AI receptionists filter out the large majority of robocalls by testing whether a caller responds like a real person — not just by matching phone numbers against a blocklist. That distinction matters more in 2026 because a growing share of the scam calls reaching business lines are themselves AI-generated, meaning both sides of the call can now run on the same technology.
The volume problem isn't going away
Businesses aren't imagining the flood. The U.S. saw an estimated 52.5 billion robocalls placed in 2025, only about 1% fewer than the year before, according to the YouMail Robocall Index (GetVoIP, 2026). At the same time, the average American now fields 9.9 unwanted calls a week — north of 500 a year — a figure growing at a 16% compounded annual rate since 2023 (Hiya, 2026). None of that volume distinguishes between a household landline and the one line a two-person shop uses to run its whole business.
The new wrinkle: the scam caller can sound like someone you know
What's changed isn't just volume — it's who's on the other end. Hiya's 2026 State of the Call report found that one in four Americans have received an AI deepfake voice call in the past 12 months, and a further 24% say they aren't confident they could tell a cloned voice from a real one (Hiya, 2026). For a business, that's not an abstract consumer-fraud statistic. It's the "vendor rep" who sounds exactly like the person who's called every month for two years — except this time the urgent ask is a wire transfer or a gift-card payment against an unpaid invoice.
It's already illegal — that hasn't slowed it down
The FCC settled the legal question in February 2024, ruling that an AI-generated voice on a robocall counts as an "artificial voice" under the Telephone Consumer Protection Act — meaning it requires the called party's prior express consent, the same standard as any prerecorded call (Conference Board, 2024). The ruling followed a deepfake robocall impersonating a sitting U.S. president ahead of the New Hampshire primary. Two years later, the illegality hasn't dented the volume; it's mainly given prosecutors a cleaner case once a scammer is caught.
Why blocking known numbers alone doesn't catch this
Most legacy call-blocking still works off two things: a list of known bad numbers, and STIR/SHAKEN attestation, the caller-ID authentication standard carriers use to flag spoofed numbers. Both help, and both share the same blind spot — they evaluate the number, not the conversation. A freshly spoofed number with clean attestation, or a cloned voice reading off a real vendor's name, sails through both checks. AI-based call screening works differently: it reads how the caller behaves in the first several seconds of live conversation — whether they respond naturally to a direct question, pause and react like a person would, or just run a fixed script regardless of what's said back to them.
| Screening method | What it actually checks | Blind spot |
|---|---|---|
| Number / blocklist | Caller ID against known spam numbers | Beaten by any new or spoofed number |
| STIR/SHAKEN attestation | Whether the carrier verified the calling number wasn't spoofed | Verifies the number, not what's said — a clean number can still carry a scam script |
| Behavioral AI screening | Whether the caller responds like a real person to a real question | Can still produce a false positive on an unusual but legitimate caller |
The businesses with the least cushion carry the most risk
A large call center can staff a dedicated line to vet vendor and payment calls before anyone with authority to move money touches them. A five-person plumbing outfit in Kalispell answering its own phones on a normal Tuesday can't build that same layer — whoever picks up is usually also the one who could approve a payment or a supply order. That's the exact structural gap voice-cloning scams are built to exploit, and it's also exactly the gap AI call screening is built to sit in front of: catching a scripted, non-responsive caller before it ever reaches whoever's fielding the front desk today.
What to actually check before a call like this costs you money
- Treat any unexpected request to change payment details, wire funds, or share account access as a reason to hang up — regardless of how familiar the voice sounds.
- Call the vendor or contact back on a number you already have on file, never the number that just called you.
- Confirm large or unusual payment changes through a second channel — email to a known address, or a call to a known contact — before acting.
- Ask a live question a script can't anticipate. Scripted and cloned-voice callers struggle with anything outside their prepared flow.