AI Calling Laws & TCPA Consent Requirements in 2026

A financial services company in Texas deployed an AI voice agent to follow up with leads in January 2025. The system made 4,200 calls in its first month. The leads had opted in to receive "phone calls about loan options." None of the consent forms mentioned "artificial or prerecorded voice."

Every one of those 4,200 calls was a TCPA violation worth up to $1,500. Total exposure: $6.3 million — from a single month of an AI tool the company thought was just a more efficient way to make follow-up calls.

This is the new reality of AI calling under the TCPA (the federal law that restricts business calls and texts). Since the FCC's February 2024 ruling, AI-generated voices are legally classified as "artificial or prerecorded voice" — the same category that has triggered billions of dollars in robocall penalties over the past three decades.

This post explains what the ruling actually means, what consent you need for different types of AI calls, which scenarios are always illegal, and how to set up your AI calling workflows so they do not become your most expensive business decision.

What the FCC's February 2024 AI Calling Ruling Actually Says

On February 8, 2024, the FCC issued a declaratory ruling that settled a question the telecom industry had been debating for years: do AI-generated voices count as "artificial or prerecorded voice" under the TCPA?

The answer is yes. Unambiguously.

The ruling states that any call using a voice generated or substantially modified by artificial intelligence technology falls under the TCPA's existing restrictions on artificial and prerecorded voice calls. This is not a new law — it is a clarification that the 1991 TCPA already covers AI voices.

What this means in practice:

• AI voice agents that call leads, customers, or prospects are regulated exactly like robocalls.
• AI-generated voicemails (ringless or otherwise) require the same consent as a live prerecorded message.
• Voice cloning — using AI to replicate a real person's voice — is treated identically to a prerecorded message from that person.
• AI-assisted calls where a human initiates the call but AI handles part of the conversation are in a gray area, but the FCC has signaled these will be treated as artificial voice calls.

The ruling did not create a separate consent framework for AI calls. It folded AI calling into the existing prior express consent and prior express written consent requirements — the same framework that has governed robocalls since 1991.

Human Calls vs. AI Calls: Consent Requirements Side by Side

The consent you need depends on two factors: the purpose of the call and whether AI is involved.

The critical difference: a human sales rep can call a lead who gave their phone number on a web form. An AI voice agent making that same call needs prior express written consent — and that consent must include specific language about artificial or prerecorded voice.

If your existing consent forms say "I agree to receive phone calls from [Business Name]" without mentioning artificial or prerecorded voice, your AI calling program has no valid consent.

3 AI Calling Scenarios That Are Always Illegal

These three uses of AI calling violate the TCPA regardless of what consent you have:

1. AI cold calls to numbers without any prior consent

An AI voice agent calling through a purchased lead list where none of the contacts opted in to hear from your specific business is illegal. This is true even if the leads "opted in" on a third-party form that listed dozens of companies. Under the FCC's one-to-one consent standard, consent must name your specific business.

2. AI calls to numbers on the National Do Not Call Registry without prior express written consent

Numbers registered on the DNC list cannot receive marketing calls — human or AI — unless the consumer has given prior express written consent to your specific company. An AI system does not bypass DNC restrictions. Calling DNC-listed numbers with an AI voice agent without written consent is a per-call violation.

3. AI-generated calls that impersonate real people without disclosure

Using AI to clone a celebrity's voice, a politician's voice, or even your CEO's voice without disclosing that the call is AI-generated violates both the TCPA and a growing number of state deepfake and consumer protection laws. California requires AI call disclosure within the first 15 seconds. Several states are enacting similar requirements.

The AI Calling Decision Flowchart

Before you deploy an AI voice agent, walk through this decision tree:

Step 1: What is the purpose of the call?

• Marketing, sales, or promotional → Go to Step 2A
• Informational, transactional, or service-related → Go to Step 2B
• Emergency or health/safety → Exempt from consent requirements (proceed with caution)

Step 2A: Marketing AI calls

• Do you have prior express written consent? → If no, STOP. You cannot make this call.
• Does the consent form include "artificial or prerecorded voice" language? → If no, STOP. Your consent does not cover AI calls.
• Is the number on the National DNC list? → If yes, confirm your written consent explicitly authorizes marketing calls.
• Does the consent name your specific business? → If no, STOP. Generic "partner" consent is not sufficient.
• All yes → Proceed. Record and store the consent proof.

Step 2B: Informational AI calls

• Did the consumer provide their phone number to your business? → If no, STOP.
• Would the consumer reasonably expect this type of call? → If no, STOP.
• Does the call contain ANY promotional content? → If yes, go back to Step 2A. It is a marketing call.
• All clear → Proceed. But add any promotional content later and you need written consent.

5 Real Enforcement Patterns You Should Know

While the February 2024 ruling is relatively new, enforcement has already begun to take shape through a combination of FCC actions, state AG investigations, and private TCPA lawsuits.

1. The $299 million auto-warranty signal (August 2023). The FCC fined a robocall operation $299,997,000 for billions of prerecorded auto-warranty calls. While this predates the AI ruling, the FCC explicitly cited this case when announcing the AI voice clarification — signaling that AI calling operations at scale will face the same enforcement posture.

2. State AG actions against AI voice scams (2024–2025). Multiple state attorneys general have used the FCC's ruling as the legal basis to pursue AI voice scam operations. These cases establish that state-level enforcement agencies view AI calling violations as within their jurisdiction, not just the FCC's.

3. Insurance industry AI dialer exposure. Insurance agencies using AI dialers to contact Medicare leads face specific prohibitions. Medicare supplement leads require a recorded Permission to Contact (PTC) less than 12 months old. AI dialers used without PTC are classified as ATDS violations. For more on insurance-specific requirements, see our TCPA consent language guide for insurance lead forms.

4. GoHighLevel AI workflow liability. Businesses using GoHighLevel's AI follow-up workflows should be aware that automated AI outreach through GHL is legally classified as ATDS. Every AI-initiated text or call through a GHL workflow without proper consent is a separate TCPA violation.

5. Private plaintiff lawsuits targeting AI voice agents. TCPA plaintiff attorneys are actively monitoring businesses that deploy AI voice agents for outbound calling. The combination of high per-call penalties ($500–$1,500) and easy proof of violation (the consumer can simply record the AI voice) makes AI calling a high-value target for class action litigation.

The Consent Language You Need for AI Calling

If you want to use AI voice agents legally, your consent form must include specific language. Here is a template that covers both AI calling and SMS:

"By submitting this form, I authorize [Business Name] to contact me by automated text message and phone call, including by artificial or prerecorded voice, regarding [specific purpose — e.g., 'loan options,' 'insurance quotes,' 'appointment scheduling']. Consent is not required to purchase. Msg & data rates may apply. Reply STOP to opt out."

Key elements in this disclosure:

Without the phrase "artificial or prerecorded voice," your consent covers human calls only. You would need to go back to every contact and obtain new consent before using AI calling — or face per-call liability.

For more consent language templates covering different industries and use cases, see our template library.

State Laws That Add AI Calling Restrictions

The federal TCPA is the floor, not the ceiling. Several states have enacted or proposed laws that add requirements specifically targeting AI-generated calls.

Rule of thumb: Comply with the strictest applicable law. If you are calling a Florida resident from a Texas office, you need to satisfy both Florida FTSA and Texas identification requirements. The consumer's home state law typically controls.

For a full state-by-state breakdown of text message and calling laws, see our text message opt-in laws guide.

How to Set Up AI Calling the Right Way

If you are going to use AI voice agents, here is the compliance infrastructure you need:

1. Update your consent forms immediately.

Add "artificial or prerecorded voice" to every consent form that covers phone calls. If your forms only mention "text messages," they do not cover voice calls at all — AI or human.

2. Separate marketing consent from transactional consent.

A customer who books an appointment has given implied consent for appointment reminders. They have not given consent for marketing calls. Keep these two consent types in separate data fields.

3. Record consent proof for every contact.

When a contact opts in through your form, capture the full consent record: timestamp, IP address, user agent, the exact disclosure text they saw, and ideally a session replay of the opt-in interaction. This proof is what protects you when — not if — someone challenges whether they actually consented.

4. Log every AI call with metadata.

For each AI-generated call, log: the number called, date and time, duration, consent record ID that authorized the call, and the AI system used. This creates an auditable trail linking every call back to its consent basis.

5. Honor [consent revocation](/glossary/consent-revocation) across all channels.

When someone says "stop calling me" during an AI call, by text, by email, or by any other reasonable method, that revocation must propagate to your AI calling system, your SMS platform, your CRM, and any other system that initiates contact. The FCC's 2024 rules require you to honor revocation by any reasonable means within 10 business days.

6. Build a kill switch.

Your AI calling system should have an immediate shutdown capability. If you discover a consent gap, you need to stop all AI calls within hours, not days. The difference between 4,200 violations and 42,000 violations is how quickly you can stop the system.

Why Standard CRM Consent Fields Are Not Enough for AI Calling

Most CRMs store consent as a single checkbox field: "Phone consent: Yes/No." That field tells you nothing about:

• Whether the consent form mentioned "artificial or prerecorded voice"
• What version of the disclosure the consumer saw
• Whether consent was given before or after you added AI calling to your workflows
• The full metadata courts require to validate consent (IP, timestamp, user agent, form version)

A plaintiff's attorney does not need to prove you made the AI call — phone records and the consumer's testimony handle that. They need to prove you lacked adequate consent. A CRM checkbox with no supporting documentation is not adequate consent proof.

This is where tamper-proof consent storage matters. OptInFix captures every consent interaction with full metadata, session replay, and SHA-256 hashing — creating the kind of evidence package that withstands litigation. When the question is "did this person specifically agree to receive AI-generated calls?", you need to produce the exact form they saw, the exact words they agreed to, and proof they were not pre-checked or auto-submitted.

Frequently Asked Questions

Does the FCC AI calling ruling apply to AI chatbots and text messages?

The February 2024 FCC ruling specifically targets AI-generated voice calls — calls where an artificial intelligence system produces or mimics human speech. It does not directly regulate AI-generated text messages or chatbots. However, AI-powered systems that auto-dial text messages are already regulated under the TCPA's ATDS provisions. The safest approach is to obtain prior express written consent for any AI-initiated outreach, whether voice or text.

Can I use AI voice agents for appointment reminders without written consent?

Yes, but with conditions. Appointment reminders are classified as informational or transactional calls, which require prior express consent — a lower bar than the written consent required for marketing. The consumer must have provided their phone number to your business and reasonably expect the call. However, the moment you add any promotional content to that reminder — like "and we have a special offer this month" — it becomes a marketing call requiring prior express written consent that includes "artificial or prerecorded voice" language.

What consent language do I need for AI calling?

Your consent disclosure must include the phrase "artificial or prerecorded voice" to cover AI-generated calls. A compliant template: "By submitting, I authorize [Business Name] to contact me by automated text message and phone call, including by artificial or prerecorded voice, regarding [purpose]. Consent is not required to purchase. Msg & data rates may apply. Reply STOP to opt out." Without those specific words, consent obtained for regular calls does not extend to AI calls.

Are there state laws that restrict AI calling beyond the federal TCPA?

Yes. Several states have enacted or proposed laws that go further than the federal TCPA on AI calling. Florida's FTSA requires written consent for any automated sales call to Florida residents, with penalties of $500 per violation and no cure period. California's SB-1228 requires AI-generated calls to disclose they are AI within the first 15 seconds. New York and Illinois have additional restrictions on automated calling systems. Always comply with the strictest applicable law — which is usually the consumer's home state.