What 568 AI-Booked Clinic Appointments Taught Us

Between 1 June and 31 October 2026, our AI receptionist booked 568 appointments for a five-location clinic group in Queensland and Tasmania. It replied to patients in a median of 52 seconds, 86.6% of the appointments that reached their date were attended, and more than half of the bookings were made outside business hours. This is the full breakdown, how we measured it, and what we would tell any clinic owner thinking about an AI receptionist.

The client is a bookings-led health clinic group that runs Facebook ads to generate enquiries. The AI works over text, SMS and chat, not phone calls: it replies to each enquiry, answers common questions, books a free consultation into the right location's calendar, and follows up the patients who go quiet. Client details are anonymised. You can read the story behind it in the clinic group case study.

The Headline Numbers

• 568 appointments booked by the AI, for 559 individual patients, across 5 locations.
• 318 attended, 49 were no-shows, 185 were cancelled and 16 were still open or invalid when we froze the data.
• 86.6% show rate: 318 of the 367 appointments that reached their date. Cancellations are left out of this rate because they never reached their date.
• 72 patients went on to buy, worth $262,760 at an average of $3,649. That is about one in five attended appointments.

How Fast Did the AI Reply?

Across 6,671 replies to patient messages, the AI's median reply time was 52 seconds. 75% of replies went out within a minute and 93% within five minutes.

Speed matters because interest fades fast. In the well-known MIT and InsideSales lead response study, leads contacted within 5 minutes were 21 times more likely to qualify than leads contacted after 30. A front desk juggling phones, walk-ins and patients can struggle to reply within five minutes to every enquiry. Software does not get busy.

How Quickly Did Enquiries Turn Into Bookings?

From the moment a patient's enquiry arrived to the moment they were booked, the median time was 3.8 hours. 37% were booked within an hour and 80% within 24 hours, measured across 506 patients.

Not everyone books on the spot. Some people reply in the evening, some the next morning. The AI keeps the conversation open and picks up where it left off, so the booking happens when the patient is ready rather than when the clinic happens to call back.

When Were the Bookings Made?

307 of the 568 bookings, 54%, were made outside 9am to 5pm on weekdays. Those are enquiries that would otherwise have waited in an inbox until someone was back at the desk. Australian data shows how common that gap is: an AIIMS Group analysis of more than 207,000 calls to 858 Australian businesses found around 1 in 10 inbound calls was dropped.

Where the Leak Actually Was: Cancellations

The biggest loss in this data was not no-shows. It was cancellations. 185 of the 568 bookings, a third, were cancelled before their date. Once a patient reached their appointment date, they turned up 86.6% of the time.

That changes where the next improvement comes from. Reminders and easy rescheduling matter more than chasing no-shows, because a cancelled booking that gets rebooked is a patient kept. A BMJ Open review of 21 studies found text reminders cut no-shows from 21% to 15%, and the same logic applies to keeping cancelled patients in the calendar.

What We Would Tell a Clinic Owner

• Speed wins the booking. A reply inside a minute keeps the patient in the conversation while they are still interested.
• Half your demand arrives after hours. If nobody replies until morning, someone else may have already booked them.
• Watch cancellations, not just no-shows. That is where a third of these bookings went.
• Measure against your own records. Every figure here comes from the clinic group's booking and sales data, not from the AI's own reporting.

How We Measured This

• Source: the clinic group's own booking and sales records, mirrored into our reporting system, frozen as at 10 November 2026.
• AI booking: an appointment created through the AI's booking connection, with no staff member involved. Staff and self-service bookings are excluded.
• Window: appointments booked 1 June to 31 October 2026, Melbourne time. The AI's first booking in this record was on 3 June. Bookings from 3 to 9 June were made before our message tracking began, so they are identified by the same booking signature (created through the AI's booking connection, no staff member) as every later AI booking.
• Show rate: attended appointments divided by attended plus no-show appointments. Cancelled and still-open appointments are excluded.
• Reply time: the gap between each patient text and the AI's next reply, within 24 hours.
• Enquiry to booking: from the time a patient's enquiry was first recorded to their first AI booking, for patients first seen from 15 June.
• Sales: purchases marked by the clinic's team for patients whose first booking was made by the AI. Sales are marked by hand, most recently on 5 October 2026, so October is likely under-counted.

Limits: this is one client in one industry, analysed by us. We do not publish a before-and-after comparison, because the clinic's booking history before our system was connected is incomplete, and a comparison built on it would be misleading.

Frequently Asked Questions

How fast does an AI receptionist reply to new enquiries?

In this clinic group, the AI replied to patient messages in a median of 52 seconds across 6,671 replies, with 75% of replies sent within a minute and 93% within five minutes (1 June to 31 October 2026).

Do patients turn up to appointments booked by an AI?

Yes. 86.6% of AI-booked appointments that reached their date were attended (318 of 367), between 1 June and 31 October 2026. A third of bookings were cancelled beforehand, which is where most of the loss sat.

How many bookings happen outside business hours?

307 of 568 AI bookings (54%) were made outside 9am to 5pm on weekdays: 187 on weeknights and 120 on weekends.

Is this data independent?

No. It is one client, analysed by Stacked Digital, using the clinic group's own booking and sales records frozen as at 10 November 2026. We publish the method and its limits so you can judge it for yourself.