Do AI Receptionists Have to Say They're AI? The 2026 Disclosure Rules for Your Phone Line

Short answer: In most of the U.S., no law yet forces an inbound AI receptionist to announce that it's AI. But Maine already bans AI that could pass for human in a sale unless the caller is clearly told, Utah requires an honest answer whenever a caller asks, and federal rules already make outbound AI calls name the business in the first breath. Florida's attorney general wants undisclosed AI voices used to take money treated as fraud. Every agent Wrk Less builds says it's an AI assistant in its opening line, and we think yours should too.

This is not legal advice. Have your attorney review your scripts, especially for outbound calling.

We hear this question from owners all the time: "Do we have to tell people it's a bot?" Owners worry the disclosure will make callers hang up. Then they worry that skipping it will get them sued. Here's what the rules actually say as of October 9, 2026, and how we wire disclosure into a voice agent so it holds up.

Is there a federal law that says an AI receptionist must disclose it's AI?

Not yet. In August 2024 the FCC proposed rules that would require callers using AI-generated voices to disclose that at the start of each call. The proposal was published in the Federal Register on September 10, 2024. It has never been adopted. We checked the current text of the FCC's calling rules, 47 CFR 64.1200, on October 9, 2026, and there's still no AI-disclosure requirement in it.

What is settled federally matters more for outbound calls:

• AI voices count as "artificial" voices. The FCC's February 2024 declaratory ruling (FCC 24-17) put AI-generated voices under the Telephone Consumer Protection Act. An AI agent that calls people needs the same consent a prerecorded robocall does.
• Name the business first. Under 64.1200(b)(1), every artificial or prerecorded voice message must "at the beginning of the message, state clearly the identity of the business" responsible for the call, using its registered business name. It must also give a callback number. Telemarketing calls to residential lines need an automated voice or keypress opt-out within two seconds of that identification.
• Opt-out wording is about to change. On September 30, 2026, the FCC adopted FCC 26-67, which lets callers designate an exclusive opt-out method if it's clearly disclosed on the call. We covered that in detail in our breakdown of the new FCC opt-out rules. As of October 8, the order hadn't been published in the Federal Register, so its 30-day clock hasn't started.

Notice the gap. Federal law makes an outbound AI agent say who is calling. It doesn't yet make the agent say what is calling. The states are filling that gap.

Which states already require AI disclosure on a phone call?

Two state laws reach an AI receptionist answering your phone today.

Maine. 10 M.R.S. §1500-DD defines an "artificial intelligence chatbot" as software that simulates human conversation "through textual or aural communications." Aural means voice. A business may not use one in trade "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 clearly notified. A violation counts as a violation of the Maine Unfair Trade Practices Act. Today's best voice models sound human, which makes that "may mislead" test easy to fail.

Utah. Under Utah Code §13-77-103, a business using generative AI in a consumer transaction must tell the person they're talking to AI "if the individual asks or otherwise prompts" about it with a clear request. Licensed occupations such as medicine, law and accounting have a stricter rule: for high-risk interactions they must disclose "verbally at the start of a verbal interaction." The safe harbor in §13-77-104 protects businesses whose AI discloses at the outset and throughout the interaction that it is AI, not human, or "an artificial intelligence assistant."

Other headline laws mostly miss the phone line. California's bot-disclosure law defines "online" as websites and apps, so calls fall outside it, and the newer companion-chatbot laws in California and New York exclude customer-service bots.

What's changing in Florida and other states?

On September 17, 2026, Florida Attorney General James Uthmeier proposed the Florida Anti-Spam Communications Act. It would treat "using an undisclosed AI voice to obtain money" as communications fraud, according to News4JAX's summary of the release. It would also tighten the state's telemarketing consent law so a STOP request is honored within days rather than weeks. As of the announcement, no bill had been filed, and it still has to pass the Legislature.

The proposal targets scammers, not plumbers with an after-hours line. But we're an Orlando shop, and here's what we take from it: Florida's top prosecutor now names undisclosed AI voices in the same breath as fraud, and the 2027 session will debate it. That's not a phrase you want anywhere near your phone system.

Michigan is moving the same way. House Bills 6393 and 6394, introduced in September 2026, would fine AI robocalls $500 per call.

AI disclosure rules at a glance (as of October 9, 2026)

Rule	Status	What it requires	Reaches an inbound AI receptionist?
FCC 24-84 AI-call disclosure	Proposed in 2024, never adopted	Disclose AI-generated voice at call start	No, not in force
47 CFR 64.1200(b)	In force	Outbound artificial-voice calls name the business first, give a callback number and, for telemarketing, an opt-out	No, outbound only
Maine 10 M.R.S. §1500-DD	In force	Clear notice if AI could pass for human in trade	Yes
Utah Code §13-77-103	In force	Honest answer when asked; licensed occupations disclose up front	Yes
Florida Anti-Spam Communications Act	Proposed Sept. 17, 2026, not filed	Undisclosed AI voice used to obtain money treated as fraud	Would reach outbound scams
Michigan HB 6393/6394	Introduced Sept. 2026	Per-call fines for AI robocalls	Aimed at robocalls

Will disclosing it's AI make callers hang up?

That's the fear, and it's backwards. The call that goes wrong is the one where the caller figures it out halfway through and feels tricked. A caller who hears "I'm the AI assistant, I can book you right now or get you to a person" in the first five seconds knows what they're dealing with and has no reason to feel tricked.

The hang-ups worth worrying about come from somewhere else: slow, robotic agents that can't do anything useful. Fix latency, give the agent real tools like calendar booking and CRM lookup, and offer a human handoff. Disclosure stops being the problem. Our AI receptionist buying guide covers how to test for that before you sign anything.

There's also a newer wrinkle. AI assistants are now calling businesses on behalf of consumers, and some front desks hang up the moment they hear a bot. Being upfront on your own line is the consistent position.

How does Wrk Less build disclosure into a voice agent?

We don't leave disclosure to the language model's judgment. Here's the standard we build to, and the one we'd hold any vendor to:

• Hard-code the opening line. The disclosure goes in the agent's fixed welcome message, not the prompt. The pattern is business name, "I'm the AI assistant," then a recording notice when the call is recorded. Florida requires all-party consent to record, which we covered in our real-estate after-hours post.
• Never deny being AI. The prompt has a non-negotiable rule: if someone asks "Is this a real person?" in any phrasing, the agent says it's an AI assistant and offers a transfer or callback.
• Test the question before launch. Our go-live test calls include "Am I talking to a robot?", "Are you human?" and "Is this a recording?" The agent has to answer honestly every time.
• Log which script ran. The post-call webhook writes the script version and disclosure flag to the CRM with the transcript, so you can prove what callers heard. Our Retell integrations and webhooks page shows how that handoff works.
• Outbound gets extra lines. Callbacks and reminders state the registered business name first, give a callback number and offer a spoken or keypress opt-out. Once FCC 26-67 takes effect, the agent also says how to opt out if you've chosen an exclusive method.
• One script, every state. We write to the strictest rule in force, which is Utah's safe harbor, so a caller from Maine or Utah gets the same compliant greeting as a caller from Orlando.

If you're on a platform tool instead of a custom build, look for the same controls. HighLevel's Voice AI, for example, added AI disclosure and verbal opt-out settings to its Agent Builder in September, which we compared in our Jobber vs. GoHighLevel voice AI post.

Bottom line

Disclosure costs about four seconds of airtime. Skipping it puts you on the wrong side of Maine and Utah today, and maybe Florida and Michigan tomorrow. It also makes a federal rule harder to live with if the FCC ever finalizes its 2024 proposal. Say it up front, answer honestly when asked, and keep a log that proves both.

Want us to check your greeting?

We'll listen to how your current AI receptionist or voice agent opens a call, test whether it answers "Are you a real person?" honestly, and fix the script, the logging and the outbound opt-outs. See how we set up Retell voice agents, or skip ahead and book your AI systems assessment.