Your texts didn't fail. They were filtered. Here's how carrier filtering, A2P 10DLC registration, and consent actually work — and how to build automated follow-up that keeps arriving.
Most teams treat texting as a button inside their CRM. It isn't. Between your CRM and your lead's phone sits a chain of carriers and messaging aggregators who decide, message by message, whether your traffic is legitimate. When they decide it isn't, they don't bounce it back. They accept it, mark it delivered, and drop it.
That silent failure is why teams discover the problem late — usually as a drop in reply rate that gets blamed on the market or the lead source. Nothing in the dashboard says "blocked." The messages simply stop landing.
What follows is operational guidance from running messaging stacks for real estate businesses. It is not legal advice. Consent rules and carrier policies change, and they interact with your state's rules and your brokerage's own policy. Confirm anything here with your broker and your attorney before you rely on it.
SMS still has one enormous advantage over every other channel: people read it. Carriers know that, and they know the moment texting fills up with unwanted marketing, subscribers stop trusting it. So they built enforcement — registration requirements, reputation scoring, and content filtering — to keep commercial traffic accountable.
Real estate lands squarely in the crosshairs, and not unfairly. The industry buys aged lead lists, blasts identical templates, texts at scale from brand-new numbers, and rotates numbers when one gets burned. Every one of those behaviors is also what an actual spammer does. The filters cannot read your intent — they read your pattern.
Filtered messages usually report as delivered. Your reply rate is the only honest signal you have — which is why deliverability has to be monitored, not assumed.
A2P means application-to-person: a message generated by software rather than typed by a human thumb. 10DLC means a standard ten-digit local number — the kind with your area code that leads are willing to answer. Put together, A2P 10DLC is the framework carriers use to let businesses send software-generated messages from ordinary local numbers, on the condition that they say who they are first.
The important thing to understand is that registration is an identity system, not a permission slip. You are telling the carriers which legal business is behind the number and what you intend to send. Once that record exists, your traffic has an owner, a reputation, and something to lose. Unregistered traffic has none of that, which is exactly why it gets treated as suspect by default.
Your actual business identity — legal entity name, tax ID, address, website, and a real point of contact. This is where most real estate registrations fail: the DBA on the application doesn't match the entity on file, or the website has no working contact information behind it.
What you intend to send and to whom. You describe the use case — lead follow-up, appointment reminders, transaction updates — and provide sample messages. Vague or aspirational descriptions get rejected. Describe the traffic you actually send.
You have to state how people opted in and show it. That usually means a live URL to the form or landing page carrying the disclosure. If your opt-in language lives only in a slide deck or a screenshot, you don't have proof — you have a claim.
Each sending number is attached to the approved campaign. Numbers are not interchangeable and shouldn't be swapped casually. A number that has been sending clean, consented, replied-to traffic for months is an asset you have built — treat it like one.
Registration runs through your messaging platform and approval is not instant. Build the lead time into your launch plan rather than discovering it the week you wanted to start following up.
Federal telemarketing law — the framework people mean when they say TCPA — has been interpreted and re-interpreted for years, and the details genuinely do shift. But the operating principles underneath it have stayed remarkably stable, and you can build a system around them without needing to be a lawyer. Again: confirm the specifics with your broker or counsel, especially if you operate in a state with its own rules.
For automated marketing texts the standard is prior express written consent, and each of those words matters. Prior means before the first message, not implied by a later reply. Express means the person actually agreed to receive texts — a phone number in a form field is contact information, not agreement. Written includes a checked box or a submitted web form, as long as the disclosure was visible when they submitted it. Good disclosure names your business, says texts may be automated, notes that message and data rates may apply, and states clearly that agreeing is not a condition of buying or selling a home. A pre-checked box is not consent. Consent buried in a linked privacy policy is not consent.
Someone who wants out has to be able to get out, in the same channel, without effort. Carriers automatically process the standard keywords, but that only blocks one number-to-number pair. The obligation is broader: the contact must be suppressed in your CRM so no campaign, drip, or agent re-adds them, and the suppression must follow them across every number and every list you own. Treat opt-out as a permanent property of the contact record, not a filter on one sequence. And remember that plain English counts — "please stop texting me" is an opt-out even though no keyword was used, which means someone or something has to be reading replies.
If your consent is ever questioned, the burden of showing it lands on you. That means every contact should carry a durable record of where the consent came from, the exact disclosure language shown at the time, a timestamp, and ideally the source page or form. Screenshots in a folder are not a system. The record should live on the contact in your CRM, survive imports and migrations, and be retrievable years later without anyone reconstructing it from memory. This is also the single most useful thing you can hand your attorney if a complaint ever arrives.
Resist the urge to buy a new number. Rotating away from a filtered number is the reflex, and it is almost always the wrong first move — you lose whatever reputation you had, and the underlying cause follows you to the new number. Diagnose first.
Only after all of that should you consider new numbers — and then only as part of a fixed system, not as a way to keep the old behavior alive. This is the same discipline that governs email sending reputation; if you are also seeing campaign email land in spam, the companion guide on real estate email deliverability covers the authentication and reputation side of the same problem.
Compliance fails in real estate for a boring reason: the rules live in someone's head instead of in the system. An agent knows not to text an opted-out lead, and then the agent leaves, or a list gets imported at 11pm, or a new drip gets cloned from an old one. The fix is to make the correct behavior structural — enforced by the platform, not by memory.
Every form, landing page, and sign-in sheet that produces a phone number carries the disclosure, and the record written to the CRM includes the language, the source, and the timestamp. If a lead source can't provide that, its contacts get flagged and stay out of text automation.
Opt-out is a property of the contact, checked before any message is queued by any sequence or any user. Not a list to remember to exclude — a gate the message physically cannot pass. The same gate handles quiet hours and duplicate-send protection.
Messages that identify you, reference something specific to that person, and invite a reply behave like conversation and are treated like conversation. Sequences should also stop themselves the moment a human replies, so a real conversation is never talked over by a robot.
Texts, emails, calls, and appointments on the same contact timeline, so anyone picking up the relationship can see the whole history — including the consent record and any opt-out. Split systems are how the same lead gets texted by two agents and two sequences.
This is the design we implement in the Genesis CRM layer — pipelines, campaigns, booking, and follow-up sequences with consent and suppression enforced at the platform level rather than left to individual discipline. It sits on top of the Microsoft 365 foundation we provision, so identity, email, calendars, and messaging are one accountable stack instead of four vendors pointing at each other. For teams with multiple agents touching the same database, the team technology guide goes deeper on how shared ownership and handoffs should be structured.
AI is legitimately good at the parts of follow-up that burn an agent's evening: reading a long thread and summarizing where the conversation actually stands, drafting a reply that references what the lead said instead of what a template assumed, flagging which conversations have gone cold, and keeping tone consistent across a team.
What AI must not do is decide who gets messaged. The moment a model has open access to your database and a send button, every consent rule you have is only as reliable as a prompt. Build the boundary in the infrastructure: the system determines eligibility — consent on file, no opt-out, no active human conversation, inside allowed hours — and the AI operates only inside that set. That separation is not a setting you toggle on; it is an architecture, which is why we treat messaging as part of the stack we provision and manage rather than an app a client wires up alone. You can see how that works in our process.
Bring us your current setup and we'll walk the checklist above with you on a call — registration status, consent trail, sequence design, content triggers. If the fix is something you can do yourself in an afternoon, we'll say so. That's the point of the consult.
Genesis AI plans start at $750/mo for Foundation. Genesis CRM, where the follow-up automation lives, is included from the Professional plan at $1,750/mo.
In most cases nothing broke on your end — a carrier started filtering you. The usual causes are an unregistered or misregistered sending number, a sudden jump in volume from a number with no history, message content that trips spam heuristics, or a pattern of recipients ignoring and reporting your messages. Your sending platform will still show the message as sent, which is why the problem often goes unnoticed for weeks.
If you are typing messages one at a time from your personal phone, no. If you are sending through any software — a CRM, a dialer, a follow-up automation, a mass-text tool — then yes, that traffic is application-to-person and needs to be registered, regardless of how small your list is. Volume is not what triggers the requirement; sending through an application is.
Only if the form actually asked. A phone field with no disclosure is not consent to send automated marketing texts. The form needs clear language stating that the person agrees to receive texts, from whom, roughly what kind, that message and data rates may apply, and that consent is not a condition of buying or selling. That disclosure should sit next to the submit button, and the submitted record should capture it.
Legally that is a question for your broker or counsel, and the answer depends on what the form said and what has happened since. Operationally, texting a cold two-year-old list is one of the fastest ways to get filtered: recipients do not remember you, they report the message, and the carrier degrades your number. If you want to re-engage an aged list, do it by email first and let the people who respond opt back into texting.
The opt-out has to be honored immediately and everywhere. Carriers block the number pair automatically, but that is not enough — your CRM has to mark the contact as opted out so no sequence, campaign, or agent re-adds them later, and the suppression has to apply across every number you send from. Send one brief confirmation and then stop. Reviving an opted-out contact months later is a serious mistake.
AI is genuinely useful for drafting, summarizing a thread, and suggesting the next message based on what the lead actually said. Where teams get into trouble is letting a model send freely to anyone in the database. The safe pattern is that consent and suppression rules are enforced by the system before any message is queued, and the AI only ever writes inside that boundary — it can decide the wording, never the audience.
This page is operational guidance based on running messaging infrastructure for real estate businesses. It is not legal advice, and carrier policy and consent law both change. Confirm your own obligations with your broker and a qualified attorney before launching automated texting.
Send your details and we'll set up a consult — a working session on your current setup and what a managed AI foundation would change. No pressure, no generic sales funnel.
