Phone enquiries

Live chat on agency websites: does it convert?

Staffed live chat, chatbots and form-only agency websites compared on leads, response time and workload, with a decision checklist.

Ava research teamPublished 6 min read

On this page
  1. Where your website sits in the buyer's journey
  2. The three models
  3. What visitors expect from chat now
  4. Decision criteria
  5. Scoping a chatbot safely
  6. Handover rules that make hybrid work
  7. Writing the bot's answers
  8. Chat and lettings fairness
  9. Staffing a chat rota
  10. Privacy and transcripts
  11. Example: the workload sum
  12. Measuring whether chat converts
  13. Frequently asked questions
  14. References

Key takeaways

  • 44% of UK adults used an AI chatbot at least monthly in 2024, so chat is familiar to many buyers and tenants[1].
  • Portals dominate browsing; your website mostly serves people who already know you, so chat should help them act.
  • Staffed chat converts well but only if someone answers within a minute or two during the hours you advertise.
  • A chatbot should answer facts and capture details, then hand over; 91% of adults worry about decisions made without human input[2].

Where your website sits in the buyer's journey

Most property searching happens on portals. Rightmove reported a 90% share of time spent on UK property portals in the first half of 2026[3]. People land on an agency website for narrower reasons: to check who you are before a valuation, to look for a listing they saw elsewhere, to find a phone number or to report a repair. Live chat earns its place if it helps those visitors do what they came for faster than a phone call or a form.

Browsing time has also become less predictable. Visitors spent 8.4 billion minutes on Rightmove in the first half of 2026, down from 9.1 billion a year earlier[4]. When people browse less, each enquiry is worth more, and the speed of the first reply matters more.

Minutes spent on Rightmove, first half of each year
Minutes spent on Rightmove, first half of each yearLine chart. H1 2024: 8.3 billion minutes; H1 2025: 9.1 billion minutes; H1 2026: 8.4 billion minutes.8 billion minutes8.25 billion minutes8.5 billion minutes8.75 billion minutes9 billion minutes9.25 billion minutesH1 2024H1 2025H1 2026H1 2024: 8.3 billion minutesH1 2025: 9.1 billion minutesH1 2026: 8.4 billion minutes8.4 billion minutes8.3 billion minutesMinutes spent on Rightmove, first half of each yearLine chart. H1 2024: 8.3 billion minutes; H1 2025: 9.1 billion minutes; H1 2026: 8.4 billion minutes.8 billion minutes8.25 billion minutes8.5 billion minutes8.75 billion minutes9 billion minutes9.25 billion minutesH1 2024H1 2026H1 2024: 8.3 billion minutesH1 2025: 9.1 billion minutesH1 2026: 8.4 billion minutes8.4 billion minutes8.3 billion minutes

Source: Rightmove plc, 2026[4]

Show the data
PeriodValue
H1 20248.3 billion minutes
H1 20259.1 billion minutes
H1 20268.4 billion minutes

The three models

Agency websites usually fall into one of three models, or a hybrid of them. Each has a different effect on leads and on staff time.

Website enquiry models compared
ModelLead qualitySpeed for the visitorStaff workloadMain risk
Form onlyGood details, low volumeSlow: waits for a callbackLow and predictableVisitors leave rather than wait
Staffed live chatHigh when answered fastFast in staffed hoursHigh; needs coverUnanswered chats look worse than no chat
ChatbotMixed; depends on designInstant, around the clockLow, plus set-up and reviewWrong answers or dead ends
Hybrid: bot first, human handoverHigh if handover worksInstant, then humanModerateHandover gaps out of hours

What visitors expect from chat now

Chat is familiar. In government tracking research, 44% of UK adults said they used an AI chatbot for personal or work purposes at least once a month in 2024, and 23% at least once a week[1]. Portals are experimenting too: Rightmove said early results from its conversational search showed close to double the propensity to send leads[5]. Visitors will increasingly expect to ask a question in plain words and get a useful answer.

Familiarity is not the same as trust. The same research found 91% of UK adults were concerned about important decisions being made by computers without human input[2]. For an agency, that means a bot can answer "is it still available?" or "what is the council tax band?", but questions about offers, deposits, referencing outcomes or valuations need a person.

Decision criteria

  1. Can you answer a staffed chat within two minutes during every hour it shows as online? If not, do not offer staffed chat in those hours.
  2. Do most of your web enquiries arrive out of hours? A bot or a well-designed form with an instant acknowledgement may serve them better.
  3. Is your property data accurate and current on the website? A bot answering from stale data will create complaints.
  4. Do you have someone to review chat transcripts weekly? Without review, small errors repeat for months.
  5. Can chats flow into your CRM with the property and the visitor's details? If not, leads will be lost between systems.

Scoping a chatbot safely

A good agency chatbot does a small number of things well and hands over everything else. Keep the scope tight and written down.

Chatbot scope: what to automate and what to hand over
Let the bot handleHand to a person
Availability and key facts from approved listing detailsPrice negotiation, offers and "what will the seller take?"
Booking a valuation request or viewing requestValuation opinions
Office hours, branch addresses, how to report a repairHolding deposits, referencing results, disputes
Capturing name, contact details and positionAnything from a distressed or vulnerable visitor
Explaining your fees page and linking to itComplaints

For lettings, the bot must give the same information to every applicant and quote only the advertised rent. Bidding rules and fair treatment requirements apply to written messages as well as calls.

Handover rules that make hybrid work

  1. Detect the need

    Trigger handover on keywords such as offer, complaint, deposit or urgent, and whenever the visitor asks for a person.

  2. Say what happens next

    "I'll pass this to the lettings team. They're online until 6pm and usually reply within five minutes."

  3. Out of hours, be honest

    "The team is back at 9am. Leave your number and they will call you before 10am." Then make sure they do.

  4. Pass the context

    The person taking over should see the whole transcript and the property, so the visitor never repeats themselves.

Writing the bot's answers

A chatbot is only as good as the information behind it. Before switching one on, write the answers to the 30 or so questions your branch hears most, in your own voice, and check them with the people who handle those enquiries. Keep each answer short, give a next step and avoid anything that sounds like a promise you cannot keep.

  • Listing facts: tenure, council tax band, service charge, EPC rating, parking. Only from details the seller or landlord has approved.
  • Process questions: how viewings work, what referencing involves, how offers are handled.
  • Branch information: hours, addresses, how to reach the out-of-hours line.
  • Fees: link to your published fees page rather than paraphrasing it.

Then decide who owns updates. When a property goes under offer or a rent changes, the bot needs to know the same day. Stale answers are the most common reason website chat causes complaints.

Chat and lettings fairness

Lettings chats need extra care. Since 1 May 2026, written adverts in England must state a specific rent, and agents must not invite or encourage offers above it[6]. Chat replies count as written communication, so a bot or a negotiator must never suggest that "offers are being considered" on a rental, or hint that a higher rent would help. Equally, every applicant who asks about the same property should receive the same information about availability and the next step. Build those rules into the bot's answers and into the guidance for anyone staffing the chat window, and spot-check rental chats in your weekly transcript review. Applicants increasingly screenshot conversations, so assume any chat could be shown to a council or the redress scheme.

Staffing a chat rota

If you choose staffed or hybrid chat, treat it like the phones. Give one person the chat window for set hours, with a named backup for lunch and viewings. Show the chat as offline when nobody is watching it, rather than leaving visitors typing into silence. Decide how chat ranks against a ringing phone, because one person cannot do both well at the same time.

Privacy and transcripts

Chat transcripts are personal data. Mention website chat in your privacy notice, say whether a bot or a person is answering, and set a retention period for transcripts. Do not let visitors upload ID documents or bank statements through chat, and train staff to move sensitive conversations to a secure channel.

Example: the workload sum

Example (fictional): a branch gets about 90 web enquiries a month, half of them after 6pm or at weekends. Staffed chat in office hours would cover only the daytime half and need a negotiator watching the screen. The branch instead uses a bot to capture details and answer listing facts around the clock, with handover to a person between 9am and 6pm. Out-of-hours visitors get a callback promise for the next morning. The negotiators deal with qualified enquiries rather than "are you open Saturday?"

Run your own sum before choosing: monthly web enquiries, the share out of hours, and the staff time each model would need. The answer is often a hybrid.

Measuring whether chat converts

  • Chats started, and the share that leave contact details.
  • Time to first human reply in staffed hours.
  • Chats that became a booked valuation or viewing within seven days.
  • Wrong or unhelpful bot answers found in weekly transcript review.
  • Visitor drop-off at handover.

Compare those numbers with your form and phone enquiries over the same months. Large language model referrals are still tiny for portals, less than 0.5% of Rightmove's traffic in the first half of 2026[7], so judge chat on the leads it turns into appointments, not on novelty. If chat is not adding booked appointments after three months, a better enquiry form and fast follow-up of online valuation leads may be the better investment.

Frequently asked questions

Is live chat worth it for a small agency?

Only if someone can answer quickly in the hours it shows as available. Otherwise a bot that captures details, or a good form with an instant acknowledgement, usually performs better.

Can a chatbot give a valuation?

It should not give a price opinion. It can take the address and details and book a valuation with a person.

Do visitors trust chatbots?

Many use them, 44% of adults at least monthly in 2024[1], but most worry about decisions made without human input[2]. Make it easy to reach a person.

What should happen to chats out of hours?

Capture contact details and promise a specific callback time, then keep that promise first thing the next morning.

References

  1. Department for Science, Innovation and Technology (GOV.UK) (2024). Public attitudes to data and AI: Tracker survey (Wave 4) report (opens in a new tab) ↑

  2. Department for Science, Innovation and Technology (GOV.UK) (2024). Public attitudes to data and AI: Tracker survey (Wave 4) report (opens in a new tab) ↑

  3. Rightmove plc (2026). Rightmove plc unaudited results for the six months ended 30 June 2026 (opens in a new tab) ↑

  4. Rightmove plc (2026). Rightmove plc unaudited results for the six months ended 30 June 2026 (opens in a new tab) ↑

  5. Rightmove plc (2026). Rightmove plc unaudited results for the six months ended 30 June 2026 (opens in a new tab) ↑

  6. GOV.UK (2026). Assured periodic tenancies: rental bidding (opens in a new tab) ↑

  7. Rightmove plc (2026). Rightmove plc unaudited results for the six months ended 30 June 2026 (opens in a new tab) ↑

Book a demo

See Ava handle your calls.

Book a short demo. We’ll walk through your calls and workflow, and agree what Ava should take care of.

Book a demo