Real estate chatbots that book viewings, not just chat

Real estate chatbots are almost all identical: a scripted “hi there, are you looking to buy or rent?” that leads into a form. Every visitor sees the same script. Almost none of them actually leave contact details.

Here’s the version that works.

Start with the property, not the form

The moment a visitor lands on a property page, the chatbot’s first turn should be about that property. Not “how can I help you today?” — that’s a chat about you. The visitor is thinking about the house they can see.

Set the greeting per property (dynamic per URL) to something like:

“Hi — questions about the 3-bed on Willow Lane? I can pull up floor plans, viewing availability, or the schools list.”

That greeting produces 3–4× the engagement of “hi, how can I help?” because it signals the bot knows the context.

Load per-property knowledge

Anserra’s knowledge base can be per-chatbot, but it can also be per-property if you scale. Each property gets:

• The listing description.
• Floor plan bullet points (rooms, dimensions).
• Nearby schools, transit, amenities — pulled from your existing data source.
• The council tax band, if you have it.
• The current viewing calendar (integrated live, not embedded as text).

The chatbot answers property questions from this content. When someone asks “what’s the council tax?” it says the actual number, from your data.

Qualify before you capture

The mistake I see: chatbots that pop a lead form after two messages. This kills the conversation.

The pattern that works:

1. Answer 2–3 property questions.
2. Naturally ask “Would you like to see it in person?”
3. If yes, offer specific times: “I can fit you in this Thursday at 5pm or Saturday at 11am — which works?”
4. Only THEN capture name + phone/email + preferred time.

You’re not selling contact details for content. You’re offering something they came for (viewing the property), and the contact detail is how you deliver it. Conversion goes up 2–3× in the data I’ve seen.

Handoff to the agent

The moment a viewing is booked, notify the assigned agent immediately — SMS, Slack, or push notification. The visitor is thinking about the property right now; if an agent calls back within an hour, close rates go up dramatically.

Set the escalation to the agent’s phone number for the specific listing, not to a shared inbox where it gets lost.

Handling the “how much can I offer” question

Never let the chatbot answer this. It doesn’t have your negotiation strategy, it doesn’t know the seller’s motivation, and every answer it gives is either too low (annoying the seller) or too high (over-committing the buyer).

Configure the system prompt: “If the visitor asks about offer strategy, price negotiation or urgency, always escalate to an agent and never suggest a specific number.”

Lead quality

Lead volume is easy. Lead quality is what matters.

Track two numbers:

1. Viewings booked per lead. > 40% means the bot is qualifying well.
2. Show-up rate. > 80% means your reminders and confirmations are working.

If lead volume is high but viewings booked is low, the chatbot is capturing tyre-kickers. Tighten the qualification: ask about timeline (“looking to move in the next 3 months, 6 months, longer?”) and budget (“what’s your budget range?”). Set the escalation to only trigger when both are within range.

Multi-language

Real estate skews international. If you sell to expat buyers, load the chatbot in three languages: your primary market’s language, English, and Mandarin. Configure the chatbot to detect the visitor’s language automatically and respond in kind.

Anserra does this automatically when the model is set to “match visitor’s language.” The system prompt handles it.

The one thing everyone gets wrong

They treat the chatbot as a lead-gen tool. It should be treated as a viewing-booking tool. Volume of viewings booked is the number that maps to revenue. Everything else is theatre.

If you want to build this: get an Anserra embed key, load your listings, wire up your viewing calendar via webhook. Done in an afternoon.