Lost bookings rarely announce themselves. There is no error message and no complaint. A visitor opens the chat, something about the exchange feels like hard work or slightly wrong, and they leave. Your analytics show a conversation. Your diary shows nothing.
The same handful of patterns turn up again and again when businesses review their first months of transcripts. None of them are exotic. Most come from reasonable decisions made during setup that turned out differently in front of real customers.
Here are seven AI receptionist mistakes that cost bookings, grouped by when in the conversation they happen, each with the fix.
Before the visitor has had an answer
1. Asking for contact details first
The receptionist opens with "Before we start, can I take your name and email?" It feels efficient from the business side: you get the lead even if the visitor wanders off. From the visitor's side, it is a form wearing a chat costume, and they have not yet learned whether you can help.
The fix: answer first, ask later. Let the receptionist handle the question, then offer to take details at the point where doing so helps the visitor, such as requesting an appointment or getting a quote. Asking for contact details politely covers the timing.
2. Too many questions too early
A cousin of the first mistake. A visitor asks "do you have anything Thursday?" and gets five questions back: service type, stylist preference, new or existing client, how they heard about you, phone number. Each one is reasonable. Together they feel like an application.
The fix: ask only what the next step needs, one or two questions at a time, and explain why when it is not obvious. Save marketing questions for later or drop them.
3. Greeting the wrong people, the wrong way
A proactive greeting that pops up two seconds into every page, including blog posts and the careers page, trains visitors to close the chat before reading it. Meanwhile the pricing page, where people actually have questions, shows the same generic "How can I help?"
The fix: use proactive greetings on high-intent pages with a short delay, and write page-specific openers for your pricing, contact and main service pages.
In what it knows
4. Stale hours, prices and services
A very common cause of a lost booking is also the least interesting: the receptionist is working from out-of-date facts. Christmas hours from last year. A price that went up in March. A treatment the business stopped offering. The receptionist is doing exactly what it should, answering from its content, and the content is wrong.
The fix: tie updates to the moment things change. When you change a price on the till, change it on the site and in the receptionist the same day. Re-crawl after website edits. Keep holiday hours in a Q&A pair you can edit in a minute. Keeping your chatbot's knowledge up to date sets out a routine.
At the moment of booking
5. Implying the booking is confirmed
This one does real damage. A website receptionist that captures booking requests is doing a useful job, but if its wording suggests the slot is secured, customers turn up to appointments that do not exist.
Visitor: Can I book a massage Saturday at 11?
Assistant (mistake): Perfect, you're booked in for Saturday at 11am. See you then!
Assistant (fixed): I can send a request for Saturday at 11am to the team. They'll check availability and confirm by text, usually within a few hours during opening times. Could I take your name and mobile number?
The fix: make the confirmation step explicit in every booking reply, with a realistic time frame. If you are weighing up request-based booking against a direct booking tool, booking widgets vs conversational requests compares the two honestly.
6. A dead end with no person behind it
The visitor has a question the receptionist cannot answer, or wants to talk to someone, and the conversation simply stops: "I'm sorry, I can't help with that." No details taken, no ticket, no handover.
The fix: every conversation the receptionist cannot finish should end in one of three places: a captured enquiry, a support ticket or a live handover to a team member who is on duty. Decide which applies to which situation. When should a chatbot hand over to a human sets out clear triggers.
After the conversation ends
7. Nobody owns the captured requests
The receptionist did everything right. It answered, qualified, and captured a booking request at 7.20pm on Sunday. The request sits in an inbox until Wednesday, because everyone assumed someone else was checking it. By then the customer has booked elsewhere.
The fix: name one person responsible for captured requests each working day, and set a target such as replying before 10am. Send leads to wherever that person already works, whether that is email, a CRM or Slack via a webhook. Speed to lead explains why the first hours matter so much.
A fifteen-minute audit
Run through this with your own receptionist. Each check takes a minute or two.
| Check | How to test it | Warning sign |
|---|---|---|
| Details before answers | Open the chat and ask a simple question | Any request for name or email before an answer |
| Question load | Ask to book something | More than two questions in one message |
| Greeting placement | Visit a blog post, then the pricing page | Same pop-up and same wording on both |
| Fresh facts | Ask about hours next bank holiday and your newest price | Any outdated answer |
| Booking wording | Request a specific time | Words like booked, confirmed or see you then |
| Dead ends | Ask something it cannot know | No offer to take details or pass it on |
| Follow-up | Check when last week's requests got a reply | Anything waiting more than a working day |
Fix the two that cost most first
If you only have an hour, start with numbers 5 and 7. Misleading booking wording creates no-shows and angry customers, and unowned requests throw away enquiries the receptionist already won. Both are quick to fix: rewrite the booking reply to name the confirmation step, and put one person's name against the morning check.
Then run the audit table once a month to catch new AI receptionist mistakes early, and again whenever your hours, prices or services change. Read five real transcripts each time. The mistakes above are easy to spot in transcripts and nearly invisible anywhere else.