it is placed across the website.
Start with a small set of pages and review the first conversations. If users consistently ask questions the sources cannot answer, decide whether to add approved information or narrow the assistant’s scope. Expanding coverage should follow content readiness.
Maintain accuracy as the portfolio changes
Give one person responsibility for the source list and agree how listing changes reach them. Record which information is stable and which needs frequent review. Service descriptions may remain accurate for months, while individual property details can change much sooner.
Review conversations for recurring uncertainty about viewing routes, fees, office responsibilities or property references. These patterns can improve the website’s information architecture as well as the bot. Measure whether enquiries arrive with useful context and whether reviewed answers match the published facts.
A real estate chatbot succeeds when it helps a visitor understand the available information and reach the right team with a clear question. It should save repetitive explanation without manufacturing availability, appointments or individual decisions. That balance makes it a useful extension of the agency’s public website.