Status: planned, not started

How we’ll test AI receptionists.

No test calls have been placed and no scores exist yet. This page describes the method we intend to use, and any change to it will be dated.

What we call

  • Published demo lines only. We only call numbers a vendor publicly labels as a demo of its AI receptionist. We never call a vendor’s real customers, sales lines or support lines.
  • 3 scripted calls per vendor, every month, using the same scripts for every vendor in that round.
  • Vendors without a published demo line are listed but not scored, and their pages say so.

The three calls

The draft script previewed on the home page. The exact wording will be published with each round.

  1. Call 1 · Booking

    A homeowner with a leaking sink asks for a Thursday-morning visit and gives a name and callback number.

    Scored on: Answer speed, booking success, accuracy

  2. Call 2 · After-hours emergency

    Late at night, water is coming through the ceiling. What should I do?

    Scored on: Escalation, safety advice, no over-promising

  3. Call 3 · Price question

    Roughly how much does a call-out cost?

    Scored on: An honest answer, no invented quote

How the calls are made

  • Calls are placed by an automated caller. At the start of each call it says that it’s an automated test call and that the call is recorded.
  • Each call is recorded with each side on its own channel, transcribed, and the transcript is scored against the rubric below.
  • The scripts play a typical small-business caller (for example, someone who wants to book a plumber). Results reflect the vendor’s own demo setup, not how it would be configured for your business.

Scoring rubric

CriterionWhat we look at
Answer speedTime from the first ring to the first spoken greeting.
Booking successWhether the caller can book (or request) an appointment and gets a clear confirmation.
AccuracyWhether the details it captures and repeats back (name, number, time, reason) are correct, and nothing is made up.
After-hours emergencyHow it handles an urgent out-of-hours problem: escalation, safety advice, not over-promising.
Price questionHow it answers “how much will this cost?”: honestly, on-script, with no invented quotes.

Weights and exact scoring scales will be published here before the first results go live.

Vendors may respond

Every vendor can send a correction or a response to its results, and we’ll publish it alongside them. Submit a response or correction.

Prices

Starting prices are copied word-for-word from each vendor’s site with the date checked, and re-checked weekly. Changes go in the price change log. A blank price means we couldn’t find one published.

Independence

We may earn affiliate commissions in future. Commissions never affect scores, and vendors can’t pay to be tested, ranked or removed. See our disclosure.