AI voice agents have become cheap enough for a dental office, a roofing contractor, or a property manager to deploy in an afternoon. They call customers to confirm appointments, chase unpaid invoices, follow up on web leads, and pitch renewals, in a voice that sounds like a friendly human.
Under federal law, every one of those outbound calls is a call made with an “artificial or prerecorded voice.” That puts it under the Telephone Consumer Protection Act (TCPA), a statute that lets each person called sue for each call. And on November 25, 2026, the FCC's rules on how customers stop those calls change.
The 2024 ruling: an AI voice is an artificial voice
The TCPA has restricted calls using an “artificial or prerecorded voice” since 1991. In February 2024, the FCC removed any doubt about AI. In Declaratory Ruling FCC 24-17, adopted February 2, 2024, the Commission confirmed that “the TCPA's restrictions on the use of ‘artificial or prerecorded voice’ encompass current AI technologies that generate human voices.” As a result, such calls “require the prior express consent of the called party to initiate such calls absent an emergency purpose or exemption.”
The ruling covers both a wholly synthetic voice and a cloned voice taken from a real person's audio. It also reminds callers that the identification, disclosure, and opt-out rules for artificial-voice calls apply, and that those requirements reach “any AI technology that initiates any outbound telephone call using an artificial or prerecorded voice to consumers.”
Inbound and outbound are different problems
The word that matters is “initiates.” The TCPA's robocall restrictions are about making calls. An AI receptionist that only answers calls your customers place to you is not initiating calls, so the consent rules discussed here don't apply to that conversation. Other laws can, including recording-consent statutes if the call is recorded or transcribed (see our piece on AI notetakers and recording consent) and state bot-disclosure laws (see the 2026 chatbot disclosure map).
The moment that same receptionist calls the customer back, sends a reminder call the next day, or dials a list of leads, it is placing an artificial-voice call. Many voice-agent products do both, so map which of your call flows are outbound before anything else.
What consent you need for an outbound AI call
The FCC's rule, 47 CFR 64.1200, splits calls into two kinds. An informational call contains no advertising or telemarketing: an appointment reminder, a delivery notice, a past-due notice. A telemarketing call encourages a purchase or includes an advertisement, such as a renewal pitch, an upsell, or a promotion.
- Informational calls to cell phones need the called party's prior express consent. Since 1992 the FCC has said that calling a number “provided as one at which the called party wishes to be reached” does not violate its rules, so a customer who gives you a cell number for their account has generally consented to calls about that account. Keep a record of where and why the number was given.
- Informational calls to residential landlines for a commercial purpose need no consent if you make no more than three such calls in any 30-day period and honor opt-out requests, under 64.1200(a)(3)(iii).
- Telemarketing calls with an artificial voice, to cell phones or residential lines, require prior express written consent under 64.1200(a)(2) and (a)(3). That means a signed agreement, electronic signatures included, that clearly authorizes calls using an artificial or prerecorded voice to a specific number, and that discloses the customer is not required to sign as a condition of buying anything.
One appeals court has pushed back on the written-consent rule. In Bradford v. Sovereign Pest Control of TX, decided February 25, 2026, the Fifth Circuit held that the statute itself requires only “prior express consent,” which can be oral or written, for any prerecorded call. It said the statute “provides no basis for concluding that telemarketing calls require prior express written consent but not oral consent,” contrary to the FCC's regulation. The customer there had given his cell number on a service agreement and confirmed in later conversations that the company could call him.
Bradford binds federal courts in Texas, Louisiana, and Mississippi. Elsewhere, the FCC's written-consent requirement remains the rule courts start from, and even inside the Fifth Circuit, a signed written consent is better evidence than a remembered conversation. For an outbound AI sales campaign, get written consent.
What every AI-voice call must say
Consent is only half of it. Under 64.1200(b), every artificial or prerecorded voice message must:
- state clearly, at the beginning, the identity of the business responsible for the call, using the name under which it is registered to do business with the state;
- state clearly, during or after the message, a telephone number for that business (not the number of the system that placed the call); and
- for telemarketing calls, and for informational calls to residential lines made under the no-consent exemptions, provide an automated, interactive voice or key-press opt-out mechanism, with brief instructions, so the person can make a do-not-call request during the call.
A voice agent that opens with “Hi, it's Jessica!” and never names the business fails the first requirement on the first sentence. Write the opening into the script, and don't let the model improvise it.
What changes on November 25, 2026
On September 30, 2026, the FCC adopted a Report and Order (FCC 26-67) revising how consumers revoke consent. It was published at 91 FR 67214 on October 26, 2026, and takes effect November 25, 2026. Three changes matter for voice agents.
Revocations can be limited to the category of informational call. Under the revised rule, a revocation made in response to an informational call may be treated as applying “only to the specific category of informational robocalls and text messages to which the revocation was directed.” A patient who opts out of payment-reminder calls can still get appointment reminders. That flexibility does not extend to marketing: a revocation in response to a telemarketing call revokes consent to all future telemarketing calls from that caller.
You may designate an exclusive way to revoke. A caller may designate one or more of these as the exclusive revocation method: an automated, interactive voice or key-press opt-out on the call; a reply text using a standard word (“stop,” “quit,” “end,” “revoke,” “opt out,” “cancel,” or “unsubscribe”); or a website or phone number the caller designates. The catch is disclosure: “The caller must clearly and conspicuously disclose the designated method to revoke consent on the call or in the text.” A caller that designates no exclusive method must process revocations “made by any reasonable means,” and the customer gets a rebuttable presumption that they revoked.
This is where AI voice agents create a specific risk. A conversational agent hears things like “please stop calling me” or “take me off your list” constantly. If you haven't designated and disclosed an exclusive method, a spoken request like that is very likely a reasonable revocation, and your system has to catch it, log it, and act on it. If you have designated a key-press opt-out, say so on every call, and consider honoring the plain-language request anyway. It costs you one customer you were going to lose, and it avoids an argument about whether your disclosure was clear and conspicuous.
Ten business days. All revocations must be honored within a reasonable time not to exceed ten business days from receipt. The new rule also supersedes a delayed portion of the earlier revocation rule that had been pushed to January 31, 2027.
The FCC isn't finished. A companion further notice of proposed rulemaking (91 FR 67239) asks whether to shorten the ten-day window, require two-way texting, require a single “revoke all” method, and change how revocations apply to affiliates. Comments are due November 25, 2026, and replies December 28, 2026. As of November 2026, those are proposals, not rules.
State rules: California's artificial-voice disclosure
States add their own requirements. California's automatic dialing-announcing device statute, Public Utilities Code § 2874, was amended by AB 2905 (Statutes of 2024, chapter 316). When the statute applies, the call must begin with an unrecorded, natural-voice announcement that identifies the caller and asks whether the person consents to hear the prerecorded message, and that announcement must now inform the person if the message uses an artificial voice. The amendment defines an artificial voice as one generated or significantly altered using artificial intelligence. Other states have their own telemarketing and do-not-call statutes, so check the law in each state you call into.
Why the stakes are high
The TCPA gives the person called a private right of action. Under 47 U.S.C. § 227(b)(3), a plaintiff can recover actual monetary loss or $500 for each violation, whichever is greater, and a court may increase the award up to three times if the violation was willful or knowing. State attorneys general can also sue for $500 per violation. Damages are counted per call. An agent that places a few thousand reminder calls a month without a clean consent record can produce a very large number.
A deployment checklist for AI voice agents
Consider a 15-employee HVAC company that wants an AI agent to confirm service appointments and, in the off season, call past customers about maintenance plans. Here is how it might set up the program.
Separate the call types. Appointment confirmations are informational. Maintenance-plan calls are telemarketing. Run them as separate campaigns with separate consent checks.
Capture written consent for marketing calls. Add a clear, separate checkbox or signed authorization to the service agreement or online booking form that covers calls using an artificial or prerecorded voice to the number given, with the “not required to purchase” disclosure. Store the record with a timestamp.
Lock the opening and closing script. The agent should open with the registered business name and give a callback number before it hangs up. Marketing calls need the automated opt-out mechanism and instructions.
Decide on an exclusive revocation method before November 25. If you designate one, disclose it on every call. Either way, set the agent to recognize spoken opt-out requests and route them to your do-not-call list within ten business days, and keep track of which category of informational calls each revocation covers.
Scrub and honor do-not-call lists. Telemarketing calls must respect the national registry and your internal do-not-call list.
Read the vendor contract. The rules make the business “responsible for initiating the call” the one that must identify itself, and that will usually be you, not the software vendor. Ask what the vendor does to support consent records, opt-out handling, and call logs, and who bears the cost of a violation caused by the platform. Our AI vendor contract checklist covers indemnity and liability clauses.
Watch your marketing claims about the agent itself. If you sell AI voice services, the FTC expects you to back up what you say they can do. See how the FTC polices AI marketing.
Frequently asked questions
Does the TCPA apply if a human reviews the AI agent's calls afterward?
Yes. The question is whether the call was made using an artificial or prerecorded voice, and the FCC has said AI-generated voices are artificial voices. Human review after the fact doesn't change how the call was made.
Can my AI agent call customers to remind them about appointments?
Generally yes, if the customer gave you the number and the call is purely informational. The call must still identify your business at the start and give a callback number, and you must honor requests to stop within ten business days. Adding a sales pitch turns it into a telemarketing call with stricter consent rules.
Does the Fifth Circuit's Bradford decision mean oral consent is enough everywhere?
No. Bradford binds federal courts in Texas, Louisiana, and Mississippi. Outside that circuit, the FCC's requirement of prior express written consent for artificial-voice telemarketing remains in its rules, and written consent is the safer practice everywhere.
This article reflects FCC Declaratory Ruling FCC 24-17, the FCC's Report and Order FCC 26-67 as published at 91 FR 67214 (effective November 25, 2026), 47 CFR 64.1200, 47 U.S.C. § 227, Bradford v. Sovereign Pest Control of TX (5th Cir. 2026), and California Public Utilities Code § 2874 as of November 2026. This is general information, not legal advice for your situation.
This article is for educational purposes only and does not constitute legal, tax, or financial advice. Consult a qualified professional for guidance specific to your situation.