What is SMS opt-in? The practical difference from Express written consent

SMS opt-in

That distinction matters because a customer can give you a phone number for shipping, account security, or customer service without agreeing to promotional texts. A checkout field that says “Mobile number” is therefore not the same thing as an SMS marketing opt-in.

For US programs, the consent standard that matters depends on what is being sent, how it is sent, and which rule is being applied. The TCPA and FCC rules create legal requirements in defined situations. CTIA publishes industry messaging principles and best practices. Carriers and messaging providers can impose additional operational requirements. A safe implementation does not collapse these layers into one sentence such as “TCPA requires this exact checkbox for every text.”

The useful mental model: opt-in is a permission grant

Think of an opt-in as a record with at least 6 dimensions:

Who gave permission: the phone number and, when available, the customer/profile identity linked to it
To whom: the specific business or brand that will send the messages
For what channel: SMS or MMS, rather than “all marketing everywhere.”
For what purpose: for example promotions, back-in-stock alerts, order updates, or another clearly described program
Under what disclosure: the language the person saw when they consented
When and how: timestamp, source, form/keyword/import path, and other evidence that can reconstruct the event

The grant can later be narrowed, replaced, or revoked. The platform therefore needs history, not just a boolean such as sms_subscribed = true.

Figure 1

Collection creates a record. Every send rechecks the current state

Collect
1
Disclosure shown, then affirmative action
What the person saw and did
2
Consent evidence record
Source, timestamp, brand, purpose, disclosure version
Store
3
Current permission state
4
Campaign queued
Recheck
5
Send-time eligibility gate
Never rely on the queue-time snapshot
Outcome
SEND

STOP arrives, or the gate fails: suppress the send. A later new opt-in starts a new permission event; it never edits this one.

Consent requirements vary by message type, technology, jurisdiction and provider. This is an operational model, not a universal legal test.

The evidence from signup does not expire, but it also is not the last word. The gate right before send looks at the latest state, not the state when the campaign was queued.

Checkout checkbox

A shopper enters a phone number for delivery updates and separately chooses to receive marketing texts. The marketing permission should be tied to the separate affirmative action, not inferred from the existence of the number.

Store the wording and version of the disclosure shown at that checkout. If the storefront later changes its wording, old records should still be reconstructable.

Popup or embedded form

A form might offer “Get 10% off by text.” The opt-in event should retain the campaign/form identifier, disclosure version, phone number, timestamp, and source page. If the incentive changes later, that should not rewrite historical evidence.

Keyword signup

A customer texts a keyword to a business number. The inbound message is a strong event. The program still needs a clear response and a record of the keyword's meaning at that time. “JOIN” should map to a defined program. It is not an unlimited future license to message for unrelated purposes.

Back-in-stock or product alert

A shopper can request a one-purpose alert such as “Tell me when size M is back.” That request may be narrower than ongoing promotional consent. A system should not silently upgrade a one-time alert into a general marketing subscription.

Imported contacts

An uploaded CSV containing phone numbers is not proof of SMS permission. Import workflows should ask for the consent basis. For higher-risk sources, require evidence or keep the records non-sendable until reviewed. “They are customers” and “we have their numbers” do not answer the permission question.

Store evidence, not just state

A durable consent record can include:

normalized phone number
brand or legal sender identity
channel and purpose
consent status
consent timestamp and timezone
collection source and source URL where relevant
disclosure text or immutable disclosure version
form, campaign, keyword, or API event identifier
IP/user-agent or other evidence when lawfully and usefully collected
terms/privacy links that were presented
upstream platform/provider identifiers
revocation timestamp and method, if later revoked
subsequent resubscription event, if one occurs

Do not overwrite a revoked record with a new true. Append a new event. The history should make it possible to answer: what did this person authorize, under which disclosure, and what was the latest valid instruction before this send?

Consent at collection time is not enough

The send path needs a permission gate immediately before dispatch. A practical sequence is:

Resolve the intended recipient and sending brand
Determine the message purpose or class
Load the latest applicable SMS permission state
Check suppression and opt-out state
Apply quiet-hour and other timing rules
Apply provider/carrier eligibility and sender-route requirements
Send only if all required gates pass

This protects against a common race condition. A customer opts in on Monday. A campaign is scheduled on Tuesday. The customer opts out on Wednesday morning. The queued Wednesday afternoon message must not use Tuesday’s eligibility snapshot.

Scope is part of consent

One database can legitimately contain several permissions for one number. A customer might want order-status texts but not promotions, or a specific product alert but not recurring campaigns. Your data model should preserve that distinction where your program, law, or provider policy requires it.

The opposite mistake is fragmentation without control. If a customer clearly revokes promotional messaging, the business should not route around that instruction by switching sender numbers, creating another campaign object, or relabeling the same promotion. The operational system needs brand-level suppression logic where the applicable rules or platform policy expect it.

Resubscription should create a new event

A person who previously opted out can later opt back in. Treat the new opt-in as a new permission event with its own timestamp, source, and disclosure. Do not delete the earlier opt-out. The history is useful both for compliance and for debugging why a message was or was not sent.

Consent versioning and source-of-truth rules

Consent systems become unreliable when they store the current result but discard how that result was reached. Treat the event stream as the durable source of truth and the current subscribed/suppressed value as a projection that can be rebuilt. For example, a profile might have these events:

2026-06-03 10:12 - web form opt-in under disclosure v4
2026-07-20 14:08 - inbound STOP
2026-08-11 09:31 - new keyword JOIN under disclosure v6

The current state can be “subscribed,” but a support agent should still be able to see the July opt-out and the later event that changed eligibility. This also prevents a data import from silently overwriting a newer customer instruction.

When multiple systems can write preference state, define precedence explicitly. An inbound opt-out from the messaging channel should normally reach the central preference store immediately. A nightly ecommerce sync should not re-enable the number simply because the commerce platform still has an old marketing flag. Prefer monotonic event ingestion plus conflict rules over last-write-wins synchronization between unrelated timestamps.

Shared brands and multiple storefronts

A company may operate several storefronts, regional domains or sub-brands. Do not assume that consent collected by Store A automatically authorizes Store B merely because both accounts sit in one SaaS tenant.

Your permission model should identify the sender/brand the customer was shown. If a corporate group legitimately uses a shared consent program, that relationship should be explicit in the disclosure and data model rather than inferred from ownership after the fact.

This is especially important for multi-tenant platforms: merchant isolation must prevent one customer’s consent artifact from being used to justify another merchant’s messages.

What does this page teach beyond a generic glossary definition?

It turns “SMS opt-in” from a signup action into a permission lifecycle. Scope the grant, retain the evidence, preserve history, and re-evaluate the current state immediately before every send. That model prevents a phone number collected for one purpose from silently becoming permission for another.

> This page explains messaging operations and current US regulatory/provider concepts. It is not legal advice, and state law or program-specific rules can impose additional requirements.

Worked example

A skincare store collects a mobile number at checkout for delivery updates. The shopper leaves the promotional SMS checkbox unchecked.

After purchase, the store can retain the number for the operational purpose for which it was supplied, subject to the applicable rules. It should not mark the shopper as an SMS marketing subscriber simply because the same number exists in the customer profile

2 weeks later, the shopper enters a popup that says they can receive promotional texts from the store and affirmatively accepts. That is a separate event. The marketing platform can now store the new permission with the exact disclosure version and source campaign.

If the shopper later replies STOP, future promotional sends must be gated by the resulting opt-out state even if a campaign was already queued.

What SMS opt-in requires

The reader should be able to tell what they are signing up for before acting. For a promotional program, the disclosure should identify the business and make the marketing nature of the messages clear. Where prior express written consent is required by the FCC rule, the agreement must clearly authorize the seller to deliver or cause delivery of advertising or telemarketing messages using the covered technology to the specified number; the disclosure must also make clear that consent is not a condition of purchase. Electronic and digital forms of signature can satisfy the written-signature concept when the applicable requirements are met.

This is why a marketing checkbox should not be hidden inside acceptance of general terms. Consent evidence becomes weaker when the action has several meanings at once.

CTIA’s messaging principles sit in a different layer. They are industry guidance, not a federal statute. They nevertheless matter operationally because the US messaging ecosystem expects non-consumer senders to obtain consent before messaging and to respect opt-outs. Providers and carriers may ask for screenshots, URLs, sample messages, or other evidence showing how people enter a program.

What passes and what does not

Some programs use a second confirmation step: the person submits a number, then confirms by replying or tapping a link. This can improve list quality and evidence, particularly when numbers are frequently mistyped or shared
But “SMS opt-in” and “double opt-in” are not synonyms. Whether confirmation is legally or contractually required depends on the program and applicable rules. A glossary should not turn an optional risk-control pattern into a universal legal claim
Model the states separately if you support it:
pending confirmation -> confirmed -> revoked
That also lets the platform prevent marketing sends while the record is still pending