Key takeaways
- The FCC treats AI-generated voices as artificial or prerecorded voice calls under the TCPA.
- FTC rules restrict most prerecorded telemarketing calls and impose disclosure, opt-out, Do Not Call, timing, caller-ID, and recordkeeping duties where applicable.
- Consent must be tied to the seller, number, purpose, channel, and current legal requirements—not inferred from a purchased list.
- Get qualified legal review for every jurisdiction and use case; this guide is an operating checklist, not legal advice.
01
Treat the dial as a controlled action
An outbound call crosses from software behavior into a person’s day. Before a number enters a campaign, the system should be able to show where it came from, which seller is calling, the purpose, the consent or exemption being relied on, applicable suppression checks, permitted calling time, and the approved script version.
Do not make the model decide whether consent is valid. That determination belongs in governed data and policy logic before the call is attempted.
02
AI does not create a regulatory category outside robocalls
The FCC’s 2024 declaratory ruling confirmed that AI-generated voices fall within the TCPA’s artificial or prerecorded voice restrictions. The FTC has also affirmed that voice cloning does not evade the Telemarketing Sales Rule and notes that the rule bans nearly all telemarketing robocalls to consumers, subject to its terms and exceptions.
Federal rules are not the entire picture. State laws, sector rules, international requirements, carrier policies, and the facts of the call may add or change obligations. Build jurisdiction into campaign approval rather than trying to repair it after complaints arrive.
Swipe to compare every column
| Control | Evidence to retain | Unsafe shortcut |
|---|---|---|
| Permission | Source, timestamp, language, seller, number and scope | “The lead filled a form once” |
| Suppression | National, state and seller-specific checks | A manual spreadsheet updated occasionally |
| Identity | Approved seller name, caller ID and purpose disclosure | A persona that obscures who is calling |
| Opt-out | Immediate voice or keypress path and durable suppression | Asking the caller to email later |
03
Make refusal easier than persuasion
Required disclosures must be understandable and timely; an opt-out cannot be buried after the pitch. Design spoken and keypress paths that work during the call, update the seller-specific suppression list immediately, end the call cleanly, and prevent retries from another campaign or vendor.
Test ambiguous language such as “not now,” “stop calling,” “remove me,” and interruptions during the disclosure. When intent is unclear, the safe path is to stop the marketing action and route the record for review—not to let the model argue.
04
Audit vendors as part of the campaign
Document the seller, lead source, dialer, voice provider, model provider, call-routing vendor, data processor, and any agency or affiliate. Define who maintains suppression, who responds to complaints, who can stop the campaign, and how records can be produced. Contract language does not replace a technical stop control.
Launch with a narrow, reviewed population and daily complaint monitoring. A campaign that produces meetings while generating unauthorized or deceptive calls is not commercially healthy; it is accumulating liability and distrust.
Primary sources and further reading
Use the source material to validate details against your own context and current platform configuration.
- FCC: AI-generated voices in robocalls are artificial under the TCPA
- FTC: Complying with the Telemarketing Sales Rule
- FTC: Protections against AI-enabled scam calls
- FTC: Telemarketing business guidance
This field note follows the XenGrowth editorial policy: primary sources where available, visible limitations, material review dates, and no invented first-hand experience.
Stay with the problem
