A frightened child calling from jail, a grandchild begging for emergency money or a company executive ordering an immediate wire transfer may sound completely real.
The voice may also be artificial.
Scammers are increasingly using artificial-intelligence tools to imitate relatives, friends, executives and public officials. A few seconds of audio collected from social media, a voicemail, a video or a public interview can sometimes provide enough material to create a convincing synthetic voice.
The technology does not need to reproduce an entire natural conversation perfectly. Criminals often use short emotional phrases, crying, background noise and manufactured urgency to prevent the target from pausing long enough to notice inconsistencies.
A 2026 Hiya survey of more than 12,000 consumers across six countries found that one in four Americans said they had received a deepfake voice call during the previous 12 months. Another 24 percent were unsure whether they could reliably recognize one.
McAfee’s 2026 US research separately found that one in 10 Americans said they had personally experienced a voice-cloning scam.
The “Most People Lose Money” Statistic Needs Context
A widely repeated statistic says that 77 percent of AI voice-scam victims lost money.
That figure comes from a McAfee survey released in 2023. It referred to respondents who reported that they or someone they knew had been targeted by an AI voice-cloning scam. Among that affected group, 77 percent said money had been lost.
It does not mean that 77 percent of everyone who answers a suspicious phone call sends money.
Many people recognize the scam, hang up or verify the supposed emergency through another channel. The statistic nevertheless demonstrates how effective the tactic can become after a target believes the cloned voice belongs to someone they trust.
Among respondents who reported losses in McAfee’s research, 36 percent said the amount was between $500 and $3,000. Seven percent reported losses between $5,000 and $15,000.
Why Hearing a Familiar Voice Is So Persuasive
Traditional impersonation scams require the criminal to convince someone through a fabricated story.
Voice cloning adds what feels like direct proof.
When a parent hears what appears to be a crying son or daughter, the emotional reaction can happen before rational analysis begins. The target may focus on helping the person rather than checking whether the call makes sense.
Scammers deliberately intensify that pressure. They may claim the relative has been arrested, injured in an accident, kidnapped or taken to a hospital.
The fake relative may speak only briefly before another person takes over. That caller may pretend to be a lawyer, police officer, doctor or kidnapper and provide instructions for sending money.
The short appearance by the cloned voice can be enough to make the entire story seem credible.
A Few Seconds of Public Audio May Be Enough
People regularly publish their voices online without considering how the recordings could be reused.
TikTok videos, Instagram clips, YouTube channels, podcasts, workplace presentations and voicemail greetings can all provide samples.
McAfee researchers previously reported that some voice-cloning systems could produce a recognizable imitation from approximately three seconds of audio. Longer and cleaner recordings may improve the result.
A scammer may combine the audio with information gathered from breached databases or public social-media profiles.
Knowing a person’s name, relatives, workplace, travel plans and recent activities allows the caller to construct a believable emergency around the artificial voice.
The voice clone is therefore only one part of the attack. Personal information makes the story convincing.
People Are Not Good at Identifying Synthetic Voices
Listening for a robotic tone is no longer a dependable defense.
A small 2026 controlled study found that participants correctly distinguished AI-generated and human voices only 37.5 percent of the time in simulated voice-phishing scenarios. Most AI clips were classified as human by a majority of listeners.
A larger 2026 study involving 35,532 judgments found that people’s accuracy on synthetic audio remained limited, while their ability to recognize genuine human recordings had also declined. Participants were increasingly suspicious of real voices as well as fake ones.
These findings suggest that people should not base financial decisions on whether a caller “sounds real.”
Modern systems can reproduce pauses, emotion, cadence and vocal variation that people traditionally associate with genuine speech.
AI Makes Large-Scale Voice Phishing Cheaper
In the past, voice phishing required human callers who could speak convincingly, react naturally and spend time with every target.
AI can reduce that cost.
A July 2026 research paper tested voice-phishing scenarios produced with several advanced AI voice systems. Across five scam categories, 16.5 percent of participants said they would or might comply. In relative-in-distress scenarios, the rate reached as high as 36 percent.
The researchers concluded that the immediate danger was not necessarily that AI voices were superhumanly persuasive. The larger threat was that criminals could automate believable calls at low cost and contact far more people.
Even when only a small percentage comply, thousands of automated calls can become profitable.
Government Officials Have Also Been Impersonated
Voice cloning is not limited to family-emergency scams.
The FBI warned in May 2025 that malicious actors were impersonating senior US officials through text messages and AI-generated voice messages. Targets included current and former government officials and their contacts.
The campaign attempted to build trust, gain access to accounts and obtain sensitive information.
The FBI issued another warning in December 2025 stating that senior officials continued to be impersonated through malicious text and voice communications.
The agency emphasized that a message appearing to come from a familiar or prominent person should not automatically be considered authentic.
The Scam May Begin With a Text Message
Not every voice-cloning scheme starts with a live telephone call.
The target may first receive a text from an unfamiliar number claiming that a relative has lost or damaged their phone.
After an exchange of messages, the scammer may send a voice note containing the cloned voice. The audio can make the new number appear legitimate.
A business target may receive an email or chat message followed by a voice recording supposedly from a chief executive or manager authorizing a payment.
This staged approach gives the criminal more control. A prerecorded voice note does not have to respond naturally to unexpected questions.
Scammers Create Urgency and Secrecy
The story commonly includes a reason the target must act immediately.
The supposed relative may need bail money before a deadline, hospital payment before treatment or ransom before the caller hangs up.
The criminal may also warn the victim not to contact anyone else. A fake lawyer might claim that discussing the case would violate a court order, while a kidnapper may threaten the relative if police are called.
These instructions are designed to isolate the target from people who might recognize the fraud.
Requests for payment through cryptocurrency, gift cards, cash couriers, wire transfers or person-to-person payment applications are major warning signs.
Legitimate courts, hospitals and police departments do not normally demand secret emergency payments through these methods.
Hang Up and Call the Person Directly
The most effective response is to end the incoming call and contact the supposed victim through a number already stored in your phone.
Do not call a number provided by the person making the demand.
When the relative does not answer, contact someone physically close to them, such as a spouse, roommate, colleague or school administrator.
A genuine emergency will remain genuine after a short verification process. A scammer will try to prevent that verification.
The FBI advises independently confirming the identity of anyone making an unusual request and never sending money or sensitive information based solely on an incoming message.
Create a Family Safe Word
Families can agree on a private word or phrase that must be given during an emergency involving money.
It should not be a pet’s name, birthday, hometown or other information visible online.
The word should be unusual enough that a criminal cannot guess it but easy for family members to remember under stress.
A caller who avoids the question, claims to have forgotten the word or attempts to redirect the conversation should not be trusted.
The safe word is not perfect. A family member could reveal it accidentally or through another scam. It still provides an additional verification step that cannot be copied from a voice sample alone.
Ask a Personal Question the Internet Cannot Answer
Another defense is to ask something known only to the real person.
The question should concern a private shared experience rather than information available through social media.
However, this method must be used carefully. Criminals may possess extensive personal data from breached accounts, public posts or previous conversations.
A direct callback through a trusted number remains stronger than attempting to authenticate someone during the suspicious call.
Do Not Trust Caller ID
A familiar name or telephone number on the screen does not prove the call is genuine.
Caller-ID information can be spoofed, making a call appear to originate from a relative, bank, police department or government agency.
The number may also belong to an account that has been compromised.
Verification should therefore happen through a separate communication channel initiated by the recipient.
A person can hang up and call the institution using the number printed on an official statement, payment card or government website.
Businesses Need Payment Verification Rules
Voice cloning creates a major risk for companies because employees may be accustomed to receiving instructions from executives by telephone or messaging applications.
A finance employee could hear what seems to be the chief executive authorizing an urgent transfer and fear delaying an important transaction.
Businesses should require a second approval channel for unusual payments, new bank details or changes to vendor accounts.
A voice message should never be sufficient authorization for a large or unexpected transfer.
Employees should call the executive through an established internal number, verify the request through approved software or obtain confirmation from another authorized person.
What to Do After Sending Money
Victims should act immediately.
The bank, card issuer, payment service, wire-transfer company or cryptocurrency platform should be contacted and told that the transaction resulted from fraud.
Some payments can be paused or recovered when reported quickly, although recovery is not guaranteed.
Victims should preserve the phone number, voice message, text conversation, payment receipt and any account details supplied by the scammer.
The incident can be reported to the FBI Internet Crime Complaint Center and the Federal Trade Commission’s fraud-reporting service.
Local police may also need a report, especially when a bank or insurer requests documentation.
Do Not Be Embarrassed to Report the Scam
Voice-cloning schemes are built to exploit ordinary trust and emotional reactions.
Victims may feel ashamed because they believe they should have recognized the deception. That embarrassment helps criminals by reducing reporting and allowing the same operation to target more people.
A convincing voice, accurate personal details and a frightening emergency can mislead intelligent and cautious individuals.
Reporting the incident gives investigators payment information, telephone numbers and other evidence that may connect multiple victims.
It also alerts relatives and colleagues before the same criminal attempts to impersonate another person in the network.
How to Reduce the Amount of Voice Material Online
People do not need to stop speaking publicly, but they can reduce unnecessary exposure.
Social-media accounts can be reviewed to determine which videos are visible to strangers. Old public clips containing long, clean speech samples may be made private where appropriate.
Voicemail greetings do not need to include a full name or lengthy personal message.
Parents should consider how much audio and personal information they publish about children, especially when posts also reveal names, schools, routines and family relationships.
Reducing public information makes it harder for scammers to create a targeted story, though it cannot eliminate the risk.
Regulation and Technical Detection Are Still Developing
The Federal Trade Commission has warned that voice cloning can enable fraud, extortion and impersonation. It launched a Voice Cloning Challenge to encourage technical approaches that prevent or detect harmful misuse.
In August 2025, Consumer Reports delivered a petition signed by more than 75,000 people asking the FTC to hold voice-cloning companies accountable when their tools enable fraud and impersonation.
Technology companies and mobile platforms are also developing call-screening and deepfake-detection features.
Detection may help, but it should not become the sole defense. New voice models can change faster than detection systems, and no automated warning will identify every synthetic call correctly.
The Main Safety Message
AI voice-cloning scams are becoming more common, and a familiar voice should no longer be treated as proof of identity.
One in four Americans surveyed by Hiya said they received a deepfake voice call during the previous year. McAfee’s earlier international survey found that 77 percent of respondents who reported being affected by an AI voice scam said money was lost.
That does not mean most people who answer an unknown call lose money. It means the scam can be highly effective once the target believes the voice and follows the caller’s instructions.
The safest response is simple: hang up, call the person through a trusted number and verify the emergency independently.
No voice regardless of how familiar, emotional or convincing it sounds—should be enough on its own to justify sending money.
