AI voice calls and disclosure: FCC and state rules

Start with the distinction that matters: inbound or outbound

Most home service companies use AI voice in one of two ways. Inbound answering means the customer calls you and an AI system answers, takes details, books or escalates. Outbound calling means your system dials the customer, for a callback, an appointment reminder, a maintenance offer or a review request. The federal rules that get the most attention, under the Telephone Consumer Protection Act, are mainly about calls that are made to people. So the first question for any AI voice project is which direction the calls go.

Inbound answering is the lower-risk starting point because the customer chose to call you. Even so, three obligations still apply on every inbound call: honesty about what the caller is talking to, consent for any recording, and accurate information. Outbound use adds consent records, identification rules and calling hours on top of those three, which is why many companies launch inbound first and add outbound later, if at all.

The FCC's 2024 ruling on AI-generated voices

On February 8, 2024 the Federal Communications Commission unanimously ruled that calls made with AI-generated voices are artificial under the Telephone Consumer Protection Act^[1]. The practical effect is that an outbound call using a synthetic or cloned voice is treated like a prerecorded or artificial voice call. The consent requirements and other TCPA protections that apply to those calls apply to AI voice calls too.

For a contractor, the lesson is straightforward: do not treat AI voice outbound calling as a loophole. If you would need consent to send a prerecorded call to that number, you need it for an AI voice call. Keep a record of how and when each customer agreed to be contacted, and for what kind of calls.

Federal call rules that apply alongside

Two parts of the FCC's call rules are especially relevant. First, every artificial or prerecorded voice message must, at the beginning, state clearly the identity of the business responsible for initiating the call, using the name under which it is registered to do business^[2]. An AI voice callback that opens with a first name and no company name fails that test. Second, telephone solicitations to residential subscribers are not allowed before 8 a.m. or after 9 p.m., local time at the called party's location^[2].

Marketing calls also run into the National Do Not Call Registry, which had 259 million active registrations as of September 30, 2025^[5]. If an outbound campaign promotes services, such as a maintenance plan offer, scrub the list against the registry and your own internal do-not-call list, and respect opt-outs immediately.

State bot and AI disclosure laws

States have started to write their own disclosure rules, and they are not uniform. Two examples show the range.

• Utah: a supplier using generative AI to interact with a person in a consumer transaction must disclose that the person is interacting with generative AI and not a human if the person asks or prompts about it with a clear and unambiguous request. Regulated occupations face a stricter duty to disclose prominently in high-risk interactions, verbally at the start of a voice interaction^[3].
• California: it is unlawful to use a bot to communicate with a person in California online with the intent to mislead them about its artificial identity in order to incentivize a sale; disclosure that it is a bot avoids liability and must be clear, conspicuous and reasonably designed to inform^[4]. Whether and how that online rule reaches a phone call is a question for counsel.

Other states are considering or passing similar rules. Rather than tracking each one call by call, many companies adopt the strictest common denominator: disclose up front and never deny being AI. It costs a few seconds per call and matches what customers say they want. In Pew's 2025 survey, 76% of US adults said it is extremely or very important to be able to tell whether content was made by AI or by people^[6].

Recording consent

AI answering systems usually record and transcribe calls. Federal law allows recording when one party to the call consents^[7], but some states require everyone on the call to consent; Florida, for example, permits intercepting a call only when all parties have given prior consent^[8]. Because callers can be anywhere, the common approach is a short notice at the start of every recorded call, such as calls are recorded for quality and to help us serve you. Our guide to call recording consent by state goes deeper.

A practical policy by situation

Questions to ask your AI answering vendor

Your vendor will not take legal responsibility for how you use its product, but its answers tell you how much of the compliance work it has already done and how much falls on you.

1. Can I set my own greeting and disclosure wording, and is it read word for word?
2. Does the system ever place outbound calls on its own, or only when I configure it to?
3. Where is the recording notice played, and can I change it per phone number?
4. How are consent and opt-out records stored for outbound calls, and can I export them?
5. How does the system determine the called party's local time?
6. Does any feature clone a real person's voice, and can I switch it off?
7. How long are recordings and transcripts kept, and who at your company can access them?

Keep the answers in writing with your launch notes. If a regulator, a customer or your own lawyer asks later how a call flow was set up and why, you will want that record.

Texts and emails are a separate track

Many AI answering tools also send text messages, such as confirmations, reminders or follow-ups. Text messaging has its own consent expectations, carrier registration requirements and opt-out handling, and those rules are not covered by a voice disclosure line. Treat any automated text program as its own project, with its own consent language at the point of booking, its own opt-out process and its own legal review. Do not assume that a customer who agreed to a phone callback also agreed to receive marketing texts.

Compliance checklist before launch

1. 01

   Map every call flow

   List each inbound and outbound use: after-hours answering, overflow, callbacks, reminders, follow-ups, promotions.

2. 02

   Write the disclosure lines

   One sentence for the greeting and one for when a caller asks. Keep the company's registered name in both.

3. 03

   Add the recording notice

   Place it before recording starts, on every recorded line.

4. 04

   Build the consent record

   For outbound calls, store when and how each customer agreed, which number and for what purpose. Keep opt-outs and honor them quickly.

5. 05

   Set calling hours

   Configure outbound calls to the called party's local time and keep inside 8 a.m. to 9 p.m. at a minimum.

6. 06

   Get legal review

   Have counsel review outbound use, marketing content and your recording approach for the states you serve.

Example: one company's decision

Example: a fictional HVAC company in two states launches AI answering for after-hours and overflow calls, with a greeting that names the company and says it is an AI assistant, and a recording notice. It decides not to use AI voice for outbound marketing at all. Callbacks for urgent calls come from the on-call technician, a person, and appointment reminders go only to customers who agreed to them at booking, opening with the company's registered name. A lawyer reviews the flows once before launch and again when the company adds a third state.