AI Chatbot for Real Estate: What It Can and Cannot Do for Agencies and Property Managers

Most "real estate chatbot" pages promise MLS search, automatic viewing bookings and CRM pipelines. Before choosing a tool, separate what a website chatbot does from what needs a system integration.

If the left column covers what your visitors actually ask, a website chatbot is the right, inexpensive tool. If your day depends on the right column, read the "other tools" section below first.

Buyers browsing at 22:00 rarely call. They read a listing, form two or three questions, and move to the next portal if there is nowhere to ask. A chatbot on the listing page changes that in four specific ways:

• Listing questions answered from the listing. The bot retrieves the crawled listing pages that match the question, so "is there parking" or "what is the service charge" is answered from what you published, with a link to the listing. When the detail is not on the page, it says so rather than inventing a figure, which matters more in property than in most industries.
• Process FAQs. Offer procedure, deposits, chain timelines, agency fees, what documents a seller needs for a valuation. Upload one clear document per process and the same answer is given every time.
• Lead capture with a 1–5 trust rating. The bot collects contact details in chat, through a pre-chat form or a contact form, rates each lead 1–5 automatically (editable afterwards), and emails the agent instantly. A rated lead lets a busy agent call the serious enquiry first.
• Handover for viewings. When a visitor wants to see the property, the bot offers a person. During agent hours the agent joins the same chat; outside them the widget says so, takes the preferred times and contact details, and the agent confirms in the morning. It is a request, not a calendar booking, and that should be said plainly in the widget text.

For sellers, a "what is my home worth" question is a lead, not a chatbot answer. Configure the bot to collect the address and contact details and hand the valuation to an agent.

An apartment chatbot has a different job from a sales chatbot: fewer high-value decisions, far more repeated questions from prospects and current tenants, many of them after hours.

Two rules for maintenance. First, the bot should never appear to have logged a repair it has not: it collects the details and hands them to a person, or gives the emergency number for gas, water and electrical faults. Second, keep the emergency contact in the knowledge base as plain text with the hours it applies, and test the question "there is water coming through the ceiling" before going live.

Availability changes daily; the bot only knows what the site said at the last crawl. Re-run the crawl or paste the updated availability page whenever units change, and prefer answers that link to the availability page rather than quoting a unit count.

Overseas buyers, relocating professionals and international students enquire in their own language and often outside your office hours. A RAG chatbot detects the visitor's language and replies in it from your English (or any other) content, across 95+ languages, and every widget label can be translated per site so the whole experience matches. Two practical points:

• Keep legal and financial terms consistent in your source documents; the bot translates what you wrote, so a vague clause becomes a vague clause in Portuguese.
• Set a fallback language for the rare case detection fails, and decide which languages your agents can actually take a handover in. The bot can answer in Mandarin; the viewing confirmation still has to come from a person.

The lead email arrives instantly and the full transcript is in the team inbox, so the agent can see what was asked even if the exchange happened in a language they do not read.

Plan limits to know: knowledge items are capped at 100 on Free, 500 on Starter and unlimited on Grow; conversations at 15, 150 and unlimited a month. An agency with a few dozen listings and a handful of guides fits Starter; a portfolio of hundreds of pages needs Grow at $7.49 a month. See pricing and how to train a chatbot on your data.

1. Install. On WordPress use the Vatdi plugin; on any other site, including custom builds and Webflow or Squarespace, paste the one-line script. Brand colours and logo are detected from the site automatically.
2. Crawl and upload. Run the website crawl over listings, area guides and process pages, then upload the fee schedule, application guide and tenant handbook.
3. Write the 50-question test set from your inbox and phone log: "is 14 Elm Road still available", "do you accept pets", "what documents do I need to apply", "what is your fee for selling", "how do I report a leak". Include five questions the bot must not answer, such as a legal opinion or a valuation.
4. Score the answers 0–10 and fix gaps by adding documents, not by rewording prompts.
5. Configure leads. Choose pre-chat form or in-chat capture, decide which agent or inbox receives the instant email, and check the 1–5 trust rating on a few test leads.
6. Set handover topics and agent hours. Viewings, valuations, repairs, complaints and application status go to a person; enter the office hours so the widget is truthful at night.
7. Translate the widget labels for the languages your visitors use, review the AI translations, and set the fallback language.
8. Go live, read every transcript for two weeks, then move to a weekly review of quality grades and leads.

All reports are included on every plan; leads export to CSV for whatever CRM you use.

• No MLS, IDX or portal integration. The bot answers from your website and uploads. If a listing is only on a portal and not on your site, the bot does not know it.
• No CRM or calendar sync. Leads arrive by instant email and CSV export; viewing requests are handed to a person, not booked. There are no Zapier or webhook connectors.
• Website widget only. No WhatsApp, Messenger, Instagram or SMS. Enquiries through those channels still need your existing process.
• Freshness depends on re-syncing. Availability and price changes reach the bot when the pages are re-crawled or re-pasted, not the instant your listings system changes.
• No tenant account access. Balances, application status and ticket numbers live in your property-management system; the bot explains and hands over.
• Regulated advice. Configure the bot to decline legal, mortgage and valuation opinions and hand those to a qualified person.
• Attachments. The widget does not accept file or image uploads, so a tenant cannot send a photo of the fault in chat; direct them to email or your portal.

Data handling details, including where conversation data is processed and how deletion requests work, are on the trust page.

If pipeline sync is essential, look first at the chat product from your own CRM vendor; CRM-native chat writes leads straight into the pipeline, which Vatdi cannot. HubSpot has a free chatbot builder inside its free tools and paid hubs from about $15–20 per seat a month as of September 2026, and is a reasonable choice for a brokerage already living in that CRM. For tenant self-service such as balances and ticketed repairs, the resident portal in your property-management software is the right place, with a website chatbot handling the questions around it. Drift is built for B2B revenue teams at quote-only pricing and is not a fit for agencies. A website chatbot such as Vatdi is the low-cost layer that answers the questions and captures the lead; it sits alongside those systems rather than replacing them. Features and integrations list exactly what is included.