Should your AI agent say it is an AI?
A rule requiring disclosure was proposed by the FCC in September 2024. Even where it is not yet mandatory, finding out later costs more trust than saying so upfront.
Want to see it first? Book a demo
The short answer
- In September 2024 the FCC proposed a rule that would make callers tell people when a call uses an AI voice. That would add to the consent rule set by its February 2024 ruling on AI voices under the TCPA.
- Whether or not the law requires it where you are, a caller who finds out later that they were talking to an AI tends to feel misled. That costs you trust, apart from any legal requirement.
- Disclosure doesn't need a long disclaimer. A brief, plain line near the start of the call meets the intent and keeps the talk on track.
- Check the FCC's site for the current status of the September 2024 proposal before assuming it is finalized rule, since a proposed rule and an adopted rule are not the same thing.
Why this is unsettled, not simple
Since the FCC's February 2024 ruling, AI voices count as artificial voices under the TCPA. That settles the consent question: a call with an AI voice needs the same prior express consent an automated or prerecorded call already needs. Disclosure is a separate, newer question. The FCC's September 2024 proposal would add a requirement to tell the recipient the call is AI-generated, but a proposed rule is not the same as an adopted one, and the status can change. If you use AI voice calling, check the FCC's own site for where that proposal stands. Don't assume it's already law, or that it never will be.
Separate from the regulatory question is the trust question, which does not wait for a rule to be finalized. Picture a caller who has a full talk, books something or shares details, thinking it's a person. When they later learn it was AI, they tend to feel the call wasn't what it claimed to be. That reaction happens whether or not disclosure was legally required at the time.
Disclosure now vs. disclosure later
| Disclosed upfront | Discovered after the fact | |
|---|---|---|
| Caller's sense of the interaction | Knows what they are dealing with from the start | Feels the earlier trust was based on a false premise |
| Legal risk if disclosure becomes a rule | Already compliant | Retrofit required, possibly under time pressure |
| Effort to implement | One short line near the start of the call | None until a rule or a complaint forces it |
What a reasonable disclosure looks like
A short, plain line near the start of the call is most often enough. Say the call is automated or AI-assisted, then move on to why you're calling. It does not need to repeat itself throughout the conversation to be genuine, and it does not need to be defensive or apologize for existing.
Disclosure: SalesCrew is our product. SalesCrew places outbound AI voice calls to consented contacts, and every external send from a SalesCrew agent waits in an approval queue by default. The reasoning here applies to any vendor's AI calling capability.
A proposed rule is not a finalized one
Questions
- Is disclosure legally required today?
- The FCC proposed a rule in September 2024 that would require you to disclose AI calls, on top of the TCPA consent rule. Check the FCC's site to see if that proposal is final before treating it as settled law. This is not legal advice.
- Does disclosure hurt conversion?
- No outside data you can check covers this for sales calls. So treat any claim that disclosure helps or hurts conversion as unproven until you've tested your own numbers.
- What is a natural way to disclose without derailing the call?
- A brief line near the start keeps it factual. Say the call is automated or AI-assisted, then get to the reason for it, with no long disclaimer to sit through.