How much does an AI receptionist cost? The verified numbers

That spread is real, not a rounding error. The $49 product and the $2,497 product both answer a ringing phone, and past that they have almost nothing in common. What you are actually shopping for is how far the call gets carried after hello.

Every figure in this post was read off a public pricing page in July 2026, and each page is linked in the sources at the bottom. Vendors reprice constantly, so treat these as a snapshot with a date on it. Any vendor quoting you a competitor's price without a URL is negotiating, not informing.

One vendor is missing on purpose: anyone whose page says "book a demo" where the number should be. Unpublished pricing is not a red flag by itself. It just means the price is whatever the salesperson thinks you will pay, and you should shop accordingly.

Per-minute is the answering-service tradition carried over. Ruby sells 50 human minutes for $250 and 500 for $1,725, which works out to $3.45 to $5.00 a minute. A qualified trade call — name, address, symptom, window — runs four to six minutes, so a per-minute plan is really a per-call plan with the math hidden.

Per-call is more honest about what you are buying. Smith.ai's AI receptionist runs $1.67 to $3.00 per answered call depending on volume, and its human plans run roughly $7 to $10 a call before add-ons. Spam still counts. A robodialer that gets answered is a call you bought.

Flat monthly puts the surge risk on the vendor instead of you. Goodcall charges $79 to $249 with unlimited minutes, capped by unique customers per month. We price the same way: one monthly number with a volume ceiling agreed in writing before you start, and a conversation — not an invoice — if you outgrow it.

Read the third and fourth columns together, not the third alone. Rosie at $49 and Goodcall at $79 are genuinely cheap if all you need is a voice that picks up after hours. Smith.ai's human tier at $2,100 is not overpriced; it is a different product with people in it.

We put ourselves in the table because leaving your own row out of a price comparison is the oldest trick in content marketing. We are the most expensive AI row and the cheapest way to get the whole loop run for you. Both of those facts should survive contact with our own blog.

• Feature gating. The advertised price is usually the message-taking tier. The thing you actually wanted — a job on your calendar without a callback — lives one or two tiers up. Price the tier that books, not the tier on the billboard.
• Overage. Every metered plan has a rate for the calls past the bucket: $0.50 a customer at Goodcall, $2.17 to $2.50 a call on Smith.ai's AI plans, $8.50 to $11.50 a call on its human ones. Your surge month is priced at the overage rate, not the plan rate.
• Per-call add-ons. On Smith.ai's human plans, booking, Spanish, and text follow-up are line items on every call that uses them. A booked bilingual call with a follow-up text carries $3.00 of add-ons on top of the bundle rate.
• Annual lock-in. Goodcall takes 15% off for annual; Rosie offers two months free. Real discounts, but you are prepaying a year for software you have run for zero. Buy monthly until the thing has survived your busy season.

The good news: setup fees have largely died in this category. Smith.ai and Ruby both state there are none, and ours is waived right now. If a vendor still quotes one, ask what specifically it buys, in hours of whose time.

Here is the honest version of why our floor is ten times Rosie's. A $49 tool hands you a settings screen: connect the calendar, write the qualifying questions, encode your slot lengths and buffers, test it, tune it. That work is real, it is why the tool is $49, and most owners never finish it. The tool then answers beautifully and books nothing.

What the three tiers actually buy: $497 is our line, scripts, and templates with you doing the wiring. $997 is us building it — script written for your shop, calendar connected, a text to you on every booked job, review requests after every completed one, and a recovered-revenue number you can check. $2,497 adds monthly script tuning against real recordings, warm transfer rules, and a named person.

If someone in your shop genuinely likes setting up software, a self-serve tool plus a free weekend gets you most of the way there, and you should try that first. The tiers exist for the owner whose free weekend is in November.

Do not use a vendor's ROI calculator with the vendor's default numbers in it, including ours. The defaults are set to make the product look inevitable. The calculator at /roi is only worth your time because you can type your own numbers into it — and if yours say no, believe yours.

1. Median invoice: pull your last twenty paid invoices and take the middle one, not the average. One big job drags an average somewhere useless.
2. Missed calls: export last month from your carrier portal, count unanswered inbound, delete repeat dials within ten minutes and obvious spam. Count, don't estimate.
3. Close rate: use the fraction of answered calls you actually book today. If you book half of what you answer, recovered calls will not do better.
4. Multiply the three. That is the monthly revenue on the table. Compare it to the monthly price of whatever you are considering — including the overage-month price, not the brochure price.

A worked example with round numbers, not a claim about results: a shop with a $400 median ticket, 15 genuinely missed calls a month, and a 40% close rate is leaving about $2,400 on the table. Against that, $49 pays for itself on the first job and $997 on the third. A shop missing four calls a month at a $250 ticket should buy the $49 tool or nothing.

Run the count before you talk to any vendor, us included. If your number is small, close all six tabs and go to bed. If it is big, now you know what you are negotiating with.