Search for AI cold calling real estate tools and you will find a dozen products promising to dial expired listings all day in a voice indistinguishable from a person’s. What you will struggle to find on any of those pages is a clear statement of the rule that governs them.
So here it is, in one sentence: on 8 February 2024 the FCC ruled that AI-generated voices count as “artificial” under the Telephone Consumer Protection Act, which puts a call made by an AI voice under exactly the same restrictions as a recorded robocall. The same cloned voice in criminal hands is a different problem entirely.
That single ruling changes the answer to most questions agents ask about these tools, and it is the reason this article exists. What follows is general information about how the rules are structured, not legal advice, and it is not a substitute for asking your broker and a lawyer who knows your state. Where that line sits when the tool is answering a contract question is covered in the piece on AI and contract review.
The three rules stacked on top of each other
Cold calling in the United States is governed by more than one thing at once, and a tool can comply with one while breaking another.
| Layer | Who sets it | What it restricts |
|---|---|---|
| TCPA | FCC | Artificial and prerecorded voices, and calls to mobile numbers |
| Telemarketing Sales Rule | FTC | The Do Not Call Registry, calling hours, prerecorded sales calls |
| State law | Your state | Additional consent rules, and in some states AI disclosure |
A vendor telling you their product is “TCPA compliant” is describing one of these three, usually by pointing at a consent checkbox you are expected to have collected yourself.
What the FCC ruling actually did
Before February 2024 there was an argument, made mostly by people selling AI callers, that a voice generated live by a model was not a “prerecorded” message and therefore sat outside the rules.
The FCC closed it. The Declaratory Ruling holds that AI-generated voices are artificial voices within the meaning of the statute, and the Commission’s own announcement puts it plainly: using AI to generate the voice in these calls is itself illegal where the underlying call would be. The practical consequence is that the consent standard for telemarketing applies, which is prior express written consent from the person you are calling.
Read that back against how prospecting actually works. An expired listing owner has not given you written consent. Neither has a for-sale-by-owner, or a name from a circle-prospecting list, or anybody scraped from public records. That is the entire premise of cold calling, and it is precisely the situation the consent requirement is about.
The Do Not Call Registry, which people forget entirely
Separately from anything to do with AI, the FTC’s Telemarketing Sales Rule requires telemarketers to scrub against the National Do Not Call Registry and to keep doing it regularly, with narrow exceptions for an established business relationship or written permission.
It also sets the boundaries most people half-remember: no calls before 8 a.m. or after 9 p.m. in the recipient’s time zone, prerecorded telemarketing calls require a written agreement in advance, and those messages must carry an automated way to opt out.
None of this is new and none of it is about AI. It applies to a human dialling manually. Automation simply makes it possible to break it two hundred times an hour instead of twice.
So what can you actually do?
Quite a lot, once you separate the two things that get bundled together and sold as one product.
The AI speaks: high risk
A synthetic voice holding a conversation with someone who did not ask to be called. This is what the FCC ruling is aimed at, and it is the version being marketed hardest. Outbound prospecting in this form is the part to leave alone unless a lawyer who has read your specific setup tells you otherwise.
The AI prepares, you speak: low risk
Everything that happens before and after the call is untouched by any of this, because a person is still making the call.
- Research before dialling. Summarising public information about a property so you open with something specific rather than a script. Worth knowing what these tools can and cannot actually see before you rely on anything they tell you about a property.
- Objection practice. Have a model play an irritated expired-listing owner and argue back, out loud, in the car. Genuinely the most useful thing on this list and nobody sells it.
- Call notes. Dictate thirty seconds afterwards and let a transcription tool produce the CRM entry, rather than typing it up between appointments or not at all.
- The follow-up that comes after a call that went well. Which is where most of these conversations are actually won or lost, and there are prompts for it in the piece on follow-up emails.
Inbound is a different question
Someone who calls you, or who submits a form asking to be contacted, is in a different position from someone you dialled unprompted. Many agents use AI to answer inbound enquiries out of hours for exactly that reason. It is not automatically unrestricted, and consent language and state rules still matter, but it is a far more defensible starting point than outbound.
Text messages are calls too
Worth a paragraph because agents assume texting is the loophole. Under the TCPA framework, text messages to mobile numbers are treated as calls. Automated texting to people who never consented sits in the same category as automated calling to them, and the fact that a model wrote the message changes nothing about it.
Email is governed by a different regime with lighter requirements, covered briefly in the follow-up emails piece. Do not reason from email to text. They are not the same.
Questions to ask any vendor before you sign
- Does your product use a synthetic voice on outbound calls to people who have not given written consent? If the answer is long, it is yes.
- Who is liable if a call breaks the rules, you or me? Read the contract rather than accepting the answer. It is almost always you.
- Do you scrub against the National Do Not Call Registry, and how often?
- How do you handle the time zone of the number rather than mine?
- Does the call disclose that it is AI, and can I turn that off? If it can be turned off, ask yourself why that switch exists.
- Which states do you not operate in, and why? The honest vendors have a list. It tells you more than any compliance page.
The uncomfortable business question
Set the law aside for a second, because there is a second reason to be careful about AI cold calling real estate prospecting.
Prospecting works when someone decides you are a person worth calling back. A homeowner who discovers, thirty seconds in, that the warm and attentive voice asking about their property is software does not think “how efficient”. They think they were tricked, and that impression attaches to your name, in a business built almost entirely on referral and reputation.
Every other use of AI in your business happens behind the scenes, where the client meets your work rather than the machine. Outbound calling is the one place where the machine meets the client directly, which is exactly why it is the one place to be conservative. There is a broader version of that argument in the piece on what to tell clients about AI.
Common questions
Is it illegal for me to use an AI dialler at all?
The distinction that matters is not the dialling, it is the voice. A tool that puts calls in front of a human who then speaks is a different thing under these rules from a tool that speaks with a synthetic voice. Both still have to respect the Do Not Call Registry and calling hours.
What if the AI says it is an AI at the start?
Disclosure is good practice and some states require it, but the FCC ruling is about the voice being artificial, not about whether the recipient was told. Announcing it does not by itself supply the consent the rules ask for.
Everyone in my market is doing this. Why should I not?
Enforcement in this area has historically come in waves and lands on whoever is visible when it arrives. “Common practice” has never been a defence, and TCPA exposure is per call, which is what makes automated volume the dangerous part rather than the helpful part.
Does this apply to calling my own past clients?
An established business relationship changes the analysis for some of these rules and not for others, and how far it stretches is fact-specific. It is the one situation genuinely worth putting in front of your broker rather than reading about, because the answer depends on how recent the relationship is and what they agreed to.
Sources
- Federal Communications Commission, Declaratory Ruling on AI-generated voices under the TCPA, 8 February 2024.
- Federal Trade Commission, Complying with the Telemarketing Sales Rule.
Recording those calls raises a separate question with its own state-by-state answer: see when you need every party’s consent to run an AI notetaker before you let one join a client call.
Checked August 2026. This is general information, not legal advice. Telemarketing rules change and state requirements vary; confirm with your broker and an attorney licensed in your state before running any calling campaign.