AI disclosure laws: what a business using an AI receptionist should know

Several U.S. states have laws that require disclosing an automated system in certain interactions. California's bot disclosure law requires that a bot used to communicate with a person online to encourage a purchase, or to influence a vote, must not mislead about being a bot, and a clear disclosure satisfies it. Utah's artificial intelligence law requires disclosure when a person asks whether they are interacting with AI, and up front in some regulated professions. Other states have passed or are considering broader rules. The simple answer for a business is to disclose everywhere, clearly, at the start. This post is a summary, not legal advice.

What the existing laws look like

California (Business and Professions Code, the "bot" law) makes it unlawful to use a bot to communicate with a person in California with the intent to mislead them about its artificial identity, in order to incentivize a sale or influence a vote. A clear, conspicuous disclosure that the communication is from a bot is a defence. An AI receptionist that books appointments is plausibly in scope, and the disclosure is easy.

Utah (the Artificial Intelligence Policy Act) requires that if a person asks whether they are talking to AI, the business must clearly disclose it; for regulated occupations (healthcare, law, and others licensed by the state) the disclosure must be made up front, before the interaction.

Other states have passed rules with disclosure elements, some effective in 2026, and more bills are moving. The details differ; the direction is the same.

Why "disclose everywhere" is the easy answer

Tracking which state each customer is in and which law applies is error-prone, and the cost of disclosing is nothing. A customer who knows they are talking to an assistant asks better questions and is not surprised later. And the customers who mind are the ones who would be angriest to find out afterwards. Disclose at the start of every conversation, on every channel, and in the voicemail greeting if calls are answered by AI.

How to word it

Plain and brief. "Hi, this is Northside Dental's assistant. I can book appointments and answer questions; a person is always available if you ask." Or on voice: "You have reached Northside Dental. I am an automated assistant." The word "assistant" is honest; "AI" is fine too. Avoid a human name presented as if a person were typing.

When a person takes over

Say so, because the reverse is also a disclosure question. "I am passing this to Dana, who will text you this morning." The customer should always know whether the current message came from software or a person. Some systems mark the difference visibly; at minimum, the handoff is announced.

Answering "are you a real person?"

Yes, immediately and plainly, every time, on every channel. This is the one question the AI must never dodge, deflect or answer cleverly. "No, I am Northside Dental's automated assistant. Would you like me to get a person?" Load it as a fixed response.

The regulated professions

Healthcare, legal, financial and other licensed services face stricter rules in some states, with disclosure required before the interaction rather than on request. If you are in one of those, disclose up front, in the first message, and keep the AI out of anything that touches the professional judgement the licence covers.

Keeping up

The laws are changing quickly. Ask your vendor what disclosures their product makes by default and whether you can edit them. Review annually, or when you expand into a new state. And when in doubt, disclose more, not less.

Frequently asked questions

Does disclosing lower response rates?

In our experience, no. Customers respond to fast, useful replies. Knowing the reply came from an assistant does not change that.

Do I need to disclose in appointment reminders?

Reminders are automated messages from the business, and customers understand that. Disclosure matters most in conversations where the customer might think a person is replying. Naming the business as the sender is the norm for reminders.

What about recorded calls answered by AI?

Call recording has its own consent rules by state. Disclose the AI and the recording together in the greeting.

What Frontiva does here

In Frontiva you describe each AI agent's personality and rules in plain language, so the disclosure wording is yours to set, and the playground lets you check it before a customer sees it. A request for a person is one of the readiness scenarios every agent is scored on, and escalation keywords hand a conversation over instantly. The AI agents page shows the settings; call recording consent basics covers the phone side.