About 1 in 4 Americans got a deepfake voice call this year, a report found.

A phone rings, and the voice on the other end sounds exactly like a daughter, a grandson, or a boss. The panic in the message is real even when the voice is not. That is the premise behind a wave of AI-generated phone scams that, according to a recent consumer report, has already reached roughly a quarter of the adult population in the span of a single year.

The technology behind these calls has existed in research labs for years, but it has only recently become cheap, fast, and convincing enough to run at criminal scale. What used to require a Hollywood sound studio can now be approximated with a few seconds of audio pulled from a social media video, a voicemail greeting, or a public interview.

What the Report Found

A survey-based report published by AllAboutCookies found that close to one in four Americans said they had received a phone call featuring a cloned or artificially generated voice within the past year. The report ties the surge to the falling cost of voice-cloning software, much of which is now available through consumer-facing apps rather than specialized criminal tooling. Scammers no longer need weeks of a target’s recorded speech; a short clip, sometimes just a few seconds long, can be enough to generate a passable imitation for a scripted call.

The report frames the trend as an extension of long-running impostor scams, particularly the so-called grandparent scam, in which a caller claims to be a relative in trouble and pressures the listener to wire money or buy gift cards immediately. What has changed is the audio itself. Instead of a stranger’s voice attempting an accent or a vague approximation of a loved one, victims now hear something that matches the pitch, cadence, and inflection of a person they know, which short-circuits the skepticism that might otherwise slow the interaction down.

Why the Numbers Keep Climbing

Voice-cloning tools have moved from research demos to widely available commercial products in a short window of time, and that shift shows up directly in fraud complaint data. The Federal Communications Commission has warned that deepfake audio embedded in robocalls and text-message links is making it harder for consumers to distinguish a legitimate call from a fabricated one, since the same synthetic-voice techniques used for entertainment and accessibility tools can be repurposed for impersonation with little technical skill required.

Financial pressure compounds the problem. A caller claiming to be a family member in a car accident, an arrest, or a medical emergency creates urgency that discourages the kind of careful verification people might otherwise perform, such as hanging up and calling the relative back on a known number. Scammers often request payment through channels that are difficult to reverse, including wire transfers, cryptocurrency, or prepaid gift cards, which makes recovery of lost funds unlikely once the transaction clears.

How a Cloned-Voice Call Typically Unfolds

Reports gathered by researchers describe a consistent pattern. The call often opens with a distressed greeting rather than a full explanation, prompting the listener to ask what is wrong and, in doing so, supply the caller with confirmation of identity or emotional cues to exploit. A second voice sometimes joins the call, posing as a lawyer, bail bondsman, or law enforcement officer, adding a layer of institutional authority to the request. The scripted urgency rarely leaves room for the listener to verify details independently, and callers frequently discourage hanging up, citing a supposed ongoing emergency that requires immediate resolution.

Because the underlying audio model only needs a short sample to work from, publicly posted content such as voicemail greetings, wedding videos, podcast appearances, or social media clips have become raw material for scammers. This is part of why the fraud has spread quickly across age groups; older adults with adult children on social media and younger adults with public-facing careers can both become targets, either as the person being impersonated or the person being called.

Building a Verification Habit

Consumer advocates and federal agencies have converged on a similar recommendation: treat any unexpected, urgent phone request for money as a signal to slow down rather than speed up. Hanging up and calling the person back using a number already saved in a contact list, rather than one provided during the suspicious call, removes the caller’s ability to control the follow-up conversation entirely. Family members are increasingly encouraged to agree on a simple verification phrase or question in advance, something a cloned voice would have no way of knowing, to use during any call involving a request for money or personal information.

The Internet Crime Complaint Center, run by the FBI, has continued to log a steady rise in impostor and grandparent-style scam reports as part of its broader tracking of internet-enabled fraud, and it recommends reporting any suspected scam call regardless of whether money changed hands, since aggregated complaint data helps investigators identify active criminal networks and calling patterns. Reporting also creates a record that can assist banks or payment platforms if a victim later attempts to dispute a transaction.

The Broader Shift in Phone Fraud

What distinguishes this current wave from earlier robocall scams is not the underlying con, which remains a variation on urgency and impersonation that has existed for decades, but the fidelity of the impersonation itself. A poorly accented stranger claiming to be a grandchild used to be relatively easy to catch. A voice that sounds correct in tone and rhythm removes one of the most reliable instincts people have used to detect fraud: recognition.

That shift is prompting a broader rethink of how households handle unexpected calls involving money, medical emergencies, or legal trouble. Rather than relying on voice recognition as proof of identity, security guidance increasingly points toward independent verification steps that do not depend on how convincing the caller sounds. As voice-cloning tools continue to improve and become more accessible, the report’s authors suggest that call-based verification habits, not detection skills, will matter most for avoiding losses.

Morning Overview produced this article with AI assistance and reviewed it against the cited sources.

More from Morning Overview