Do I Have to Tell Callers They're Talking to an AI? 2026
You set up an AI receptionist, it answers your line, and then someone — a customer, your partner, a competitor in a Facebook group — asks the uncomfortable question. Are you allowed to do that without telling people? Do you have to announce it? Is there a fine attached to getting this wrong?
It is a fair question and the answers circulating online are mostly wrong, because they collapse two completely different legal situations into one. NoMissedCall makes an AI receptionist, so we have an obvious interest in the answer; read this with that in mind and check the links, all of which go to primary sources — the regulators' own documents, not other vendors' blogs. Nothing here is legal advice, and where the answer genuinely depends on your state or your industry we say so rather than guessing.
The distinction that decides almost everything
Nearly every rule that people cite in this argument governs calls a business places to consumers. Almost none of them govern calls consumers place to a business.
That is the whole thing. Robocall law exists because unwanted calls arrive at your dinner table uninvited. When someone dials your published business number, they initiated the contact, they want to reach you, and the entire consent architecture of US telemarketing law simply does not engage. An AI receptionist answering your own line is legally a very different animal from an AI voice dialling out to a list, and once you separate those two cases the picture becomes fairly clear.
What actually changed on 2 August 2026
Five days ago the transparency obligations in Article 50 of the EU AI Act became applicable. This is the first binding rule anywhere that speaks directly to AI systems talking to people, and it is why the question is suddenly live again.
The core duty is short. For AI systems intended to interact directly with people, providers must ensure that individuals are informed they are interacting with an AI system, unless that is obvious from the circumstances — and the notification has to come at the first interaction, in a clear and distinguishable manner, meeting accessibility requirements. That is the European Commission's own summary in its FAQ on Article 50, published alongside its guidelines on the transparency obligations.
Two details matter for anyone reading this in a panic. First, for interactive systems the obligation sits on the provider — the company that develops the system and places it on the market — not on the small business that switches it on. Deployer-side duties in Article 50 attach to emotion-recognition systems, biometric categorisation and deepfake labelling, none of which describe a receptionist that books appointments. Second, the limited grace period that runs to 2 December 2026 applies to the separate machine-readable marking obligation for synthetic content on systems already on the market, not to the interaction-disclosure duty, which is live now. Both points come from the Commission's FAQ linked above.
The penalties are real: non-compliance with the transparency obligations can draw fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher, per Cooley's 3 August 2026 analysis. Those numbers are aimed at the model and platform layer, not at a two-van HVAC company.
Does the EU rule reach a business in Ohio?
Probably not, but the answer is not automatically no. The AI Act's territorial scope extends to providers and deployers established outside the EU where the output of the AI system is used in the Union — the Commission's FAQ states plainly that compliance is mandatory regardless of where the company is established. A contractor whose callers are all within a 40-mile radius of Columbus is outside that. A business that sells into Europe, or one whose AI line regularly takes calls from EU numbers, is in genuinely uncertain territory and should ask a lawyer, not a blog.
The more useful way to read the EU rule from the US is as a preview. Article 50 is the first serious attempt to write down what "tell people it's a bot" means in operational terms, and the language it lands on — inform at first interaction, unless obvious, clearly and distinguishably — is the language other regulators will borrow.
The US federal picture: the TCPA covers calls you make
On 8 February 2024 the FCC adopted a declaratory ruling confirming that AI technologies generating human voices, including voice cloning, fall within the TCPA's restriction on using an "artificial or prerecorded voice" — the full text is FCC 24-17, in CG Docket 23-362, and the Commission's own summary page is worth reading before you trust a headline about it. The headlines at the time said the FCC had made AI voices "illegal." It did not. As Mayer Brown noted at the time, the ruling makes AI voice calls subject to the same consent, identification and opt-out rules as any other artificial or prerecorded voice call. Those rules govern calls placed to consumers.
The FCC did propose going further. On 7 August 2024 it adopted a Notice of Proposed Rulemaking, FCC 24-84, proposing to define an "AI-generated call," require callers to disclose AI use at the start of a call, and tie consent language to it — the Federal Register notice set comments for 10 October 2024. As of August 2026 that proposal has not been adopted as a final rule. Even if it were, its in-call disclosure requirement is written for outbound calls to consumers.
Which state rules actually reach an inbound business line
Four states come up constantly in this conversation. Only one of them clearly reaches a small business answering its own phone.
Utah is the one to know. The Utah Artificial Intelligence Policy Act requires disclosure that a consumer is interacting with generative AI, and the 2025 amendments in SB 226 narrowed that duty to cases where the consumer makes a clear and unambiguous request — plus "high-risk" interactions involving health, financial or biometric data or advice on financial, legal or healthcare matters. Crucially, the amendments added a safe harbour: a business is shielded from disclosure enforcement if the AI itself clearly and conspicuously discloses that it is not human at the outset of and throughout the interaction. Davis Polk's analysis and the Future of Privacy Forum's overview both walk through it. Read that safe harbour again, because it is the single most actionable sentence in this entire article: a one-line disclosure in your greeting is the compliant path, and it is cheaper than working out whether any of the exceptions apply to you.
California is widely misquoted here. SB 1001, the B.O.T. Act, makes it unlawful to use an undisclosed bot to incentivise a sale — but it defines a bot as an automated online account and defines "online" as a public-facing website, web application or digital application, per the bill text and Perkins Coie's read of it. Telephone calls are not online accounts. And SB 243, effective 1 January 2026, covers companion chatbots — systems built for sustained social or emotional interaction — as the bill's own text makes clear. An assistant that quotes your hourly rate and books a Tuesday slot is not a companion chatbot.
Texas gets cited for two laws, neither of which does what people think. SB 140, effective 1 September 2025, broadened the state's telephone-solicitation statute to cover text and multimedia marketing, added registration duties and expanded the private right of action — Paul Hastings has the detail. It regulates how you contact consumers, not how you answer them, and it contains no AI-voice disclosure rule. The Texas Responsible AI Governance Act, HB 149, effective 1 January 2026, does impose disclosure — on government agencies, before or at the point of interaction, and on healthcare providers using AI in diagnosis or treatment. Norton Rose Fulbright and Holland & Knight both cover the sector split. If you run a medical or dental practice, that healthcare provision is worth a proper read.
Colorado is the cautionary tale about planning around pending law. Its 2024 AI Act was delayed, then challenged, then repealed outright: Governor Polis signed SB 26-189 on 14 May 2026 replacing it with a narrower disclosure framework for automated decision-making technology, effective 1 January 2027, per McDermott's summary. Two years of compliance anxiety about a law that no longer exists.
What applies to an AI answering your own number
| Rule | Reaches an inbound AI receptionist? | Source |
|---|---|---|
| TCPA / FCC 24-17 (Feb 2024) | No — governs outbound artificial-voice calls to consumers | FCC 24-17 |
| FCC 24-84 AI-call rules | No — still a proposal, and written for outbound calls | FCC 24-84 |
| EU AI Act Article 50 | Yes if outputs are used in the EU; duty sits on the provider | Commission FAQ |
| Utah AI Policy Act (as amended by SB 226) | Yes — on clear request, or in high-risk interactions; safe harbour for upfront disclosure | Davis Polk |
| California SB 1001 (B.O.T. Act) | No — limited to online accounts, not phone calls | Bill text |
| California SB 243 | No — companion chatbots only | Bill text |
| Texas SB 140 | No — outbound telephone solicitation and marketing texts | Paul Hastings |
| Texas HB 149 (TRAIGA) | Only if you are a government agency, or a healthcare provider using AI in diagnosis or treatment | Norton Rose Fulbright |
One more thing this table does not cover: call recording. If your AI receptionist records or transcribes calls — most do, including ours — you are in two-party-consent territory in a dozen states, which is an older and better-settled body of law than anything above, and a much likelier source of trouble. That deserves its own article, and it is on our list.
The better reason to disclose, which has nothing to do with law
Set the statutes aside. The commercial case for one honest sentence at the top of the call is stronger than the legal one.
Consumer trust in the voice channel is deteriorating badly, and AI is why. In Hiya's State of the Call 2026, a survey of more than 12,000 consumers across six countries, one in four Americans reported receiving a deepfake voice call in the previous twelve months, and a further 24% said they were not confident they could tell the difference. Nearly half the adult population is now either primed for voice fraud or uncertain about their own ability to spot it.
That is the room your greeting walks into. A caller who suspects mid-conversation that the friendly voice is synthetic — and who was not told — has just learned something unflattering about how you do business, at the exact moment you were trying to earn a job. A caller told upfront gets to make an informed choice: keep going, or ask for a human. In our experience the disclosure costs you far less than the discovery does.
There is also a straightforward operational benefit. Callers who know they are talking to an assistant speak more clearly, stick to the question they called about, and ask for a callback when their situation is genuinely complicated. Pretending otherwise buys you a slightly smoother first ten seconds and a worse call.
How to word it
One sentence, at the top, doing three jobs: name the business, flag that the assistant is automated, say what it can do.
"Thanks for calling Ridgeline Plumbing — this is our automated assistant. I can book you in, get a message to the team, or answer questions about our services. What can I help with?"
That is illustrative, not a real customer's script. Note what it does not do: it does not apologise for being an AI, it does not spend fifteen seconds on a legal notice, and it does not say "I am a bot" in a way that invites the caller to hang up. "Automated assistant" and "virtual assistant" both read as honest and competent. If you are in healthcare, legal or financial services, or you operate in Utah, put the disclosure in the first sentence and leave it there — the Utah safe harbour rewards exactly that.
Two things worth checking on whichever product you use. Can you edit the greeting yourself, or is it locked to a vendor template? And does the disclosure survive a call transfer, so the caller knows when they have moved from the AI to a person? Those are reasonable questions to ask before you buy, and they are worth more than any vendor's compliance badge.
Where we sit
NoMissedCall's greeting is written by you, in the app, and can be changed in seconds — which means the disclosure decision is yours rather than ours, and we would rather it stayed that way. Our recommendation to every customer is the same as above: say it in the first line. Every call also produces a transcript and summary you can read afterwards, which is what makes it possible to check what your assistant actually said rather than what you assumed it said. If you want to see how the setup works before deciding, the Help Center walks through the greeting, forwarding and voice settings, and both plans have a 7-day free trial through the App Store and Google Play.
What we cannot tell you is what your specific state and industry require, and any vendor who says otherwise is selling. If you are a medical or dental practice, a law firm doing intake, or a financial adviser, the sector-specific rules — TRAIGA's healthcare provision, professional-conduct rules, HIPAA — matter more than anything in this article, and thirty minutes with a lawyer is cheap next to the alternative.
The bottom line
As of August 2026, no US federal rule requires you to announce that an AI is answering your own inbound line. Utah requires it when a caller asks, and offers a safe harbour if you just say it upfront. Texas requires it of government agencies and of healthcare providers using AI in treatment. California's two headline bot laws do not reach a phone receptionist at all. And the EU's Article 50, live since 2 August 2026, puts the duty on the vendor rather than on you — but is the clearest signal yet of where the whole category is heading.
Which makes the practical answer simple, and it was never really about the law: say it in one sentence, at the start of the call, because it is the version of your business you would want to be on the receiving end of. If you are still choosing a product, we ranked the ones that run entirely from a phone in our 2026 apps guide, priced the alternatives against a human answering service in this cost breakdown, and costed the DIY route in the Twilio build post.
Frequently asked questions
Do I legally have to tell callers my receptionist is an AI?
In the United States, as of August 2026, there is no general federal rule requiring a business to announce that an AI is answering its own inbound phone line. The FCC's February 2024 declaratory ruling brought AI-generated voices under the TCPA, but the TCPA governs calls you place to consumers, not calls consumers place to you. Some state duties can reach you — Utah requires disclosure when a consumer clearly and unambiguously asks, and Texas imposes disclosure duties on government agencies and on healthcare providers using AI in diagnosis or treatment. If any of your callers are in the EU, the AI Act's transparency rule has applied since 2 August 2026.
What changed on 2 August 2026?
The transparency obligations in Article 50 of the EU AI Act became applicable. For systems that interact directly with people — chatbots and voice assistants included — the provider must ensure people are informed they are dealing with an AI, unless that is obvious from the circumstances. A limited transition to 2 December 2026 applies to the separate machine-readable marking duty for synthetic content, not to the interaction-disclosure duty.
Does the EU AI Act apply to a business in the United States?
It can. The AI Act reaches providers and deployers outside the EU where the output of the AI system is used in the EU. A plumber in Ohio taking calls from Ohio customers is outside its scope. A business that markets into the EU, or whose AI receptionist routinely handles callers in EU member states, is a different question and worth putting to a lawyer rather than a blog post.
Does the TCPA make my AI receptionist illegal?
No. The FCC's 2024 ruling confirmed that AI-generated voices count as an artificial or prerecorded voice under the TCPA, which triggers consent, identification and opt-out rules for outbound calls to consumers. An AI answering an inbound call to your published business number is not a robocall, because the consumer initiated the contact. The FCC's proposed rules on defining and disclosing AI-generated calls remain a proposal, not a final rule.
Does California's bot disclosure law cover phone calls?
No. California's B.O.T. Act, SB 1001, defines a bot as an automated online account and defines online as a public-facing website, web application or digital application. Telephone calls fall outside that definition. California's SB 243, effective 1 January 2026, covers companion chatbots — systems designed for ongoing social or emotional interaction — not a receptionist that books appointments.
What should my AI greeting actually say?
One sentence at the top of the call is enough for almost every business: name the business, say the assistant is virtual or automated, and say what it can do. Something like "Thanks for calling Ridgeline Plumbing — this is our automated assistant, I can book you in or take a message." It satisfies every disclosure duty currently written down, it costs about two seconds, and it sets the caller's expectations before they ask a question the AI can't handle.
Try it on your own phone
No Missed Call is an AI receptionist that answers your business line 24/7, books appointments straight into your calendar, and sends you a transcript and summary of every call. Setup takes minutes from your phone, and there is a 7-day free trial.
Download on the App Store · Get it on Google Play · Read the Help Center
