Using AI in your business · Updated October 5, 2026
Should an AI receptionist tell callers it is an AI?
Updated October 5, 2026
Yes. An AI receptionist should tell callers it is an AI in the first sentence of its greeting. Maine requires clear notice when a chatbot, voice included, could pass for a person, and Utah requires an AI to admit it when a consumer clearly asks. Federal robocall rules cover outbound AI calls, so they apply once it calls customers back.
Put the disclosure in the first sentence the caller hears
If an AI voice agent answers your phone, have it say it is an automated assistant before it asks the caller anything. One sentence at the start, plus a straight answer whenever a caller asks, meets the state disclosure laws covered below, and it costs nothing to say.
The law does not yet say this everywhere. The federal robocall rules do not require an inbound AI receptionist to announce itself, and the states that have written a rule wrote it in different ways: one says notify when a reasonable consumer could be misled, one says answer honestly when a consumer clearly asks, one only covers online bots. Writing a greeting for whichever state the caller happens to be in is harder than writing one that meets the strictest of them.
Callers ask anyway. A good voice agent sounds close to a person. A caller told in the first five seconds can carry on, ask for a person, or call back. A caller who finds out halfway through wonders what else was not as it seemed. That is a judgment, not a statute, and it is the reason to disclose even where no law asks yet.
For example
"Thanks for calling Ridgeline Plumbing. I'm the company's automated assistant, and this call is recorded. I can book a visit, answer questions or put you through to the team. How can I help?" Three facts, under ten seconds: what is answering, that the call is recorded, and that a person is available.
Sort every rule by the direction of the call
Many of the AI phone rules people cite were written for robocalls, so check which way the call goes before deciding whether a rule applies to you.
Calls the AI answers.
State AI disclosure laws can apply, along with call recording laws if the system records. The federal robocall rules do not.
Calls the AI places.
The federal robocall rules apply: consent before most calls, the business named at the start, and written consent for anything that sells. State autodialer rules can apply too.
The FCC drew this line itself. In an August 2024 proposal it wrote that "The TCPA's requirements do not extend to technologies used to answer inbound calls." The TCPA (Telephone Consumer Protection Act) is the main federal robocall law. The same system can sit on both sides of that line: a receptionist that answers calls is inbound, and the same agent dialing back a missed caller is outbound.
State laws that reach calls the AI answers
If you take calls from Maine consumers, tell them clearly that no person is on the line
Maine's law is the clearest fit for an AI receptionist, because it covers voice. A business may not use "an artificial intelligence chatbot or any other computer technology" in trade with a consumer "in a manner that may mislead or deceive a reasonable consumer into believing that the consumer is engaging with a human being unless the consumer is notified in a clear and conspicuous manner that the consumer is not engaging with a human being."
The law defines a chatbot as software that "simulates human conversation and interaction through textual or aural communications." Aural means spoken. A natural sounding voice agent fits that definition, and a violation counts as a violation of the Maine Unfair Trade Practices Act.
What it asks for is notice, not a script. The statute sets no wording and no point in the call. A plain statement at the start of the greeting is the surest way to make it "clear and conspicuous", because a caller who has already given an address and a problem has already been engaging.
If you serve customers in Utah, the AI must say so when a caller asks
Utah passed its Artificial Intelligence Policy Act in 2024 and added consumer disclosure rules in 2025 (S.B. 226), now in Title 13, Chapter 77 of the Utah Code. They cover "generative artificial intelligence" designed to simulate conversation through text, audio or visual communication that "generates non-scripted outputs", so a voice agent built on a language model fits.
- When a caller asks. A business that uses generative AI to interact with someone in a consumer transaction "shall disclose to the individual that the individual is interacting with generative artificial intelligence and not a human" if the person asks. The question "must be a clear and unambiguous request" to find out whether it is a human or AI.
- Licensed occupations, high-risk calls. Someone providing services in a regulated occupation, one regulated by the Department of Commerce that needs a license or state certification, must "prominently disclose" the AI when the interaction is high risk: collecting health, financial or biometric data, or giving advice such as financial, legal, medical or mental health advice. That disclosure comes "verbally at the start of a verbal interaction."
- The safe harbor. A business is not subject to enforcement under the disclosure section if its AI "clearly and conspicuously discloses" at the outset of the interaction, and throughout it, that it is generative AI, is not human, or is an AI assistant.
The same chapter says it is not a defense to a consumer protection violation that generative AI "made the violative statement." A wrong quote or a broken promise from the receptionist is still the business's problem. A greeting that discloses the AI, plus a rule that it never commits to prices it cannot read from your own list, covers both.
If you also run a website chat, California's bot law covers that, and probably not the phone
California's bot disclosure law is often cited for AI receptionists, and it is written for online accounts. Disclose on your website chat for certain; treat the phone line as a matter of the greeting you already have.
The law makes it unlawful to use a bot to talk with someone in California "online, with the intent to mislead the other person about its artificial identity" to "incentivize a purchase or sale of goods or services in a commercial transaction or to influence a vote in an election." A business "shall not be liable under this section if the person discloses that it is a bot," and the disclosure must be "clear, conspicuous, and reasonably designed to inform" the person.
Online is defined. It means "appearing on any public-facing Internet Web site, Web application, or digital application, including a social network or publication," and a bot is "an automated online account." A phone call is not mentioned. Reading that text, a chat widget on your site appears online in the statute's sense and a phone line is not named. A disclosing greeting makes the question moot either way.
Does the FTC's impersonation rule apply to an AI receptionist?
Not to one answering honestly for your own business. The FTC's rule makes it a deceptive practice to "materially and falsely pose as, directly or by implication, a business or officer thereof." An assistant that names your company and says it is automated is not posing as anyone.
The trap is a human persona. A receptionist that calls itself "Sarah from the office" and denies being automated when asked is the pattern Maine's law targets. Give the assistant a name if you like; never let it claim to be a person.
Before the AI calls anyone back
If the system will dial out, get consent first and name the business at the start
Turn on outbound calling only after checking the consent you hold for each kind of call. The moment an AI voice places a call, it is an artificial voice under federal law.
In February 2024 the FCC ruled that the TCPA's limits on "artificial or prerecorded voice" calls "encompass current AI technologies that generate human voices," and that such calls "require the prior express consent of the called party to initiate such calls absent an emergency purpose or exemption." It added that the TCPA does not allow "any carve out of technologies that purport to provide the equivalent of a live agent." An AI that holds a conversation is still an artificial voice.
- Calls to cell phones. An AI voice call to a mobile number needs the person's prior express consent, unless it is an emergency call.
- Calls that sell anything. If the call "includes or introduces an advertisement or constitutes telemarketing," it needs prior express written consent: an agreement signed by the person called that authorizes your business to call with an artificial voice, and gives the number to call.
- Non-sales calls to a home landline. A commercial call with no sales content needs no consent if you make "no more than three calls within any consecutive 30-day period" to that line and honor a request to stop.
- Every AI voice call. It must "At the beginning of the message, state clearly the identity of the business" responsible, using the name it is registered under, and give a callback number during or after the message.
For example
A homeowner calls at 7 p.m., hangs up before the assistant answers, and the system offers to call back. The callback is an outbound artificial voice call. To a cell phone, it needs that person's prior express consent. If the callback pitches a maintenance plan, it needs signed written consent, which a missed call does not give you. A callback that only confirms a booked visit is a different call from one that sells.
If you call California numbers with an autodialer and no prior agreement, a live person speaks first
California allows automatic dialing-announcing devices only with the called person's prior agreement or under its announcement rule, which requires an "unrecorded, natural voice announcement" before the device plays its message. Since January 1, 2025 that announcement must also "Inform the person called if the prerecorded message uses an artificial voice."
The law defines artificial voice as "a voice that is generated or significantly altered using artificial intelligence," and bars these devices from placing calls that arrive "between 9 p.m. and 9 a.m. California time." It covers equipment that stores or generates numbers and plays a message to them, so it is an outbound rule.
Is the FCC going to require AI disclosure on every call?
It has proposed it for outbound calls. In August 2024 the FCC proposed "requiring callers using AI-generated voice to, at the beginning of each call, clearly disclose to the called party that the call is using AI-generated technology." The proposed definition of an AI-generated call ends "over an outbound telephone call," and the FCC said it chose that wording to avoid burdening "widely used existing customer service technologies on inbound calls."
A proposal is not a rule. A greeting that already discloses the AI on outbound calls would meet it if it were adopted as written.
The script and the handoff
If the system records calls, announce it in the same greeting
If your AI receptionist records or transcribes calls, say so in the greeting too. Some states need every party's consent to record, and an announcement at the start is the simplest way to get it.
Federal law allows recording when "one of the parties to the communication has given prior consent," and the business is a party to its own calls. Several states go further. California makes it a crime to record a "confidential communication" by telephone "without the consent of all parties," punishable by a fine of up to $2,500 per violation or jail, and it counts a corporation as a person. Washington requires "the consent of all the participants", and treats consent as obtained when one party "has announced to all other parties ... that such communication or conversation is about to be recorded." If the call is recorded, Washington requires that the announcement itself be recorded too.
A ringing phone does not tell you which state the caller is in, so the safe default is to announce recording on every call, at the start, inside the recording.
Write the greeting in four parts, then test it on your own phone
Set the opening once, in your voice agent's instructions, and lock it so the model cannot paraphrase it away.
- Name the business exactly as customers know it, or as it is registered if the agent ever calls out.
- Say what is answering: "I'm the company's automated assistant" or "I'm an AI assistant." Avoid soft words like "virtual receptionist" that a caller could take for a remote employee.
- Announce the recording: "This call is recorded." Make sure the recording starts before this sentence plays.
- Offer a person: "Say 'team member' at any time to reach someone." Then ask how it can help.
- Call the line yourself from a cell phone and listen to the whole greeting. Check that the disclosure plays before the agent asks for a name or an address.
When a caller asks "am I talking to a robot?", say yes and offer a person
Write the answer into the agent's instructions word for word: "Yes, I'm an automated assistant. I can keep helping, or I can put you through to someone on the team." Never let it dodge, joke, or claim to be human.
Maine's test is whether a reasonable consumer could be misled into thinking a person is on the line, and Utah's law requires the AI to say it is AI when a consumer clearly asks. An agent that answers "I'm here to help!" to a direct question fails Utah's rule, and fails the caller everywhere else. Test it the way callers will ask: "Is this a real person?", "Are you a bot?", "Am I talking to a computer?"
If a call goes past what the AI should handle, transfer it
Give the agent a short list of moments when it stops and puts the call through, or takes a message for a callback the same day. These triggers are judgment, not law.
- The caller asks for a person. Once is enough. Do not make them repeat it.
- Anything urgent or unsafe. A gas smell, flooding, no heat in freezing weather, sparking. Tell the agent to say to call 911 when life is at risk, then transfer.
- A firm price or a promise. If the answer is not in the price list you gave it, it takes a message. Language models can give confident answers that are wrong, and a caller hears them as your answer.
- An upset caller or a complaint. An existing customer with a problem wants a person.
- Questions it fails twice. If it cannot understand or answer after two tries, hand off.
Judge the setup on booked jobs and on transfers that reached someone, not on how many calls the agent finished alone. An agent that handles every call but loses the urgent one is the wrong trade.
If you work in a licensed trade or call across state lines, check your own states
This is general information, not legal advice. State AI laws change quickly: Utah extended its 2024 law a year later. Before turning on an AI receptionist, read the current law in each state you take calls from, any rule from your licensing board, and your phone platform's recording settings. If the system will place calls, have the consent you collect reviewed before the first call goes out.