AI Text Message Follow-Up: Rules, Consent and What Not to Do

Is AI text message follow-up legal?

Yes, automated texting is legal in the US when you follow the rules around consent, identification and opt-out. The main federal law is the Telephone Consumer Protection Act (TCPA), and carriers add their own requirements on top. Using AI to write or send the messages does not change those rules. An automated system is still an automated system.

The practical point: the risk is rarely the AI, it is the list and the process. Texting people who never agreed to hear from you is the problem, whether a human or a model typed the message.

This article is plain-language orientation, not legal advice. Your industry (legal, healthcare, financial services) may add rules, so get a lawyer to review your setup before you launch.

What counts as consent for AI SMS follow up?

Consent is the foundation. In plain terms, the person needs to have agreed to receive texts from your business, and the agreement should be easy to prove later. Marketing and promotional texts generally need a higher bar (clear written consent) than purely informational ones, such as confirming an appointment the person booked.

Good consent has these traits:

• Specific: it names your business and says texts will be sent to that number.
• Not bundled: a pre-checked box buried in terms of service is weak. A visible, unchecked box or a clear statement next to the submit button is stronger.
• Recorded: store the timestamp, the number, the page or form it came from, and the exact wording shown.
• Scoped: if someone asked about a quote, do not later enroll them in unrelated promotions.

Speed-to-lead texting works because a person just filled out your form and expects a reply. That expectation helps, but it does not replace having consent language on the form. If you work in a lead-driven field, our guide on speed to lead for personal injury firms shows why the first reply matters so much, and why it still has to be compliant.

What is 10DLC and why does it matter?

10DLC stands for 10-digit long code. It is the system US carriers use to let businesses send application-to-person texts from regular local phone numbers. You register your business (the brand) and describe what you will send (the campaign). Carriers then decide how your traffic is treated.

Skip registration and you will usually see filtered or blocked messages, which looks like follow-up that mysteriously never lands. Your campaign registration also asks for sample messages and your opt-in method, so what you register must match what you actually do.

• Register the brand with accurate legal business details.
• Describe the real use case, such as lead follow-up or appointment reminders.
• Keep sample messages consistent with what your AI sends.
• Use a number tied to the registered campaign, not a random personal line.

Registration requirements change, so confirm current details with your messaging provider.

How should STOP, quiet hours and frequency work?

These three settings are where most AI texting setups go wrong, because they need to be enforced by the system, not left to the model's judgment.

STOP handling. When someone replies STOP (or similar words like UNSUBSCRIBE, CANCEL, END or QUIT), the system should suppress that number immediately across every sequence and send at most one confirmation. Handle it at the platform level with a hard rule, then also train the AI to recognize natural phrasing such as 'please stop texting me' or 'wrong number'. Never argue, never ask why, never send one more nudge.

Quiet hours. A common conservative standard is to avoid texting early morning and late evening in the recipient's local time. Federal telemarketing rules point to roughly 8 a.m. to 9 p.m., and some states are stricter. Base the window on the contact's area code or stored time zone, not your office clock, and queue overnight messages for morning.

Frequency. Set a cap and a stop condition. A simple structure: a few touches in the first days after a lead arrives, then spacing widens, then the sequence ends. Any reply from a human should pause automation so a person or a context-aware AI takes over.

What should an AI text follow-up never do?

Most complaints come from a handful of repeatable mistakes. Avoid these:

• Pretending to be a person. Identify your business in the first message and answer honestly if someone asks whether they are talking to an AI.
• Texting purchased or scraped lists. No consent means no texting, however good the message is.
• Ignoring opt-outs in other channels. If someone says stop by voice mail, email or in a call, treat it as an opt-out.
• Making claims the business cannot back up. The AI should not promise outcomes, prices or availability it has not been given.
• Sending sensitive details by text. Keep medical, legal and financial specifics for a secure channel or a live conversation.
• Continuing after a wrong number. Flag it, suppress it and stop.

For regulated, high-trust businesses this matters even more. For example, the reasoning behind an AI receptionist for med spas applies equally to texting: keep it helpful, keep it private, and hand off to a human when the question gets personal.

What does good tone look like in AI follow-up texts?

Good follow-up reads like a competent front-desk person wrote it on a phone. It is short, specific to what the lead asked about, and ends with one easy next step.

A solid first message has four parts: who you are, why you are texting, one helpful line, and a simple question or link. Include opt-out language in the first message and periodically after that, for example 'Reply STOP to opt out.'

• One idea per message, ideally two short sentences.
• Use the person's name and the service they asked about, not generic filler.
• Ask a question that can be answered in a word or two.
• Skip all caps, stacked exclamation points and fake urgency.
• Let the AI say 'I'm not sure, let me get someone who knows' instead of guessing.

What is a launch checklist for compliant AI SMS follow up?

Work through this list before the first message goes out, and revisit it whenever you change your forms or sequences.

• 1. Audit your consent language on every form, call script and booking page that feeds the texting list.
• 2. Store proof of consent with timestamp, source and wording.
• 3. Complete 10DLC brand and campaign registration with accurate sample messages.
• 4. Configure global opt-out so STOP applies to all sequences and numbers you use.
• 5. Set quiet hours by recipient time zone and queue messages outside the window.
• 6. Cap touches per lead and define when a sequence ends.
• 7. Write human handoff rules for questions, complaints, pricing and anything sensitive.
• 8. Test with your own phones, including STOP, a wrong-number reply and an after-hours lead.
• 9. Review transcripts weekly for the first month, then on a regular schedule.
• 10. Have an attorney review the whole setup.

Steps 9 and 10 are the ones teams skip, and they are why ongoing oversight matters. That is the idea behind a managed setup, which we explain in what an AI ops retainer is. Our ClawOps approach is to only pitch what we can demo live, so you can see the opt-out and handoff behavior working before you commit.