Is AI calling legal in the US?
Yes, AI calling is legal in the US when you have the right consent for the type of call. Since February 2024 the FCC treats AI-generated voices as "artificial" voices under the Telephone Consumer Protection Act, so an AI voice agent calling a mobile number needs the same prior express consent as a prerecorded call, and prior express written consent if the call is marketing.
That one sentence covers most of the risk. The ruling is the FCC's Declaratory Ruling FCC 24-17, adopted on February 2, 2024 and released on February 8, 2024. It did not create a new law for AI. It confirmed that the existing TCPA rules for "artificial or prerecorded voice" calls already cover voices generated or cloned by AI. So the question is never "is AI allowed", it is "did this person agree to this kind of call".
What follows is general information for business owners, not legal advice. If you run outbound campaigns at volume, have a lawyer review your consent language once. It costs far less than one class action.
What consent does an AI voice agent need?
It depends on the purpose of the call. An informational call, such as an appointment reminder or a callback the person asked for, needs prior express consent, which usually means the person gave you their number for that purpose. A marketing or sales call made with an AI voice needs prior express written consent: a signed agreement, electronic signatures included, that names your business.
Which consent does this AI call need?
| Call type | Example | Consent needed for an AI voice |
|---|---|---|
| Inbound call answered by AI | AI receptionist picks up when a customer calls you | No TCPA consent needed; the caller started it. Disclose recording if you record. |
| Informational outbound | Appointment reminder, "your quote is ready" | Prior express consent |
| Requested callback | Web form says "call me about my quote" | Prior express consent for that request; keep the form record |
| Marketing outbound | Promoting a new service to past leads | Prior express written consent naming your business |
| Cold call to a bought list | Numbers with no relationship to you | Not possible with an AI voice without written consent; do not do it |
The written consent definition sits in the FCC's rules at 47 CFR 64.1200: "an agreement, in writing, bearing the signature of the person called that clearly authorizes the seller to deliver" telemarketing using "an artificial or prerecorded voice." A checkbox on a web form, unticked by default, with clear wording and your business name, is the usual way to get it.
Does the answer change for inbound AI calls?
Mostly, yes. When a customer calls you and an AI answers, the TCPA's outbound consent rules do not apply, because you did not place the call. What still applies is call recording law, honesty about what the caller is talking to, and any follow-up texts the AI sends after the call.
That last part is where inbound setups get into trouble. An AI receptionist that books the appointment and then fires a marketing text sequence has turned an inbound call into outbound marketing. Reminders about the booking are informational. A "we have a special offer" text is marketing and needs its own consent, and in GoHighLevel it also needs a registered texting campaign, covered in our A2P 10DLC registration guide.
On recording: about a dozen states require every party to agree before a call is recorded. The simple fix is a short line at the start of each call: "This call is answered by an AI assistant and may be recorded." It covers disclosure and recording in one sentence, and callers rarely hang up over it.
What happens if an AI call breaks the TCPA?
The TCPA lets people sue for $500 per call or text, and a court can raise that to $1,500 per call if the violation was willful or knowing. The amounts come from the statute itself, 47 U.S.C. 227(b)(3), and they apply per call, so a campaign of 2,000 calls without consent is a potential $1 million to $3 million exposure before legal fees.
Potential statutory damages, 2,000 calls without consent
That arithmetic is why TCPA suits are popular with class-action firms. State attorneys general and the FCC can also act. In 2024 the FCC proposed a $6 million fine, later issued, against the consultant behind the AI-cloned voice robocalls sent before the January 2024 New Hampshire primary. You do not need to be a scammer to be sued, though. Most suits against ordinary businesses start with one missing checkbox or one ignored "stop".
How do opt-outs work with an AI voice agent?
A person can revoke consent by any reasonable means, and you must honor it within a reasonable time, no more than 10 business days under the FCC's rules. For an AI voice agent that means it has to recognise "stop calling me", "take me off your list" and similar phrases mid-call, end the call politely, and write the opt-out to your CRM so no workflow calls that number again.
How an opt-out should flow during an AI call
The FCC changed parts of this in 2024 and was scheduled to vote on September 30, 2026 on a draft order that would let an opt-out from an informational call apply only to that category. We track that in our news post on the FCC opt-out rule changes. Whatever the final wording, the safe build is the same: detect the request, tag the contact, remove it from every matching workflow within minutes, not days.
- In the call script: the agent is told to stop the pitch and confirm the opt-out when it hears any opt-out phrase.
- In the CRM: a Do Not Call tag or DND setting is written on the contact during the call, with the date and the call ID.
- In every outbound workflow: the first step checks that tag and ends the workflow if it is present.
How do you run AI calling without legal risk?
Collect consent before you call, keep proof of it, call only in allowed hours, identify your business at the start, honor opt-outs instantly, and never use an AI voice on numbers you bought or scraped. Those six rules cover the large majority of TCPA exposure for a service business.
- Consent at the form. Separate, unticked checkboxes for informational calls and for marketing calls, each naming your business.
- Proof stored on the contact. Keep the form version, the consent text, the timestamp and the page URL. If you are ever asked, you can show it in minutes.
- Calling hours. Federal rules allow telemarketing calls between 8 a.m. and 9 p.m. in the called person's local time, and some states are stricter. Our page on what time businesses can text and call customers has the state detail.
- Identify yourself. Say the business name and that the caller is an AI assistant in the first sentence.
- Scrub against the Do Not Call Registry for any marketing call to someone who is not an existing customer or enquirer.
- Keep a human path. "Would you like to speak to someone on the team?" routes to a person or books a callback.
This is how we build every AI calling agent: consent and opt-out handling go in before the first script line is written. If you use GoHighLevel's own voice agent, the Voice AI outbound consent change explains what HighLevel now records per location and what it leaves to you. For how an AI agent fits into a real lead flow, see the Ultra Cryo & Recovery case study.
Questions about AI calling and the law
Do I have to tell people they are talking to an AI?
Federal law does not yet require a spoken AI disclosure on every call; the FCC proposed AI call disclosure rules in August 2024, so check the current status of that proceeding before you rely on either answer. Several states have their own disclosure rules for bots, and the prerecorded-message rules already require you to identify your business. We add a one-line AI disclosure to every call anyway, because it costs nothing and removes the argument.
Can an AI agent call leads who filled in my web form?
Yes, if the form asked for consent to be called and the call is about what they asked for. A quote request with a clear "we will call you about your quote" line supports an informational callback. A marketing call about other services needs the written consent checkbox.
Is AI calling legal for healthcare practices?
The FCC's rules allow certain healthcare messages covered by HIPAA, such as appointment reminders, with lighter consent rules, but they still limit frequency and require an easy opt-out. The recording and data side is covered by HIPAA itself; our guide to HIPAA and automation tools lists which platforms sign a business associate agreement.
Does the TCPA apply to business-to-business calls?
The artificial voice rules for mobile numbers apply whoever owns the phone, so an AI call to a business owner's mobile needs consent just like a call to a consumer. The Do Not Call Registry rules mainly protect residential numbers.
Are AI calls to existing customers treated differently?
An existing relationship helps with Do Not Call Registry rules, but it does not replace consent for artificial voice calls to mobile numbers. Reminders about their own appointments are informational and are the lowest-risk use of an AI voice.
Sources and further reading
- FCC, Declaratory Ruling FCC 24-17 (AI-generated voices are "artificial" under the TCPA), released February 8, 2024.
- FCC, FCC confirms TCPA applies to AI technologies that generate human voices, February 8, 2024.
- Code of Federal Regulations, 47 CFR 64.1200 (consent, calling hours, opt-outs), via Cornell LII.
- US Code, 47 U.S.C. 227 (the TCPA, including the $500 and $1,500 damages), via Cornell LII.
