# How AI receptionist vendors define a booking

> AI receptionist and voice-agent vendors bill per call, minute, caller, location or contract, and apart from ServiceTitan none of the pages read defines a booking. Marketplaces that bill per booking draw the line later.

Published 2026-11-12 · Reviewed 2026-11-12. Canonical: https://attify.app/compare/ai-receptionist-booking-definitions

Apart from ServiceTitan's booking rate, none of the AI receptionist and voice-agent vendor pages read for Attify's research defines what counts as a booking. Those that state a billing unit bill for activity (per call, per minute, per unique caller, per location or by annual contract), so the booking counts they report are not invoiced. Marketplaces that do bill per outcome draw the line later or narrower: seated diners, visits that took place, or bookings by patients new to the provider.

Facts below come from each company's own pages, read on 31 October 2026 (OpenTable's on 9 November) for the [research](https://attify.app/research/bookings) behind the [Attify Audited Booking Standard](https://attify.app/standards/bookings). This page describes rules, not results.

## How AI receptionist and voice platforms bill

| Billing unit | Vendors (from their own pricing pages) |
|---|---|
| Per call | Smith.ai; ServiceTitan's AI voice agents |
| Per minute | Ruby, Sameday, Retell AI, Vapi, Bland |
| Per unique caller | Goodcall |
| Per location | Slang.ai |
| Annual contract | Synthflow |
| Not published | Arini, Weave, NexHealth, Assort Health, Hello Patient, Avoca |

Details, prices and links for each are in the [research](https://attify.app/research/bookings), dated. Prices change.

## How booking is counted

- **Apart from ServiceTitan, none of the vendor pages read defines a booking.** Some show an "appointments booked" counter or a booking rate without a formula.
- **ServiceTitan** documents its call booking rate: booked calls over lead calls (60 seconds or longer, or tagged as a lead), with reschedules and confirmations "excused" ([ServiceTitan](https://help.servicetitan.com/v1/docs/understand-call-classifications)).
- **Voice platforms' success flags** (Retell's `call_successful`, Vapi's `successEvaluation`) are an LLM judging the transcript against a prompt the customer writes.
- **Request versus booking.** Where an AI cannot write to the practice's system, some products capture a request and staff make the booking. A count that includes both credits the AI with staff work.
- **Podium's API** marks whether its AI scheduled an appointment (`aiGenerated`), the only explicit AI-booked flag found.

## Where per-booking marketplaces draw the line

| Marketplace | Billed when | Source |
|---|---|---|
| OpenTable | Seated diners only; no-shows and cancellations not billed if marked within 48 hours | [OpenTable](https://www.opentable.com/restaurant-solutions/plans/basic/) |
| Zocdoc | At booking, only for patients new to the provider; fees may be waived for some cancellations | [Zocdoc, 17 Dec 2025](https://www.zocdoc.com/blog/pay-per-booking-fees-explained/) |
| Booksy Boost, Fresha | A new client's visit that took place; no-shows marked the same day | [research](https://attify.app/research/bookings) |

These are the exclusions a careful vendor already accepts, and the reason Attify's bookings standard checks what happened after the booking: against the claim of a vendor that bills kept appointments, beside the claim of one that counts bookings ([the claimed unit](https://attify.app/standards/bookings#the-claimed-unit)).

## Against Attify's checks

[Standard #3](https://attify.app/standards/bookings) tests the vendor's own count where one exists, in the unit the vendor claims ([the claimed unit](https://attify.app/standards/bookings#the-claimed-unit)), then checks in the business's own booking system that the AI created it (not staff from a request), it is a new booking and staff did not fix it within 48 hours, and has a judge confirm it matches what was agreed. For a claim of kept appointments it also checks that the appointment was not cancelled and the customer came; for a claim of bookings, cancellations and no-shows are reported beside the claim. No vendor definition read includes any of these.

## Vendors

A vendor described here can send corrections with a link to its own documentation under the [right of reply](https://attify.app/right-of-reply).

## Questions

### Does any AI receptionist bill per booked appointment?

None of the 19 vendors whose own pages Attify's research read says it does; those that state a billing unit bill per call, minute, unique caller, location or annual contract. A third-party comparison site describes one more vendor as billing per booked job, which could not be confirmed on its own site.

### What does a voice platform's 'call successful' mean?

On Retell and Vapi it is an LLM grading the call transcript against criteria the customer or vendor writes. It is the vendor's own verdict, not a booking record.

### Who does define a billable booking?

Marketplaces that bill per outcome: OpenTable bills seated diners, Booksy and Fresha bill visits that took place, Zocdoc bills bookings by patients new to the provider.

## Sources

- Attify: [Research: how AI receptionists count and bill a booking](https://attify.app/research/bookings), read 2026-11-09
- ServiceTitan: [Understand call classifications](https://help.servicetitan.com/v1/docs/understand-call-classifications), read 2026-10-31
- Zocdoc: [Pay per booking fees explained](https://www.zocdoc.com/blog/pay-per-booking-fees-explained/), read 2026-10-31
- OpenTable: [Basic Plan](https://www.opentable.com/restaurant-solutions/plans/basic/), read 2026-11-09
