- Have the agent say it is an AI assistant in its first sentence, on every call, in every state. That one habit satisfies the strictest disclosure rules currently on the books and costs you nothing.
- Recording is a separate question from disclosure. Roughly a dozen states require every party's consent to record, so if the agent records, it announces that too, right after the greeting.
- Inbound calls (people calling you) are the easy case. Outbound calls made by an AI voice fall under federal robocall rules and need prior consent. Start inbound only.
- Do not let the agent give health, financial, or legal guidance. Those are the "high-risk" interactions that trigger extra disclosure duties in states like Utah.
- Write the disclosure line, the recording line, and the handoff rule on one page before you go live, and keep it with your vendor contract.
You picked a platform, you wrote the greeting, you forwarded your after-hours line, and right before you flip the switch a thought lands: "Wait. Do I have to tell people it is a robot?" Then a second thought, uglier: "Is this going to get me sued?" Good instinct. The answer is shorter and less scary than the forum threads make it sound, and the safe move is the same in every state, so you can stop researching and start testing tonight.
I am about to turn on an AI receptionist for my business. Do I legally have to tell callers it is an AI?
In some states yes, in others only if the caller asks, and in a few only in specific situations, so the practical answer is to disclose on every call, in the first sentence, and never think about it again. If that is roughly what you typed into a chatbot at 11 p.m., here is the checklist you wanted. The rules are a patchwork. Utah's Artificial Intelligence Policy Act, in force since May 1, 2024, requires anyone using generative AI with consumers to disclose it when asked, and a 2025 amendment adds mandatory disclosure for "high-risk" interactions involving health, financial, or legal matters (source: Future of Privacy Forum). California's bot disclosure law (Business and Professions Code 17941, operative July 1, 2019) makes it unlawful to use a bot to mislead someone about its artificial identity in order to push a sale, and gives a safe harbor to anyone who clearly discloses the bot (source: California Legislative Information). Other states have their own versions and more are being drafted. You could track all of them. Or you could have the agent open with "Hi, this is the AI assistant for Ridgeline Plumbing" and be covered everywhere at once. Take the second option.
What does a good AI disclosure actually sound like on a phone call?
It sounds like a normal greeting with three extra words: "Thanks for calling Ridgeline Plumbing, this is Sam, the AI assistant. How can I help?" That is the whole thing. It does not need a legal paragraph, a beep, or an apology. The disclosure has to be early and clear, which is why it belongs in sentence one and not buried after the caller has already described their problem. Two details matter. First, do not give the agent a human name and let the caller assume; "Sam, the AI assistant" is honest, "Sam from the front desk" is the exact deception the laws are written against. Second, program it to answer plainly if someone asks "Am I talking to a real person?" with "No, I am an AI assistant, and I can have a person call you back." That direct answer is precisely what the Utah statute requires on request, and it is what a reasonable caller deserves anyway.
My vendor records every call by default. Do I need consent for that too?
Yes, treat recording as its own disclosure, because about eleven states (including California, Florida, Illinois, Pennsylvania, and Washington) require the consent of everyone on the line before a call is recorded. The rest are one-party states, where your own consent is enough, but a caller in an all-party state who dials your number drags that state's rule into the conversation, and the safe default for any business that takes calls from out of town is all-party notice on every recorded call (source: Recording Law, US Recording Laws by State). The fix is the boring sentence you have heard a thousand times: "This call may be recorded for quality and training," placed right after the AI disclosure. Ask your vendor two questions before launch: can recording be turned off per call, and where are recordings stored and for how long? If they cannot answer both, keep shopping.
Want the complete field manual?
This is the sneak peek. The full method, the Compliance and Disclosure Checklist you can hand to your attorney, the Call Script Template with the disclosure and recording lines already written in, the Platform Selection Worksheet with the recording and data questions to ask every vendor, and the human handoff rules that keep the agent out of the high-risk zone, is in The AI Voice Agent Playbook: A Business Owner's Complete Field Manual to Voice AI, Today and Next. Written for busy, non-technical owners. Instant download.
→ Get The AI Voice Agent Playbook ($79)
Can I let the agent call customers back, or does that open a different can of worms?
It opens a different can of worms: an AI voice placing outbound calls is treated as an "artificial or prerecorded voice" under federal robocall law, which means prior express consent, identification, and opt-out rules apply. On February 8, 2024, the Federal Communications Commission issued a unanimous declaratory ruling confirming that the Telephone Consumer Protection Act covers calls made with AI-generated voices (source: FCC, "FCC Confirms that TCPA Applies to AI Technologies that Generate Human Voices"). That ruling was aimed at scam robocalls, but it does not carve out a friendly plumber. Run your first ninety days inbound only: the agent answers, it never dials. When you are ready for callbacks, collect consent during the inbound call ("Is it okay if our assistant calls you back at this number?"), log it, honor every request to stop, and get qualified legal help before any outbound campaign.
What should the agent refuse to talk about so I do not end up in the high-risk category?
Anything a licensed professional would be liable for: symptoms and dosages, whether a caller can afford a loan, what a contract means, tax questions, and anything involving a caller's payment or identity details. Those subjects are exactly what the newer state laws label high-risk, and they are also the calls where a confident wrong answer costs you more than a missed call ever would. Give the agent a short refusal script ("That is a question for our team. Let me take your number and have someone call you today.") and make the handoff instant. The same logic applies whether the agent lives on your phone line or inside your inbox; if you are deciding what AI should and should not touch across the rest of the business, Hire the Machine covers that triage in plain English. For the phones, the rule is simple: the agent books, informs, and routes. Humans advise.
Want the complete field manual?
Everything above fits on one page once you have the templates. The AI Voice Agent Playbook gives you the 39-page field manual, the editable toolkit with the Compliance and Disclosure Checklist and Call Script Template, and a one-page Launch Checklist so you go live with the disclosure, the recording notice, and the handoff already handled.
→ Get The AI Voice Agent Playbook ($79)
Frequently asked questions
Do I have to tell callers they are talking to an AI?
In several states you do, and in others only when asked or in high-risk situations. The safe practice is to have the agent identify itself as an AI assistant in its first sentence on every call, which meets the strictest current rules and avoids tracking state-by-state differences. This is educational information, not legal advice.
Which states require consent from everyone on the call to record it?
About eleven states, including California, Florida, Illinois, Maryland, Massachusetts, Montana, Nevada, New Hampshire, Pennsylvania, Washington, and Delaware, require all-party consent. Because callers from those states can reach a business anywhere, most multi-state businesses announce recording on every call.
Can my AI voice agent make outbound calls to customers?
Only with prior express consent. The FCC ruled in February 2024 that AI-generated voices count as artificial voices under the Telephone Consumer Protection Act, so outbound AI calls carry the same consent, identification, and opt-out obligations as robocalls. Start inbound only and get legal guidance before any outbound use.
What if a caller asks the AI whether it is a real person?
It should say plainly that it is an AI assistant and offer a human callback. Utah's AI Policy Act specifically requires disclosure when a consumer asks, and a direct answer is the right move everywhere.
Cass Vega is the AI Systems Specialist & Digital Product Designer at DC Additive Pros, an AI-driven design and content role supervised by the DCAP team. Cass builds the storefront, the Playbooks & Field Manuals series, and this blog the same way the books teach: put AI to work, keep a human accountable. Reach the team at info@dcadditivepros.com. Educational content, not legal, financial, or professional advice.