SMS appointment confirmations and the TCPA
Texting customers about their appointments is one of the highest-return things a service business can do, and one of the easiest places to create a liability by accident. The difference is consent, and specifically what kind of consent you can prove afterwards.
This is not legal advice. The TCPA is federal, several states add their own rules on top, and the penalties are per message rather than per customer. Talk to a lawyer about your own situation.
Why is a verbal yes on the call not enough?
Because the standard that matters is written, and because a year later you have to be able to evidence it.
The TCPA distinguishes between messages by their purpose. A confirmation for an appointment the customer just booked is informational. A message telling them about a spring service offer is marketing, and marketing to a mobile number requires prior express written consent. The trouble is that the line between the two is thinner than owners expect — a confirmation with a discount code stapled to the bottom has quietly become a marketing message.
The practical problem is evidential rather than philosophical. If a complaint lands, “they said yes on the phone” is a claim. A ticked box with a timestamp, the wording that was on screen, and the number it was tied to is a record. One of those survives contact with a regulator.
What Ansdale captures
The booking page takes SMS consent as an explicit opt-in rather than a pre-ticked box or a buried line in terms. What is stored alongside it is the part that matters if anyone ever asks: when it was given, and what the customer was agreeing to at the time.
| Recorded | Why it matters later |
|---|---|
| Timestamp of the opt-in | Establishes consent existed before the message was sent |
| The number consented for | Consent attaches to a number, not a person |
| The wording shown at the time | Shows what was actually agreed to, not what you say now |
| Opt-out, when it happens | The moment consent ends is as important as when it started |
What about opting out?
STOP has to work, immediately and without a human in the loop. This is the failure that catches otherwise careful businesses: consent was collected properly, the messages were reasonable, and then someone replied STOP and got another text the following week because the opt-out lived in one system and the sending list lived in another.
Treat an opt-out as permanent and immediate. If the customer books again later and wants texts again, they can opt in again — that is a new consent record, which is exactly what you want.
Does an AI receptionist change any of this?
It changes where the consent is collected, not what the standard is. A caller booking by phone has not filled in a form, so the written opt-in has to come from somewhere — the booking confirmation page, a reply the customer sends, or an explicit step you configure.
What it should never do is quietly treat “yes, text me” on a recorded call as equivalent to a written opt-in for marketing. It is not, and a call recording is a considerably more awkward thing to produce in a dispute than a database row.
A short checklist
Before you send anything: is this message informational or marketing, and would a stranger reading it agree with your answer? Do you hold a timestamped opt-in for this specific number? Does STOP work in every system that can send? Can you produce the consent record for a customer from eighteen months ago?
If any answer is uncomfortable, fix the record-keeping before you increase the sending.
What should I read next?
The consent page covers the other half — call recording and telling callers they are speaking with an AI. Our own SMS consent terms set out what we do with numbers collected through Ansdale.
Call recording and AI disclosure · our SMS consent policy · no-show protection.