The phone rings at a small garage. Nobody recognises the number. It might be a customer whose car won’t start, or a recorded voice about a parcel that was never sent. Whoever is nearest has a second to decide, and letting it ring feels safest. The numbers show why.

Suspicious calls are part of everyday life

In Ofcom’s 2024 research on suspicious calls and messages, a survey of 2,128 UK adults, 39% of mobile users and 48% of landline users said they’d had a suspicious call in the previous three months.

Fraud is common too. The ONS estimates 4.5 million fraud incidents in England and Wales in the year to March 2026, out of 9.6 million incidents of headline crime: close to half.

In the US, the Federal Trade Commission told Congress in March 2026 that consumers filed 3 million fraud reports in 2025 and reported losing $15.9 billion. Over 1 million of those reports were about impostor scams.

“It sounded like a machine” is how people spot a scam

Ofcom asked people who had picked up a call that looked genuine but turned out to be suspicious what gave it away. The top answer, at 49%, was that the voice sounded recorded rather than live. And 30% said the caller didn’t sound like a real person speaking, or sounded like a bot.

Caller ID doesn’t settle the question. In the same research, 82% of people thought scammers could fake the mobile number shown on their screen, and 77% thought the same of landline numbers.

So people hold back. Ofcom found that 43% would be very unlikely to pick up a call from a withheld number, and 58% very unlikely to answer an international number they didn’t recognise.

The old test is breaking

Listening for a robot worked while scam calls sounded robotic. In Starling Bank’s 2024 survey of 3,010 UK adults, 28% thought they’d been targeted by an AI voice-cloning scam in the past year, and 46% didn’t know such scams existed. Starling notes that a voice can be cloned from as little as three seconds of audio.

Hiya, a call-blocking company, reported that 26% of UK consumers it surveyed said they’d had a deepfake voice fraud call in the past year. In early 2025 it found AI-generated audio in 1 in 4 calls to its honeypot numbers, and 55% of those were fraud. These are a vendor’s figures: a signal, not a census.

So the test now fails both ways. A cloned voice can sound like someone you trust. An honest AI assistant, booking a table for a real customer, sounds exactly like what it is.

The rules start with disclosure

In February 2024 the US Federal Communications Commission ruled that AI-generated voices count as “artificial” under its robocall rules, so callers need the same consent they’d need for any artificial or prerecorded voice. Since 2 August 2026, the EU AI Act has required that people be told when they’re interacting with an AI system, unless that’s obvious.

Disclosure is the right first step, but it isn’t proof. A scam can claim to be anyone. And to a busy receptionist, an honest “I’m an AI calling for a customer” sounds like every other unwanted call.

What follows

As more genuine calls come from AI acting for real people, they’ll land in the same pile as the scams. That hurts both ends of the line. The person who asked their assistant to move an appointment doesn’t get it moved, and the business loses a customer it never knew was calling.

Humans first, with proof

We think the way through is to keep people at the centre of every call, and to make proof something a business can check for itself.

  • Every call answered. Your Telephonist answers your business calls day and night on your own UK number, books callers in and puts them through to your team. One of our people is on call as backup around the clock.
  • Every call our AI makes carries proof. When Telephonist rings a business for someone, say after they ask in ChatGPT and approve the call, the call is signed and our AI gives a short reference when asked. It states which proofs stand behind it, such as Proof of Payment or Proof of Human, and Benaiah AI Ltd answers for it.
  • Any business can check a call. Enter the reference and the number we called on our Check a call page to see that the call is genuine, which proofs stand behind it and who is accountable. Set a front desk rule to choose which AI calls placed through Telephonist reach you: anyone, Proof of Payment and above, Proof of Human only, or none. Both are free.

To be clear about the limits: checking a call and setting a front desk rule cover only calls placed through Telephonist. They don’t detect or block anyone else’s AI calls, or scam calls. What they do is let a genuine call show where it came from.

People already hang up on anything that sounds automated. As AI calls grow, genuine ones will need to prove who sent them, and businesses will still need every real call answered. You can check a call or set your front desk today. To have every call answered, hire your Telephonist for £1, and it comes back as calls; then 30p a minute, with no monthly fee.