The question agents ask is whether chat widgets work. The better question is what happens in the ninety seconds after a visitor types a question into yours.
What the numbers do and do not say
Every chat widget vendor publishes conversion figures, and they are impressive. We went looking for a primary source behind the commonly quoted numbers and could not find one we would stand behind, so this post will not repeat them.
That is a limitation worth naming rather than papering over. The honest position is that a chat widget changes visitor behaviour in both directions, that the direction depends on how it is run, and that no published figure we could verify tells an individual agent which way it will go for them.
What follows is the mechanism instead. It is more useful anyway, because the mechanism is the part you control.
The case for chat: it lowers the cost of asking
A contact form asks for a commitment before it gives anything. Name, email, phone number, sometimes a message box and a dropdown, all submitted before a single question gets answered. For a visitor who has one small question and no intention of choosing an agent today, that is a lot to hand over.
Chat inverts it. Type a sentence, get a reply. The visitor gives up almost nothing and finds out whether you are helpful before deciding whether to identify themselves.
That difference matters most for a specific person: early in their research, uncertain, and nervous about talking to an agent because they expect to be pursued. First-time buyers are the clearest example. In NAR's 2025 Profile of Home Buyers and Sellers, 70 percent of buyers used a mobile device or tablet during their search, and typing one sentence on a phone is far easier than completing a five-field form on one.
There is a second benefit that has nothing to do with leads. The questions people type into chat are the questions your website failed to answer. Read a month of transcripts and you have a list of pages to write, in your visitors' own words. That is the same value our post on what to track in website analytics gets from search queries, delivered more directly.
The case against: the unanswered chat
Here is the cost agents underestimate.
A visitor types a question at 9pm. Nobody answers. They check back the next morning and it is still unanswered. They have now learned something specific about you: you invite questions and do not respond to them.
That is worse than having no widget. A site with no chat box makes no promise. A site with a chat box makes an implicit one, and breaking it is a live demonstration of your responsiveness on the page where a stranger is deciding whether to trust you with the largest transaction of their life.
The expectation is set by the format itself. Nobody expects an email answered in two minutes. Everyone expects a chat message answered like a text, because that is what it looks like. You do not get to reset that expectation by intending otherwise.
This is also why the honest automatic reply matters more than any other setting. A message saying "I am with clients until 5pm and will reply this evening, or call me at this number if it is urgent" is acceptable and even reassuring. Silence is not. Our post on lead response time covers why speed decides so many of these first contacts.
AI chatbots and what changes
An AI chatbot solves the availability problem. It always answers, at 9pm and on Sunday, and it never gets tired of the same question.
It introduces a different one. AI chatbots generate answers, which means they can produce something plausible and wrong. In most industries that is an inconvenience. In real estate it can be a buyer who arrives at a showing believing the strata allows two dogs, or a seller who anchors on a number a bot volunteered.
That makes the boundary worth drawing carefully. A confident wrong answer about a specific property is a worse outcome than a delayed human reply, because the visitor acts on it.
Three rules keep this manageable.
Feed the bot only facts you have published and can check. If the answer is not already on your website, the bot should not be constructing one.
Forbid the categories where being wrong is expensive: what a home is worth, what to offer, what a property's condition is, anything about financing, and anything a lawyer or an inspector should answer. Our post on home valuation tools covers why the valuation case in particular needs a human.
Read the transcripts weekly for the first month. You will find things it said that you did not expect, and that is the only way to find them.
What a bot can honestly handle
The list is shorter than vendors suggest and longer than sceptics assume.
Hours and availability. Straightforward, factual, and asked constantly.
Where you are and which areas you cover. Also factual, and useful for filtering out people outside your market before either of you spends time.
Booking a showing or a call. This is the strongest case, with one condition: it must connect to your real calendar. A bot offering times you are not free creates a cancellation and a bad first impression.
General process questions. What happens after an offer is accepted, what a home inspection involves, roughly how long a sale takes in your market. These have stable answers you can write once.
Routing. Working out whether someone is buying, selling, or renting, and sending each to the right next step.
Collecting a callback request. Name, number, and one line about what they want, delivered to your phone.
What it should hand off, always: anything about price, anything about negotiation, anything specific to one property's condition or history, anything about a person's finances, and anything where the visitor sounds upset. The handoff should be explicit and fast. "I do not have that one. I can have [agent name] call you. What number is best, and is now a good time?"
A bot that admits a limit and hands off cleanly reads as competent. A bot that produces a confident non-answer reads as a wall, and the visitor leaves.
The speed cost on mobile
A chat widget is third-party code. It loads on every page of your site, from someone else's server, and it does so for every visitor including the ones who will never open it.
Two specific effects matter. The first is weight: another file to download on a cellular connection while a visitor stands outside a listing. The second is layout movement. Many widgets load after the main content and then insert the bubble, which pushes the page as it appears. Google measures that as Cumulative Layout Shift, one of the Core Web Vitals, and the good threshold is under 0.1.
The practical version of that problem is a visitor reaching for your phone number as the layout shifts and tapping a chat bubble instead.
Whether this is acceptable depends on what else is loading. One widget on an otherwise clean site is manageable. One widget on a site already carrying two analytics tools, a heatmap recorder, a review carousel, and an IDX embed is the sixth item on a list nobody has audited. Our post on website speed and mobile performance covers how to measure the total rather than arguing about any one piece.
Two settings reduce the cost without removing the widget. Load it after the rest of the page rather than alongside it, so the content appears first. And turn off the automatic opening on phone screens, where an unprompted panel covers a meaningful portion of what the visitor came to read.
Chat versus a form: keep both
Agents sometimes treat this as a replacement decision. It is not.
The form serves the person who has decided to make contact. They are willing to give details and they want a record that their message was sent. Removing the form to push people into chat means someone ready to act has to wait for a reply before they can reach you at all.
Chat serves the person who has not decided. One question, low commitment, no form.
There is one setting that ruins chat's advantage: demanding an email address before the conversation starts. That converts chat into a form with extra steps and loses the exact visitors it was meant to help. Ask for contact details after you have answered something, when the visitor has a reason to give them.
And keep your phone number visible at the top of every page regardless. It remains the fastest path for the highest-intent visitor, and it costs nothing to load. Our post on contact pages covers how the pieces fit together.
When an agent should skip it entirely
Some honest cases for not adding one.
You cannot answer within business-hours minutes and will not use an answering service. The widget's value is response speed. Without it you are building a record of unanswered messages on your own website.
You show homes most of the day. An agent in a car for six hours is not staffing a chat. A visible phone number serves those visitors better.
Your site is already slow. Fix the weight first. Adding a widget to a site that takes five seconds on cellular data trades a load problem for a worse one.
You get few enough enquiries to handle personally. A solo agent receiving four website enquiries a month does not have a volume problem that chat solves. Effort spent on the pages that produce those four is worth more.
You would use it mainly to capture emails. A widget deployed as a collection device, opening automatically and asking for an address before helping, teaches visitors to close it. That habit costs you the visitors who had a real question.
The takeaway
A chat widget is a promise to respond, and it pays off only when the promise is kept. Judge it on what happens in the first ninety seconds rather than on any vendor's conversion figure, because that is the variable you control. If you can answer quickly, or you run a bot that handles hours, location and booking and hands everything else to you with a clear next step, it earns its place and its page weight. If you cannot, a visible phone number and a short contact form do the same job without teaching a stranger that you do not reply.
