Businesses are embracing AI in calls, and so are scammers. How can you trust a voice on the line?

Americans reported losing $665 million to phone call fraud in the first half of 2026, according to Federal Trade Commission data as of June. Nearly $128 million of that total was attributed to scammers impersonating a business. With the advances in AI, voice scams are getting harder to detect, despite growing exposure to AI in everyday life.

Meanwhile, a 2026 Sinch survey of 2,527 decision-makers across 10 countries found that nearly half of North American companies have started using AI on their phone lines, further blurring the line between legitimate and fraudulent calls.

Below, Sinch examines what this means for consumers and the businesses they trust, and how confidence in the call experience can be rebuilt.

Scam calls are using AI, and AI awareness is no protection

Most people have seen voice-cloning scams in the news: a grandchild calls needing bail money, or a sister in distress, stranded overseas. The voice sounds familiar, and before the victim knows it, their money is gone. Microsoft researchers demonstrated in 2023 that their VALL-E system could reproduce someone's voice from a three-second recording, preserving characteristics such as emotion and the recording’s acoustic environment. Anyone who’s posted a video or audio clip online may have left material that scammers could use to imitate their voice. And that number on the screen? It can be spoofed.

The threat extends beyond relatives and audio. In 2024, a finance employee at the engineering firm Arup joined a video call with people he believed were his colleagues, including the chief financial officer, and authorized $25.6 million in transfers. Hong Kong police later confirmed the other participants on the call were deepfakes.

Voice scams have been around for decades. What deepfakes add is believability and scalability. Some scam operations run like businesses, and staffing once limited the number of calls they could make. AI is helping to remove that constraint.

As demonstrated in 2025 research from Rutgers University, AI can now handle the tricky part of a scam conversation: fielding an unscripted question, absorbing pushback and staying plausible when the person on the other end turns suspicious, all without human intervention. In a simulated environment, the AI models completed scam dialogues successfully between 43% and 74% of the time.

AI scams are becoming easier and cheaper to produce, potentially allowing them to reach more people. The question is whether people can spot them. It's easy to believe that frequent AI use provides some level of protection against AI scams, but research suggests it isn’t the case. According to 2026 research from Harvard and Meta, published in Expert Systems with Applications, nearly a third of AI voices went unrecognized in an experiment that asked participants to judge whether recorded scam calls were using human or machine voices, and frequent AI users did no better at spotting a synthetic caller than people who never use it.

But as legitimate businesses begin using AI to handle phone calls, an AI-sounding voice alone is becoming a less reliable signal of fraud.

AI voice has already earned its place in customer communications

Fully autonomous voice AI is beginning to enter business calls. Among North American enterprises surveyed, 64% already had AI communications agents in production, and 97% planned to increase their AI communications investment in 2026.

Specifically on voice, 41% of enterprise leaders globally named fully autonomous AI call handling as the single biggest opportunity in voice, while 17% cited outbound AI voice use for appointments, reminders and proactive outreach.  In North America, 46% said they already use AI in their voice communications.

Consumers are fairly evenly split on whether they want an AI on the other end of the line when they call a company. In a Sinch survey of 2,501 shoppers conducted ahead of the 2026 holiday season, 48% said they'd trust a human over a voice AI agent to resolve issues, while 52% were open to AI voice as long as they could be transferred to a person if needed.

More from this section

+3

Team searching for Noah’s Ark builds 3D model of Turkish mound

+5

Still stinging? Bee’s stinger pumps venom after bee long gone

+2

Can eating a pack of peanuts daily help improve your memory?

But being open to an AI agent when you call a company is one thing. Receiving an unexpected call is another.

How businesses are building trust directly into the call experience

If consumers can no longer reliably tell who's really calling, the businesses initiating the call have to make it clear. Solutions exist to prove a caller's identity before the recipient picks up.

It starts with verifying the number. In the U.S. and Canada, under a framework called STIR/SHAKEN, the phone company placing a call digitally signs the caller ID information, and the receiving phone company verifies that signature to confirm the call really is from the number displayed. Phone companies, and the analytics firms they work with, weigh that information alongside calling patterns and complaints to decide how to handle the call. A call without a valid signature, or with a lower level of verification, may be more likely to be blocked before it rings or labeled as spam on the recipient's screen.

The system isn't perfect. It works on internet-based phone networks, so some legitimate calls still arrive unauthenticated. Legitimate calls can also be labeled as spam because of how they're placed, such as many calls in a short burst or calls from a brand-new number. Businesses can add another layer by marking numbers they only use to receive calls, such as a customer service line, as "do not originate," so any outgoing call claiming to come from one of those numbers can be blocked. While this framework doesn't prevent spoofing entirely, it makes it significantly harder.

Then, there’s what consumers actually see. A separate system called CNAM lets a business register its name so it can appear on the recipient's screen instead of a bare number. The recipient's phone company looks the name up when the call comes in, but not every phone company does this on every call, so the name doesn't always appear. And because the name is looked up by number, a scammer who spoofs a company's number may also trigger the display of that company's real name. On its own, a name on the screen isn't proof of who's calling.

A standard called Rich Call Data takes it further, letting businesses include their logo and a brief reason for the call in the signed caller identity information. Those details can appear on the recipient's screen before the call is answered, offering some reassurance about who’s calling.

Support for this capability, however, varies across handsets and carriers, so many calls don't display these elements. Plus, a call with no branding isn't evidence of fraud, because many legitimate calls still arrive without it.

What consumers can do to protect themselves

When in doubt about an incoming call, the most useful thing you can do is also the simplest: hang up, or just don't pick up, and call back using a trusted number. Not the number that called, but the one on the business card, the bank statement or the business app. Scammers can still spoof a legitimate number, but they can't redirect a call you place yourself.

A few other things worth considering:

• Treat urgent demands for money via cryptocurrency, wire transfers or gift cards as a massive red flag.
• Consider restricting access to your social media profiles to "friends only" to limit access to audio content.
• Recorded or artificial-voice calls generally must identify the business or person responsible at the beginning of the message and provide a callback number during or after it. In 2024, the FCC confirmed that AI-generated voices count as artificial under the Telephone Consumer Protection Act. A synthetic voice that never says who's calling isn't following these rules.
• Turn on the blocking tools already available on the phone. Most carriers and handsets offer them at no extra cost. In September 2026, the FCC published step-by-step guides for iPhone, Android and the major U.S. carriers in one place at fcc.gov/blockit.
• Be wary of any caller asking you to share your full card number, a one-time code or a password. These requests are red flags, regardless of how the caller sounds.
• Report any suspicious activity at reportfraud.ftc.gov. Every business day, the FTC releases the phone numbers reported by consumers so phone companies can use them to help block unwanted calls.

Tackling voice fraud is a combined effort

Beyond financial losses, voice scams also wear away trust in the calls that matter, like a doctor contacting you about a prescription or your bank calling about an actual suspicious charge.

Anti-fraud measures and the people working around them will always be racing each other, and no single fix can tackle voice scams altogether. Carriers, businesses and consumers each have a role to play in curbing AI voice fraud.

This story was produced by Sinch and reviewed and distributed by Stacker.