Consent revocation: how customers can opt out, and how fast you must honour it
A customer can withdraw consent by any reasonable means, which is more ways than a keyword list covers. What the confirmation text may and may not say, the five places an opt-out has to reach, and why one opt-out must cover the whole business. Not legal advice.
Sohail Safdar · · 6 min read
Most businesses treat opting out as a keyword problem. It is not: a keyword list is the automated part of a mechanism that also has to work when somebody says it in words, says it to a person, or says it in an email.
The governing idea in the FCC's rules is any reasonable means. The customer does not have to use your preferred word, and you cannot require one.
This is a summary in plain terms rather than legal advice, and parts of these rules have had their effective dates moved, so check the current position for anything you are relying on.
The ways an opt-out actually arrives
The keyword. STOP, and also STOPALL, UNSUBSCRIBE, CANCEL, END and QUIT, which carriers treat as opt-outs whether or not your system was watching for that particular word.
A sentence. "Please stop texting me." "Take me off your list." "No more texts." None of these contains a keyword and all of them are unambiguous to a human reader.
To a person. At the desk, on the phone, in passing. This is the one with no technical component at all, and the one most often lost: whoever heard it has to record it.
In writing elsewhere. An email, a reply to a letter, a note on a form.
In another language. A clear request to stop is a request to stop whatever language it is in.
The practical consequence: a keyword handler covers the first category and nothing else. The other four need a human process, and the business that has only the handler has an opt-out mechanism that works for the customers who happened to guess the magic word.
How fast
The rules set an outer limit in business days. Treat that as irrelevant to you.
The right target is immediate: the next message that would have gone to that number does not go. An outer limit measured in days exists for organisations running overnight batches against separate systems, and a small business on a modern platform has no reason to use any of it. If a reminder goes out the morning after somebody said stop, the fact that you were inside a regulatory window will not be what the customer remembers.
The confirmation text, and what it must not be
One confirmation is permitted, sent promptly, confirming the opt-out.
"You are unsubscribed from Northside Dental texts. Reply START to resubscribe."
That is the whole thing. What it must not carry:
- An offer or a discount.
- A reason to stay.
- A "tell us why you are leaving" survey.
- Anything about anything else.
The logic is simple once stated: at the moment they opted out, you stopped having consent to market to them. A win-back attempt inside the confirmation is marketing sent to somebody who has just told you to stop, which is the single worst message you can send in this whole area.
One opt-out, across the whole business
Record it against the phone number, not against a conversation, a campaign, a list, or a channel.
The failure mode is specific and common: a customer opts out of a marketing sequence, the opt-out is stored against that sequence, and three weeks later a different automation reaches them because it reads a different list. From the customer's side they said stop and you carried on. The internal distinction between your two lists is not a thing they can see or care about.
Narrow readings of how far a revocation reaches have been the subject of delayed rulemaking, which is a reason to be conservative rather than clever.
Scope, when the customer limited it
Some customers do say something narrower: "stop the review requests but keep the appointment reminders". That is a real preference and worth honouring exactly.
Two rules for it. Record the scope they actually stated, rather than reclassifying it into your own categories. And when it is ambiguous, ask, in the one message you are allowed, or simply stop everything, which is the safe direction and costs you a customer's reminders rather than a complaint.
The five places it has to reach
- The automated keyword handler, with no AI anywhere near it. This is a fixed rule, not a judgement, and a model that can be talked out of things should not be between a customer and an opt-out.
- Plain-language opt-outs, recognised automatically where possible and flagged to a person the same day where not.
- Front desk and phone requests, recorded on the customer's record by whoever received them. Decide who and how before you need it.
- Every send path, because an opt-out that only one of your three message sources reads is an opt-out that works a third of the time.
- A record of when and how it arrived, which is the thing you will want if anybody ever asks.
Re-subscribing
A customer who texts START, or who asks at the desk to start getting reminders again, can be re-enabled. Record it as a new consent, with its date and the wording they agreed to, rather than as the cancellation of the old opt-out.
The distinction matters if it is ever examined: a new consent is evidence of a decision they made, where a removed flag is evidence of something your system did.
In Frontiva
A STOP is recorded against the phone number, so it covers every conversation with that person including ones opened later, which is point five of the previous section made structural rather than procedural. START lifts it and HELP is answered with the carrier-required disclosure, and the AI is deliberately kept out of all three: they are fixed handlers, because an opt-out is not a thing to exercise judgement about.
Every outbound message then passes one send check that reads that state, and the check stores its outcome on the message. That is the part worth having for this specific subject: "why did you text me?" is answerable with what was true at the time rather than with a re-derivation from today's data, and since consent can have been withdrawn in between, a re-derivation would give the wrong answer with complete confidence.
The check distinguishes its refusals by reason rather than lumping them together, and two of those distinctions exist for this topic. A refusal for do not contact is different from a refusal for no consent for this purpose, because they send a practice to do different things. And a platform-level suppression, where an address has been suppressed across every business for a hard bounce or an abuse report, carries its own reason rather than reading as "they replied STOP", because a practice told the second thing would go looking for a STOP that never happened.
STOP, HELP and the texts you must answer covers the three keywords carriers require, keeping a consent record covers what to store, and the unified inbox is where a plain-language opt-out lands for a person to action.