Must an AI Phone Agent Say It's AI? Disclosure Laws

US law doesn't yet require every AI phone agent to announce it's AI, but several rules push in that direction. California bans bots that mislead people about being artificial to sell something or influence a vote, Utah requires disclosure when a consumer clearly asks (and up front for certain high-risk interactions), and the FCC has proposed AI disclosure on AI-generated calls. The practical answer: disclose at the start of the call and always answer honestly if asked.

General information, not legal advice. AI law is changing quickly; check the states where your callers are.

The rules that exist today

Rule	What it requires	Applies to
California Bus. & Prof. Code §17941	Don't use a bot to mislead someone about its artificial identity to incentivize a sale or influence a vote; clear disclosure avoids liability	Online bots communicating with people in California; commonly treated as relevant to voice agents
Utah AI Policy Act (as amended by SB 226, 2025)	Disclose generative AI use when a consumer clearly asks; disclose proactively in high-risk interactions (e.g., health, financial or biometric data, or advice relied on for significant decisions)	Businesses using generative AI with Utah consumers
TCPA identification rule (47 CFR 64.1200(b))	Artificial-voice calls must identify the business at the start	Outbound AI calls
FCC 2024 proposal (pending)	Would require AI-generated calls and texts to disclose AI use	Not final

Utah's amended law also gives a safe harbor to businesses whose AI discloses it is AI at the outset and throughout the interaction.

California's law is written for bots "online," and whether it reaches phone agents hasn't been settled by courts. Since a clear disclosure is all it takes to comply, there's little reason to test it.

Inbound vs outbound

Outbound AI calls carry the most scrutiny. They're already regulated as artificial-voice calls, need consent, and must identify the business. Adding "an AI assistant" to the opening line costs a few words. See AI outbound calls and the TCPA.

Inbound AI receptionists are lower risk because the caller chose to call you. Still, if the agent is selling (quoting, booking paid work), California's misleading-bot rule and Utah's disclosure-on-request rule can apply. Disclose in the greeting or, at minimum, answer honestly when asked.

Disclosure wording that works

Good disclosures are short, early and plain.

Inbound greeting:

"Thanks for calling Lakeside Plumbing. This is Ava, Lakeside's AI assistant. I can answer questions, take a message, or get you to the team. What can I help with?"

Outbound opening:

"Hi Dana, this is Mia, an AI assistant calling from Lakeside Plumbing about your appointment tomorrow."

When a caller asks "Am I talking to a real person?":

"I'm an AI assistant for Lakeside Plumbing. I can help with most questions, or I can connect you with someone on the team."

Avoid wording that implies a human: "Let me check with my manager," "I'm at the front desk today," or fake typing sounds. These are the kinds of details that make a disclosure-on-request law matter.

More greeting options: AI receptionist greeting examples.

Does disclosure hurt results?

Some callers prefer a person, and some hang up. Two things reduce that:

1. Offer the exit early. "...or I can get you to the team" tells callers a human is available.
2. Be useful fast. If the agent answers the actual question in the first 20 seconds, the AI label stops mattering.

Measure it rather than guess: compare hang-up rates in the first 15 seconds before and after changing the greeting. See AI agent performance metrics.

Special cases

• Health, finance, legal advice: Utah treats these as high-risk interactions that need proactive disclosure. Other regulators expect it too. Disclose up front and keep the agent to admin tasks.
• Cloned voices: using AI to imitate a specific real person's voice raises separate consent and right-of-publicity issues. Use stock voices, or a voice clone only of someone who agreed in writing.
• Recording notices: AI disclosure doesn't replace a recording notice. See recording AI phone calls.

Disclosure beyond the greeting

A one-time disclosure at the start of a call is the floor. A few other moments deserve it too:

• Before a transfer: "I'm going to connect you with Jordan on our team now." This makes the switch from AI to human obvious.
• When taking sensitive details: if a caller starts sharing medical or financial information, a reminder that they're speaking with an AI assistant, and that a person will follow up, helps them decide what to say. See keeping sensitive data off AI calls.
• In voicemails left by AI: the message should name the business and say it's an automated assistant, so the listener isn't misled later.
• In written follow-ups: emails or texts sent by the AI should be signed as the assistant, not as a staff member.

Telling customers before they call

Disclosure can start before the phone rings. Many businesses add a line to their website contact page or booking confirmation: "Calls after hours are answered by our AI assistant, who can take messages and book callbacks." It sets expectations, reduces surprise hang-ups, and shows good faith if a disclosure question ever comes up.

Your privacy policy should also mention that calls may be handled by an AI assistant and that AI providers process call audio and transcripts on your behalf. See AI receptionist privacy.

Watching for new laws

Several states have introduced or passed AI transparency bills since 2024, and some have been delayed or rewritten before taking effect. A practical approach for a small business: disclose up front everywhere, which satisfies the strictest version of most of these rules, and review your setup once or twice a year.

A disclosure checklist

• The greeting names the business.
• The greeting or first sentence says it's an AI assistant.
• The agent answers "are you a person?" truthfully, every time.
• The agent never claims to be human or uses human-only cover stories.
• Outbound calls disclose AI in the first sentence.
• AI texts are signed or identify the business.
• Someone reviews a sample of calls monthly for disclosure.

How Callata handles disclosure

Every Callata AI agent is instructed to say it's an AI assistant for your business when a caller asks whether it's a person. The default outbound opening line is "Hi [name], this is [agent], an AI assistant calling from [your business]." Inbound greetings are yours to write; the presets introduce the agent by name, and adding "the AI assistant" is a one-word edit. Callata's terms state that AI agents identify themselves as AI assistants.

Callata Office is $99 per month for five users ($20 per extra user), and AI agents run on prepaid minutes at $0.25 per minute. Build your agent in Callata.