Which US states require an AI-disclosure statement before a robocall in 2026?
This matters to any business whose phone does the work: trades, clinics, salons, agencies. The honest answer is a short list, a federal rule that isn't final yet, and a three-state consent split most compliance pages haven't caught.
Answered in short
5 things that decide this
- 01No state statute in force in 2026 requires an AI-disclosure statement before every robocall. The binding layer is federal: the FCC ruled in February 2024 that AI-generated voices count as artificial voices under the TCPA, so they need prior express consent.
- 02California is the clear state exception. AB 2905 requires a prerecorded call that uses an AI-generated voice to say so in the message.
- 03The FCC has proposed a nationwide in-call AI disclosure rule, in a 2024 rulemaking that also defines an AI-generated call. It had not been adopted as of this page's verification date.
- 04The real 2026 state split is consent, not disclosure. On 26 February 2026, the Fifth Circuit ruled in Bradford v. Sovereign Pest Control that the TCPA accepts oral consent for telemarketing calls, which puts Texas, Louisiana and Mississippi on a different standard from the rest of the country.
- 05Disclosure costs one sentence of script. Consent records cost a system. Build the call platform to say what it is and log every consent, and the state-by-state question mostly stops mattering.
The TCPA already covers AI voices
AI-voice calls are regulated today through consent, not disclosure. In February 2024, the FCC ruled that AI-generated voices count as artificial voices under the Telephone Consumer Protection Act. So an outbound AI call needs prior express consent before you place it. Marketing calls have needed written consent under the FCC's rules; a service call, like an appointment reminder, can rest on oral consent.
A disclosure duty is coming, just not in force. The FCC's 2024 rulemaking defines an AI-generated call and proposes consent language that names AI. It also proposes a clear statement, at the start of each call, that the voice is AI-generated. Until that rule is adopted, the federal duty is consent plus the older identity rules for prerecorded calls.
Texas, Louisiana and Mississippi now run on oral consent
On 26 February 2026, the Fifth Circuit held in Bradford v. Sovereign Pest Control of TX that the TCPA itself requires only prior express consent, oral or written. The FCC's written-consent rule for telemarketing, the court held, went beyond the statute. That reading binds federal courts in Texas, Louisiana and Mississippi. Everywhere else, written consent remains the standard for marketing robocalls.
For a business that calls in one of those three states, the ruling is a defense, not a strategy. A national campaign gets judged wherever the called party sues, and the circuit split invites review. Written consent still travels; oral consent only holds inside the Fifth Circuit's lines.
Build so the map stops mattering
One build satisfies every version of these rules. The call opens by naming the business and the AI. Each consent lands in a log with its channel and date, and every opt-out is honored on the spot. Disclose everywhere and the FCC's pending rule changes nothing for you. Hold written consent everywhere and the circuit split changes nothing either.
That's how we build voice agents at Hashlogics. The disclosure line sits in the script, and the consent ledger is a table, not a policy document. You might run a service desk, a clinic front office or a sales floor. Either way, the compliance layer belongs in the software, where no one can skip it on a busy day.
- Consent capturedWritten where marketing; logged always.
- Ledger entryWho, channel, date, scope.
- Call opensNames the business and the AI.
- Purpose statedReminder, follow-up or offer.
- Opt-out honoredSuppressed before the next dial.
Consent is the layer regulators check first. Disclosure is the cheapest line in the script.
Related questions
01Does an inbound AI receptionist need to disclose it's an AI?+
The TCPA governs outbound calling, so no federal telephone rule forces an inbound AI receptionist to announce itself. We build ours to say so anyway. A caller who feels tricked hangs up, and the pending FCC rule shows where expectations are heading.
02Can I skip written consent for marketing calls into Texas?+
Courts in Texas, Louisiana and Mississippi now accept oral consent for telemarketing under Bradford v. Sovereign Pest Control. The FCC's written-consent rule still stands elsewhere, and a multi-state campaign gets judged where the called party sues. Written consent remains the safe national standard.
03What are the penalties for getting it wrong?+
TCPA statutory damages are 500 dollars per call in a private suit, and up to 1,500 dollars per call for willful violations. Damages accrue per call placed, which is why robocall class actions settle large and why the consent ledger matters more than the script.
Related
- TCPA rules for automated service texts →The same consent logic, applied to texting.
- Can an AI agent answer your business phone? →What a production phone agent actually takes.
- Voice AI agent development →Our service page: agents built with the disclosure line and the consent ledger in place.
- AI receptionist vs answering service →The buy-versus-staff comparison for phone-heavy businesses.
