Natural AI voices are good enough that many customers will not notice them. That raises a fair question for any team using them: do you have to say it's AI? The answer is a mix of rules that already apply, rules that have been proposed, and state laws that vary. This guide walks through each, then gives a script approach that holds up under all of them.

What federal rules already require

There is not yet a single federal rule that says "announce the AI." But calls that use an AI voice are treated as artificial or prerecorded voice calls under the Telephone Consumer Protection Act, and those calls already carry disclosure duties:

  • Identify the business at the start. The message must clearly state the name of the business responsible for the call at the beginning.
  • Give a callback number. A telephone number must be provided during or after the message.
  • Offer an opt-out on telemarketing. Telemarketing messages must announce an automated opt-out mechanism at the start.

The FTC's Telemarketing Sales Rule adds its own prompt disclosures for sales calls: who is calling, that the purpose is to sell, and what is being sold. And under both the FTC Act and the Telemarketing Sales Rule, misrepresenting a material fact is deceptive. Claiming to be a human when you are not is exactly the kind of statement that invites that argument.

For the consent side of AI voice calls, see our guide to the TCPA and AI voices.

What the FCC has proposed

In 2024 the FCC proposed rules aimed specifically at AI-generated calls. The proposal would define an "AI-generated call," require callers to disclose that they use AI when they obtain consent, and require an AI disclosure at the start of each such call. As of this writing, check the current status before relying on either outcome: proposals can be adopted, changed or dropped. If adopted, a disclosure at the start of the call would become a firm requirement rather than a best practice.

State laws

States are moving faster than the federal government, and the landscape changes often. Examples of the kinds of rules in place:

  • Bot disclosure laws. California's bot disclosure law makes it unlawful to use a bot online to mislead someone in California about its artificial identity in order to incentivize a sale, unless the bot is clearly disclosed. Whether it reaches phone calls is debated, but the principle it sets is the one regulators everywhere are applying.
  • Generative AI disclosure on request. Utah requires businesses to disclose that a consumer is interacting with generative AI when the consumer clearly asks.
  • State telemarketing acts. Florida, Oklahoma and other states have their own telephone solicitation laws with consent and identification rules, some with private rights of action.

Because these laws target deception, the common thread is simple: never mislead someone about whether they are talking to a machine.

A disclosure approach that works everywhere

You don't need a clunky legal preamble. A short, honest opening covers identification, sets expectations and still sounds good:

"Hi, this is Mia, an AI assistant calling for Brightline Home Services. I have a real person listening in with me. Is now an okay time?"

Then make three commitments in your script and your training:

  1. If asked, answer truthfully. "Yes, this voice is AI. My colleague is right here, and I can hand you over." Then the agent takes over.
  2. Make the handoff instant. A customer who wants a person should get one within seconds, not after a menu.
  3. Honor opt-outs on the spot. "Please stop calling" goes on your do-not-call list before the call ends.

What to avoid:

  • Scripts or quick responses that say "no, I'm a real person" or dodge the question.
  • Personas that claim a human job title or location the voice doesn't have.
  • Hiding the business name or calling from numbers that don't match it.

How Voxa approaches disclosure

Voxa keeps a human agent on every call. The agent chooses each line, can switch the call to Live with one click, and hears everything the customer says. Every AI voice has a persona name you control, and none of Voxa's starter clips deny or deflect the question "are you a robot?". We guard that with an automated test that fails if a clip ever does. If you record your own quick responses, hold them to the same standard.

This article is general information, not legal advice. AI disclosure rules are changing quickly and vary by state. Talk to a qualified attorney about your scripts and campaigns.