There are free AI receptionists, and it is worth knowing what each kind of free actually is before you build a month of habits on one.
Three things get called free in this category, and only one of them is what you want.
The three kinds of free
A time-limited trial. Full product, real functionality, and it stops. This is the useful one — if the window is long enough to cover a normal month rather than a good week. Ours runs thirty days and takes no card, which is deliberate: a two-week trial straddling a quiet fortnight tells you nothing.
A usage cap. Free up to some number of calls or minutes a month. Read the number carefully. A cap that sounds generous in the abstract — say fifty calls — is a fortnight for a moderately busy salon, and the pricing past it is often the point of the free tier rather than an afterthought.
A permanently free tier that cannot book. The one to watch. It answers, it converses, it sounds impressive, and it is not connected to your calendar. What it produces is a message or a callback list. It looks fully functional right up until you notice nothing is being booked, which can take weeks because the calls genuinely are being answered.
That third kind is not dishonest exactly — the capability really is there — but it is the version of the product without the part that makes it worth having.
The question that tells you which you are looking at
On the free tier specifically, can it write into my calendar?
Not "does the product integrate with calendars" — does this tier do it. The answer is frequently no, and it is rarely on the pricing page.
Ask it in writing before you spend any time on setup, because configuring a system properly is a couple of hours and discovering the limitation afterwards means doing it twice.
What free generally does not include
The integration. Covered above, and the main one.
Text messages. Reminders and confirmations billed per message, or absent entirely. For a business that reminds every booking this is not a rounding error.
Setup help. You are on your own with your service list, durations, staff and hours. That is the fiddly part and it is where a bad configuration produces bad bookings for months.
More than one number or location.
Any real support. Fine until the day the phone does something strange and you need someone.
Using a trial properly
If you take a trial, the mistake is to spend it on the vendor's demo line rather than on your own callers.
Point real calls at it. Conditional call forwarding hands over only the calls you do not answer, so you can test on genuine callers without risking anything that currently works — your number and your existing setup are untouched. The codes by carrier are in the call forwarding guide. Testing on your own traffic is the only test that means anything.
Configure it properly on day one. Real services, real durations, real staff qualifications. A trial run on a half-finished service list measures your configuration, not the product.
Try to break it. Ask for a vague time. Interrupt it. Ask it something it cannot know. If it improvises an answer about a price or an allergen during a trial, it will do it to a customer later. The five tests are in five calls to make before you trust an AI receptionist.
Read the transcripts. Not a sample — all of them, at least for the first week. This is the single most informative thing you can do, because you will find out what your callers actually ask, which is usually not what you assumed.
Count the bookings. At the end, the number that matters is how many appointments came from calls that would previously have rung out. Everything else is impression.
The honest arithmetic on free
Free is the right starting point and a poor destination, for a reason that has nothing to do with vendors.
The cost of a receptionist that does not book is not the price. It is the bookings that did not happen. If six callers a month would have booked and instead got a message taken, at your average ticket, that is what the free tier cost you — and it is almost always larger than the paid tier it was avoiding.
So run the trial, on your own calls, configured properly. Then count. If the bookings are not there, the free version has told you something genuinely useful and you have lost nothing. If they are, you now have a real number to weigh against a real price, which is a much better position than comparing feature lists.