Do You Have to Tell Callers It's an AI? (2026 Disclosure Rules)

The Short Answer

For an AI receptionist answering calls that customers place to you, there is currently no blanket United States requirement to announce that the caller is speaking to an AI.

The rules people cite in this conversation are real, but they were written for two different things: outbound robocalls, and deceptive online bots. An inbound receptionist is neither. The one situation where honesty is clearly required, and the one you should build your policy around anyway, is when a caller asks directly.

The rest of this explains where each rule actually applies, because getting this wrong in either direction costs you. Over-disclosing costs you conversions. Under-disclosing on outbound calls costs you far more.

The Federal Layer: TCPA and the FCC

On February 8, 2024, the FCC adopted a Declaratory Ruling (FCC 24-17, Docket 23-362) confirming that an AI-generated voice counts as an "artificial or prerecorded voice" under the Telephone Consumer Protection Act.

This is the single most-cited authority in articles about AI receptionists, and it is almost always applied incorrectly. The TCPA governs calls placed to consumers. Its restriction is on making robocalls without prior express consent. It is a rule about outbound calling.

When a customer dials your business and your AI picks up, no robocall has occurred. Nobody placed a call to that consumer. The TCPA's consent requirement has nothing to attach to.

Where this absolutely does apply to you: outbound. If you use AI voice to call leads back, run reactivation campaigns, chase estimates, or dial a purchased list, you are squarely inside the TCPA and you need prior express consent. Statutory damages run per violation and there is a private right of action, which is why outbound AI calling is a genuinely higher-risk activity than inbound answering. Treat them as two different compliance problems, because legally they are.

California: Two Laws, Neither Says What You Think

AB 2905 is about outbound autodialers

AB 2905 amended California Public Utilities Code Section 2874. It is widely described as "California requires AI disclosure on calls." What it actually does is add a requirement to the announcement that an automatic dialing-announcing device must play before a prerecorded message: that announcement must now inform the person called if the prerecorded message uses an artificial voice.

The operative language governs calls placed through an automatic dialing-announcing device. It describes calls going out, not calls coming in. An AI receptionist answering your business line is not an autodialer and is not placing calls.

SB 1001 is about deceptive online bots

California's Bot Disclosure law lives at Business and Professions Code Section 17941. It makes it unlawful to use a bot to communicate with a person in California online, with intent to mislead about the bot's artificial identity, in order to knowingly deceive that person about the content of the communication to incentivize a purchase or influence a vote.

Three things matter in that sentence. It says online. It requires intent to mislead. And disclosure operates as a safe harbor, not a standalone mandate: a person is not liable if they disclose that it is a bot, and the disclosure must be clear, conspicuous, and reasonably designed to inform.

An AI receptionist that answers honestly and is not trying to deceive anyone about what it is does not engage the conduct this statute targets, and whether a voice call is "online" at all is an open question.

Utah: Disclose When Asked

Utah's AI Policy Act is the closest thing in the US to a rule directly on point, and it got narrower, not broader.

As originally enacted, it required disclosure of generative AI use when a consumer asked. SB 226, effective May 7, 2025, narrowed the general obligation further, tying it to instances where a consumer or supplier directly asks whether they are interacting with AI. It also added a statutory safe harbor for tools that make clear and conspicuous disclosures at the outset of and during the interaction.

Proactive disclosure, without being asked, is required in higher-risk contexts: interactions involving health, financial, or biometric data, and interactions providing advice on financial, legal, or healthcare matters.

For most trades and home services businesses, that means: answer honestly if asked. If you run a medical or health-adjacent practice or a legal practice, look harder at the proactive category.

So Where Does That Leave You?

A practical policy that holds up across the rules above:

Answer honestly when asked. Always, everywhere. This is the one obligation that clearly exists somewhere and is defensible everywhere. Your AI should never claim to be a human when a caller asks a direct question. Build it that way regardless of your state.

Treat outbound as a separate, stricter problem. Inbound answering and outbound AI calling are not the same compliance question. Get consent right before you dial anyone.

Raise your standard if you handle health, financial, or legal matters. Proactive disclosure is cheap insurance in those categories and is affirmatively required in some.

Check your own state. This area is moving. What is written here reflects the statutes as published at the time of writing, and more states are legislating.

The Question Underneath the Legal One

Most people asking whether they have to disclose are really asking something else: will my customers hang up?

It is a fair worry and it deserves a straight answer rather than reassurance. Some callers do react badly to learning they are talking to a machine. You will find a range of figures quoted for how many, most of them published by companies with an interest in the number being high or low, and we are not going to add another unsourced statistic to that pile.

What we can tell you is what actually moves the outcome, based on running these systems:

Voice quality does most of the work.The reaction people have is usually not to the word "AI," it is to a voice that sounds like a phone tree. Natural pacing, interruption handling, and not sounding scripted matter more than any disclosure wording.

Escalation speed matters more than escalation existing. Every AI receptionist claims it can transfer to a human. What separates them is how fast it happens when someone asks, and whether the caller has to fight for it. A fast, gracious handoff converts a frustrated caller. A slow one loses them.

The real comparison is not AI versus a human. It is AI versus your voicemail at 7pm on a Saturday. The relevant question is not whether some callers would prefer a person, it is what happens to the calls nobody is currently answering. Our missed-call calculator puts a number on that side of the ledger.

How We Handle It

Every BotPhone AI receptionist is configured to answer honestly if a caller asks whether they are speaking with a person, and to hand off to a human quickly when someone wants one. Because we build and manage the agent rather than handing you a dashboard, disclosure wording and escalation behaviour are part of the setup we do for you, and we adjust them if your requirements change.

If you operate in a category where proactive disclosure is the safer call, tell us on the discovery call and we will build it that way from the start.

Frequently Asked Questions

Does my AI receptionist have to say "I am an AI" when it picks up?

For inbound calls, generally no under current US law, with the higher-risk categories in Utah as the notable exception. It must not deny it when asked.

Is it different if I use AI to call people back?

Yes, substantially. Outbound AI voice calls fall under the TCPA following the FCC's 2024 ruling, which requires prior express consent. This is the highest-risk use and should be set up carefully.

What about recording calls?

That is a separate body of law from AI disclosure. Some states require all-party consent to record. If your AI receptionist records or transcribes, your recording notice obligations apply the same way they would with a human answering service.

Does HIPAA change any of this?

HIPAA governs how protected health information is handled, not whether you disclose that a bot is answering. Both matter for a medical practice, and they are separate requirements. We cover the health side in our guide to HIPAA-compliant AI answering.