Yes — it is legal in the United States to use an AI assistant to text and call your real estate leads, but only if the right consent exists before the first message goes out. Under the Telephone Consumer Protection Act (TCPA, 47 U.S.C. § 227) and the FCC's February 2024 declaratory ruling, an AI-generated or cloned voice counts as an "artificial voice," so AI voice calls that promote your services require prior express written consent: a signed, clear and conspicuous agreement that names the specific business doing the calling and states that consent is not a condition of purchase. AI texting falls under the same standard whenever the message is promotional; a purely responsive text back to a lead who just submitted a form asking about a listing is treated more leniently, because the number was provided for that exact purpose — but it is still safest to capture written consent at the form. The stakes are not theoretical. The TCPA carries statutory damages of $500 per violation and up to $1,500 per willful or knowing violation, with a four-year statute of limitations and a private right of action that fuels class actions — and publicly reported TCPA settlements routinely land in the seven-figure range. Layer on the National Do Not Call Registry, the FCC's revocation rules (consumers can opt out by any reasonable means, and you must honor it within 10 business days), federal calling hours of 8:00 a.m. to 9:00 p.m. in the consumer's local time, A2P 10DLC carrier registration, and a growing set of state mini-TCPAs in Florida, Oklahoma, Washington and Maryland that are often stricter than federal law. This guide walks through what consent you actually need, what your Zillow, Realtor.com and Meta lead flows really give you, how to build an opt-out system your AI cannot break, and what AI-specific disclosure laws now add in states like California, Utah and Colorado. Important: this is general information for real estate professionals, not legal advice. TCPA case law and FCC rules move quickly — confirm your specific setup with your own counsel or compliance team before you launch.
Is it legal to use an AI bot to text or call real estate leads in the U.S.?
Yes — with consent. No U.S. law bans AI from texting or calling real estate leads. What the law regulates is how the message is delivered, what it says, and whether the consumer agreed to receive it.
Three federal layers apply. First, the TCPA (47 U.S.C. § 227) and the FCC's implementing rules at 47 C.F.R. § 64.1200: marketing calls and texts sent with an autodialer, a prerecorded message, or an artificial voice require prior express written consent — and in February 2024 the FCC confirmed that AI-generated voices are artificial voices for this purpose. Second, the FTC's Telemarketing Sales Rule and the National Do Not Call Registry, which prohibit solicitations to registered numbers absent consent or an established business relationship (18 months after a transaction, 3 months after an inquiry). Third, the carrier layer: A2P 10DLC campaign registration, CTIA messaging principles, and STIR/SHAKEN caller authentication, which determine whether your AI's messages get delivered at all.
Then the distinction most agents miss. An AI text that simply answers a lead who just submitted "send me info on 2-bedroom condos in Brickell" is responsive and informational; courts have generally treated a number given for that purpose as prior express consent for that specific inquiry. The moment the AI pivots to promoting other listings, your brokerage's services, or a lender partner, it becomes telemarketing and the written-consent standard applies. Because that line moves fast inside a live conversation, most compliant brokerages simply collect prior express written consent at the form and stop worrying about the classification.
This is general information, not legal advice. Confirm your specific flows with your own counsel.
What counts as "prior express written consent" — and what doesn't?
Prior express written consent (PEWC) is a defined term, not a vibe. Under 47 C.F.R. § 64.1200(f), it is a written agreement, signed by the consumer, that clearly and conspicuously discloses that (a) the consumer authorizes the identified seller to deliver advertisements or telemarketing messages using an autodialer, prerecorded voice, or artificial voice, (b) to the specific number the consumer provides, and (c) that signing is not a condition of purchasing any property, good or service. An E-SIGN compliant electronic signature — a checkbox plus a submit action, with the disclosure visible on the same screen — satisfies the writing requirement.
What does not count, in practice: a pre-checked box; consent buried behind a hyperlink the consumer never opens; a disclosure that names only your CRM vendor or a generic "our marketing partners"; a business card handed to you at an open house; a phone number scraped from an expired listing or FSBO sign; or a number appended by a data-enrichment tool the consumer never gave you. An established business relationship helps with Do Not Call exposure, but it does not substitute for PEWC when an autodialer or artificial voice is involved.
Three operational details matter for AI systems specifically. Consent is tied to the number, not the person — if the lead gives you a new number, you need consent for that number. Consent is tied to the named seller, so a team that rebrands or a brokerage that gets acquired should re-paper. And you must be able to prove it: store the exact disclosure text shown, timestamp, IP address, user agent, form URL, and the consent version, and keep it for at least the four-year statute of limitations, ideally five. In litigation, the record is the defense.
Does the FCC treat an AI voice differently from a live agent on the phone?
Yes — and this is the single biggest change for real estate teams running AI voice callers. On February 8, 2024, the FCC issued a declaratory ruling stating that calls using AI-generated, cloned or synthetic human voices fall within the TCPA's existing restriction on "artificial or prerecorded voice." That means an AI voice agent calling a consumer's cell phone for marketing purposes needs prior express written consent, regardless of whether a human dials the number and regardless of whether the system meets the narrow autodialer definition the Supreme Court set in Facebook v. Duguid (2021), which requires random or sequential number generation.
By contrast, a human agent dialing manually from a list, with no prerecorded or artificial voice, has a materially lighter federal burden — Do Not Call rules and state law still apply, but the artificial-voice trigger doesn't. This is why the compliance question changes the moment you swap a person for a synthetic voice.
Separately, artificial-voice calls must identify at the start of the message the individual or business responsible for the call, and provide a telephone number during or after the message. Some states now require an affirmative AI disclosure as well. The FCC has also opened proceedings exploring mandatory disclosure that a call uses AI-generated content; treat proposed rules as a signal of where enforcement is heading, not as current obligations, and check the current status before launch.
Practical takeaway for brokerages: AI voice is the highest-risk channel. Many teams run AI on SMS, WhatsApp and email — where consent is easier to document and opt-out is native — and reserve AI voice for inbound calls and for leads who explicitly opted in to voice contact. Confirm your approach with counsel.
Do Zillow, Realtor.com and paid-ad leads come with consent you can rely on?
Often not the consent you need. Portal and aggregator leads are the most common source of TCPA exposure in residential real estate, because the consent was captured by someone else, on someone else's form, naming someone else.
Here's the mechanics. When a consumer submits a form on a portal or a third-party lead site, the disclosure typically authorizes contact from that platform "and its partners." Whether that language extends to you — the agent who later bought the lead — is exactly what plaintiffs' firms litigate. The FCC's "one-to-one consent" rule, which would have required lead generators to obtain consent for one clearly identified seller at a time, was vacated by the Eleventh Circuit in January 2025 (Insurance Marketing Coalition Ltd. v. FCC) before it took effect. That removed a federal bright line, but it did not remove the underlying requirement that consent be clear, conspicuous, and tied to the entity actually calling. Confirm the current regulatory status — this area continues to move.
Vicarious liability compounds it. Under long-standing FCC guidance, a seller can be liable for calls made on its behalf by lead vendors, ISAs, or contracted callers under agency principles. Buying leads does not outsource the risk.
What to do operationally: contractually require vendors to deliver, per lead, the exact disclosure text, timestamp, IP and URL — not a blanket "leads are TCPA compliant" warranty; add indemnification with real teeth; scrub against the National DNC Registry and your internal DNC list before your AI touches a record; suppress litigator and known-plaintiff numbers; and for aged or resold data, get fresh consent before any automated outreach. With U.S. real estate cost per lead running roughly $20–$60 on Meta lead forms, $40–$200 on Google Search in competitive metros like Miami, Austin and Phoenix, and $50–$500+ per Zillow Premier Agent lead depending on ZIP, a single class action erases years of media budget.
How do you build a consent and opt-out system your AI can't break?
Compliance for AI outreach is an architecture problem, not a script problem. Six components do most of the work.
One: a consent-capture layer. Put the PEWC disclosure in visible text directly above the submit button, unchecked by default, naming your legal entity, covering SMS, calls, AI/automated technology, and stating that consent isn't required to buy or sell. Log the full evidence packet with the lead record.
Two: a suppression layer that runs before every send. National DNC scrub (typically monthly at minimum), internal DNC list — which the TCPA requires you to maintain and to honor for five years, with a written policy available on request — plus wireless/reassigned-number checks against the FCC's Reassigned Numbers Database, litigator lists, and any state-specific registries.
Three: opt-out handling that understands natural language. Since the FCC's revocation order took effect on April 11, 2025, consumers may revoke consent by any reasonable means, and you must honor it within 10 business days. Your AI must recognize STOP, QUIT, END, CANCEL, UNSUBSCRIBE, REVOKE and OPT OUT — and also "please stop texting me," "take me off your list," or a spoken "don't call again." A further provision extending revocation across a sender's messages took effect in April 2026; verify the current scope.
Four: time-zone enforcement. Federal calling hours are 8:00 a.m.–9:00 p.m. in the called party's local time, derived from their location, not their area code. Some states are narrower.
Five: carrier compliance. Register your brand and campaign under A2P 10DLC (roughly a $4 one-time brand fee plus about $10–$15 per month per campaign), use a dedicated number, keep HELP responses live, and monitor opt-out rate — sustained rates above ~1–2% invite carrier filtering.
Six: immutable logging of every message, disclosure and revocation. If you cannot reconstruct a conversation four years later, you cannot defend it.
Which state laws go beyond the TCPA for AI calls and texts?
Federal compliance is the floor. A growing set of state "mini-TCPAs" and AI-disclosure statutes sit on top, and several give consumers their own private right of action with $500–$1,500 statutory damages — meaning a state claim can survive even where a federal one fails.
On the calling side: Florida's Telephone Solicitation Act (FTSA) requires prior express written consent for automated sales calls and texts, narrows calling hours to 8:00 a.m.–8:00 p.m., and caps commercial call attempts — its 2023 amendment narrowed the scope and added a 15-day cure period for text claims, but Florida remains the most litigated state for real estate texting. Oklahoma's Telephone Solicitation Act mirrors much of the Florida framework. Washington's Commercial Electronic Mail Act and its telephone solicitation statutes impose their own identification and consent duties, with Consumer Protection Act exposure. Maryland enacted a Stop the Spam Calls Act extending PEWC-style requirements to more calls and texts. Confirm current effective dates and thresholds before relying on any of these.
On the AI side, three matter for real estate. California's B.O.T. Act (Bus. & Prof. Code § 17940 et seq.) makes it unlawful to use a bot to communicate with a Californian to incentivize a sale or transaction without clearly and conspicuously disclosing that it is a bot — so your AI should identify itself. Utah's AI Policy Act requires disclosure of generative AI use, proactively for regulated occupations and on request otherwise. Colorado's AI Act (SB 24-205) treats housing as a "consequential decision," bringing duties around high-risk AI systems and algorithmic discrimination; its effective date was pushed into 2026, so verify the operative date.
And never forget the Fair Housing Act: an AI that qualifies leads differently by protected class — including via proxies like ZIP code or language — creates discrimination exposure independent of the TCPA. Again, general information only; get state-specific advice.
What does a TCPA violation actually cost a brokerage?
Do the arithmetic before you launch, because TCPA damages are per message, not per campaign. Statutory damages are $500 per violation, trebled to $1,500 for willful or knowing violations, with no cap and no requirement that the consumer prove actual harm. The statute of limitations is four years.
A simple scenario: an AI SMS sequence sends five messages each to 4,000 purchased leads over a quarter. If consent for that list is later found defective, that's 20,000 potential violations — $10 million at $500 each, $30 million if willfulness is found. Even a modest class typically resolves in the seven figures, and publicly reported TCPA class settlements have repeatedly landed in that range across industries. Defense costs alone commonly run into the hundreds of thousands before any settlement.
Three secondary costs get overlooked. First, insurance: many general liability and E&O policies contain express TCPA or "invasion of privacy / statutory violation" exclusions, so the brokerage often pays out of pocket — read your policy, and ask your broker about affirmative TCPA coverage. Second, carrier consequences: high opt-out or spam-report rates trigger 10DLC campaign suspension and number blocking, which kills your speed-to-lead advantage overnight. Third, vicarious exposure: brokers can be pulled into claims arising from an individual agent's or ISA vendor's AI outreach under agency principles.
Risk reduction is unglamorous and effective: written consent at the point of capture, scrubbing before every send, honoring revocations within 10 business days, four-plus years of immutable logs, vendor indemnification with proof-of-consent delivery obligations, and an annual review with counsel. Compare that to the cost of a class action and the ROI is obvious. This is general information, not legal advice.
How do you keep sub-5-second lead response without breaking compliance?
Speed and compliance are not in conflict — they're solved in the same workflow. Industry benchmarks consistently show that responding to an inbound web lead within the first five minutes dramatically increases the odds of making contact and qualifying versus waiting 30 minutes or more, while a large share of internet leads in U.S. residential real estate never receive any response at all. That gap is exactly why AI responders spread so fast. The fix is to put the compliance checks inside the response path rather than after it.
A workable architecture looks like this. The lead form captures PEWC in visible text with an unchecked box and writes the evidence packet to the CRM. On submit, a pre-send gate runs in milliseconds: consent present and current, number not on the national or internal DNC list, not a known-litigator number, local time inside the permitted window, state rules applied. If the gate passes, the AI replies in under five seconds by SMS or WhatsApp, identifies itself as an AI assistant working for a named brokerage, includes opt-out instructions in the first message, answers the lead's actual question, and qualifies on budget, timeline, financing and area. If the gate fails — say the lead arrives at 11:40 p.m. local time — the message queues to 8:00 a.m. rather than sending. Every exchange is logged immutably; any revocation language flips the record to suppressed instantly rather than within the 10-business-day maximum.
This is the model we build into the Estate Funnel at Growth Estate: consent capture, suppression, AI qualification and human handoff as one system, so agents get qualified conversations instead of raw form fills — and a defensible audit trail behind every one. Before you switch it on, have your own counsel review your consent language, your vendor contracts and your state footprint.
Frequently asked questions
Yes, if you have the right consent. Under the TCPA (47 U.S.C. § 227), promotional texts sent with automated technology require prior express written consent: a signed, clear and conspicuous agreement naming your business, covering the specific number, and stating that consent is not a condition of purchase. A purely responsive text answering a lead who just submitted an inquiry is treated more leniently, since the number was provided for that purpose, but capturing written consent at the form is the safe standard. You also need DNC scrubbing, 8:00 a.m.–9:00 p.m. local calling hours, working opt-out handling, and A2P 10DLC registration. This is general information, not legal advice.