Free Fix Guide 15 of 27

Tell Callers When They're Talking to AI

One sentence. It's the law in a growing number of states, and it's the right thing regardless.

Signal this fixes
ai_disclosure_present
Time
15 minutes
Cost
Free
Difficulty
Easy

What we measured

If your phone is answered by AI, we checked whether it identified itself as AI unprompted, and how quickly.

Why it matters

If a human answers your phone, this doesn't apply to you and you'll score full points automatically.

If an AI answers your phone, you have two reasons to disclose it.

The legal one. A growing number of states now require it. Maine's law (10 M.R.S. § 1500-DD, effective September 2025) is the broadest: clear, conspicuous, proactive disclosure whenever the bot may mislead a reasonable consumer - any trade or commerce, no size threshold, and no requirement that anyone was actually misled. New Jersey requires proactive disclosure at the start of the interaction for sales and advertising. Utah's law provides a safe harbor if the AI discloses itself clearly at the outset.

California's B.O.T. Act provides a similar safe harbor. Colorado, Oregon, Washington, Nebraska, Idaho and others have laws in various stages.

The practical one. Customers hate discovering it mid-call far more than they mind being told upfront. The complaint is never "they told me it was AI." It's always "I thought I was talking to a person." The good news: the union of every state requirement collapses into one cheap behaviour.

Fix it yourself

Disclose in the first sentence

Not buried, not at the end, not only if asked. First sentence, before anything else.

Working script: "Hi, thanks for calling [Business Name] - I'm an AI assistant. I can book appointments, answer questions about our services, or get you to a person. What can I help with?" That single line satisfies Maine, New Jersey, Utah's safe harbor, Colorado's consumer-interaction notice, and California's safe harbor simultaneously. Nobody has ever been penalized for over-disclosing.

Step 1

Offer the human path in the same breath

"...or get you to a person" does a lot of work. It reassures the caller who was about to hang up, and it satisfies the escalation requirement in every framework that has one.

Then honour it instantly - see Guarantee a Customer Can Always Reach a Human.

Step 2

Never let the AI claim to be a person3

Common mistakes

Giving the AI a human name presented as staff. This is the exact thing the statutes target.

Disclosing only when asked. Maine, New Jersey and others require it proactively.

Two long separate prompts. You'll lose callers before the menu. Combine them.

Assuming this is settled law. It's changing fast. Date-stamp your policy and review it twice a year.

This fix is free. You can earn and keep any Agenarys Seal tier - including Platinum - without paying Agenarys anything beyond the audit itself. If you would rather have help doing it, we sell that. We never sell passage.

No human first names presented as staff. No "let me check with my manager." No pretending to type. If a caller asks "am I talking to a real person?", the answer is an immediate, plain no.

This is where the legal exposure actually lives - the statutes target intent to mislead, and a bot with a fake human identity is the paradigm case.

If you're in healthcare, disclose twice

California's AB 3030 (effective January 2025) applies to health facilities, clinics, and physician practices using generative AI for patient communications about clinical information. For phone specifically, it requires the disclaimer verbally at the start AND at the end of the interaction, plus clear instructions for contacting a human provider.

It doesn't apply if a licensed human reviews the communication first, and it doesn't cover purely administrative content like scheduling or billing. If you're a medical or dental practice, take the safe route: disclose at both ends and always name the human path.

Step 4

Handle recording consent separately

Recording consent and AI disclosure are independent legal obligations. Satisfying one does not satisfy the other.

Federal law is one-party consent, but these states require all parties to consent for phone recording: California, Connecticut, Delaware, Florida, Illinois, Maryland, Massachusetts, Montana, Nevada, New Hampshire, Oregon, Pennsylvania, Washington. Michigan is contested - announce anyway.

Interstate calls are legally unsettled. If any of your callers are out of state, behave as if you're in an all-party state at all times.

If you record, combine both into one opening line rather than two separate prompts: "Hi, thanks for calling [Business] - I'm an AI assistant and this call may be recorded. I can book you in, answer

questions, or get you to a person."

And you must have a real way to stop recording if the caller objects. An announcement with no opt-out is weak consent.

Step 5

Your AI identifies itself as AI in the first sentence.☐

The human path is offered in the same greeting.☐

The AI never claims to be a person, even when asked directly.☐

Healthcare: disclosure at start AND end, plus human-contact instructions.☐

If you record, that's announced too, with a way to opt out.☐

Sources and further reading

Maine 10 M.R.S. § 1500-DD (eff. September 2025)

New Jersey N.J.S.A. 56:8-18.2

Utah AI Policy Act (SB 149, amended by SB 226)

California B.O.T. Act, Bus. & Prof. Code §§ 17940-17943

California AB 3030 (eff. January 2025)

FCC Declaratory Ruling 24-17, February 2024

This guide is general information, not legal advice. Laws change frequently - confirm current requirements in your states with counsel.

Agenarys · Operational Intelligence Audit · Standard v2026.1 Issued July 2026 · agenarys.com/guides/ai-disclosure-script Content reviewed July 28, 2026 | Next review: January 2027 (every 6 months)