AI tools make deepfakes easier to create and harder to detect than ever before. 

It’s important to note that anyone can fall victim to deepfakes, from individuals to global corporations. Losses range from a few hundred to tens of millions of dollars.

But just how common are deepfakes? Who’s most at risk? And what can people do to protect themselves against this growing threat?

This article covers the latest available statistics and trends related to deepfakes in 2026.

Editor’s Picks

How Common Are Deepfakes

There are seven deepfake attacks every day (Pindrop)

That’s a significant increase from 2024, when Pindrop recorded a deepfake attack every one or two days. 

There are 2,031 verified deepfake incidents every quarter (Resemble)

Resemble notes that this is a 317% increase compared to earlier in 2025. 

The number of deepfake files will hit 8 million by the end of the year (Keepnet Labs)

That’s an increase of 1,500% since 2023. 

70% of YouTube videos involve AI technology (UTSA)

That doesn’t mean they all involve deepfakes specifically. But deepfake tools now allow people to create entire realistic-looking videos, including some that mimic real people.

Prevalence of Deepfakes Among Individuals

The average American encounters 2.6 deepfakes daily (McAfee)

Deepfakes are becoming increasingly common. People who spend more time online are more likely to encounter them. In fact, young adults aged 18-24 encounter an average of 3.5 deepfakes every day.

46% of fraud experts have encountered synthetic identity fraud (UNESCO)

Synthetic identity fraud includes but isn’t limited to deepfakes. But it’s the type of deepfake-related fraud that’s most commonly encountered by experts. 37% of experts have encountered voice deepfake fraud, and 29% have encountered video fraud. 

10% of people have received a cloned audio call (Security.org)

AI tools have made it very easy to clone voices from very little data.

And 25% have experienced a voice cloning scam or know someone who has (McAfee)

This is based on a survey of 7,000 people. 

53% of finance professionals have been targeted by deepfake fraud attempts (Medius)

Finance professionals may be particularly common targets because they regularly handle relatively large volumes of money. 

72% of Americans have seen fake celebrity or influencer endorsements (McAfee)

These aren’t necessarily always created using deepfake technology, but often are. Scammers use public figures to exploit our tendency to trust people we know and who we believe are in positions of authority.

Prevalence of Deepfakes Among Organizations

There were at least 78 deepfakes targeting elections in 2024 (Wired/Princeton)

WIRED’s AI Elections Project tracked deepfakes in key elections around the world in 2024. They discovered at least 78, many (but not all) of them malicious. Princeton then further analyzed the deepfakes. 

The intent behind the deepfakes ranged widely, from misinformation to electoral campaigning to outright scams in which deepfaked versions of politicians promoted fraudulent financial platforms. Little wonder, then, that the World Economic Forum identified deepfake-related disinformation as one of the biggest risks to global democracy.

10% of organizations have dealt with attempted deepfake fraud (Security.org)

Deepfakes aren’t currently as common as other forms of fraud, like advance fee or online shopping scams.

And another 10% are uncertain if they have faced attempted deepfake fraud (Business.com)

Many organizations have likely faced deepfake attacks without realizing it. 

But just 3% of executives say they faced a deepfake attack in 2024 (Business.com)

Perhaps some executives aren’t aware of low-level attacks, don’t want to admit to having been targeted, or are simply harder to reach.

37% of organizations worldwide have experienced a deepfake voice fraud attempt (Medius)

This is notably higher than the estimate provided by Business.com. 

17% of hiring managers have encountered a job applicant using deepfake technology during a video interview (Resume Genius)

AI tech has made it possible for scammers to circumvent every step of the hiring process for remote jobs, including faking an entire video interview. According to Pindrop, speech generation tools can now produce realistic voices in near real-time. 

Millennial managers were the generation most likely to have encountered deepfake tech in video interviews, with 24% having done so. Millennials were followed by Gen Z managers (16%), Boomers (14%), and Gen X (10%).

Growth in Deepfakes

The number of deepfake attacks doubles every month (Resemble)

In Q2 2025, Resemble tracked 487 discrete deepfake incidents. That’s an increase of 312% year-over-year and 41% from the previous quarter. 

Deepfake attacks have increased by 1,300% (Pindrop)

Pindrop describes the increase in deepfake attacks as “a flood of AI-powered deception.”

51.6% of executives expect deepfake attacks against their organizations to increase in size and number (Deloitte)

Executives don’t see the growth in deepfakes slowing down. In this survey, executives were referring to deepfake attacks against their organizations’ accounting and financial data specifically.

Deepfake incidents increased by 317% between Q2 and Q3 2025 (Resemble)

In Q2 2025, Resemble verified 487 deepfake incidents. The following quarter, it identified 2,031 incidents, a significant increase. 

Money lost to deepfake fraud increased by 73.6% in the first half of 2025 (Resemble)

In Q1, $200 million was lost. In Q2, that increased to $347.2 million. 

25% of job candidates will be fake by 2028 (Gartner via CNBC)

People are increasingly using deepfake technology to get hired under fake identities. Why? In some cases, the impostors are working for foreign governments, looking to steal information or funnel money back home.

For example, in 2024, the US Justice Department alleged that over 300 companies had unknowingly hired impostors who were connected to North Korea. The impostors used deepfakes to get hired and managed to collect over $6.8 million in total. 

By 2026, 30% of enterprises will consider ID verification solutions unreliable due to deepfakes (Gartner)

As deepfake technology continues to become more sophisticated, “organizations may begin to question the reliability of identity verification and authentication solutions,” an analyst at Gartner explains, because “they will not be able to tell whether the face of the person being verified is a live person or a deepfake.”

Types of Deepfake Attacks

46% of deepfake attacks involve video (Variety)

In Q1 2025, video deepfakes were the most common type. They were followed by images (32% of incidents) and audio (22%). 

Deepfakes are linked to 20% of biometric fraud incidents (Entrust)

Biometric fraud involves people tricking or bypassing biometric security checks, such as selfies or ID verification.

They’re used in 22% of biometric fraud cases involving motion detection (Entrust)

Deepfakes are becoming a commonly used technique to bypass biometric security checks involving motion, such as recording a video. Still, the cheapfake option of recording a video of a video on a screen is still far more popular, being involved in 63% of biometric fraud cases.

As recently as 2023, recording a video of a photo on a screen was still relatively common. The technique was used in 34% of all biometric fraud cases. But detection technology has become more sophisticated, making this technique less common; in 2025, it was involved in just 9% of biometric fraud cases. 

But only in 8% of biometric fraud cases involving taking a selfie (Entrust)

Deepfakes usually aren’t necessary to bypass a biometric security check that only involves taking a selfie. The most common technique is taking a photo of a photo on a screen. Another technique, which is declining in popularity, is taking a photo of a photo ID document. 

29.9% of deepfake incidents occur on YouTube (Resemble)

YouTube is the most common site for deepfake incidents, at least in Q3 2025. It’s closely followed by Instagram (26.8% of cases), then Facebook (18.8%), TikTok (18.3%), and WhatsApp (6.3%). 

31% of fake celebrity endorsements involve giveaways or sweepstakes (McAfee)

People are, unfortunately, often willing to believe that a celebrity is inviting them to participate in a giveaway or sweepstakes. If something sounds too good to be true, then it probably is. 

Giveaways were the most common tactic. They were followed by crypto scams or fraudulent investment and trading advice (30%), weight loss programs or supplements (25%), skincare or beauty products (24%), and must-have gadgets (22%).

18% of songs uploaded to Deezer every day are completely AI (Resemble)

Deepfakes aren’t always used to defraud people. They are often used to create synthetic media to generate money via streaming services. 

Costs of Deepfakes

Deepfake fraud losses in the US will reach $40 billion by 2027 (Deloitte)

That’s based on a compound annual growth rate of 32% from 2023, when the US lost $12.3 billion to fraud. 

77% of people who received a message from a deepfake voice clone lost money (McAfee)

Only 10% of people have received messages from deepfake voice clones, according to this survey. But of those who did receive such messages, over three-quarters lost money. This underlines just how dangerous deepfakes can be.

Of those who lost money to deepfake audio clones, 36% lost between $500 and $3,000. 7% lost between $5,000 and $15,000.

Deepfake fraud cost $547.2 million in the first half of 2025 (Variety & Resemble)

That’s based on $200 million lost in Q1 and $347.2 million in Q2, as calculated by Resemble. Even without continued growth, that would mean over $1 billion directly lost to deepfake fraud in 2025 alone.

UK victims of fake celebrity endorsements lose £450 on average (McAfee)

Deepfake technology has allowed scammers to create increasingly realistic fraudulent celebrity endorsements.

Who Is Targeted By Deepfakes

38% of deepfake fraud incidents in Q1 2025 were in North America (Variety)

North America was the most common location for deepfake fraud in early 2025. It was followed by Asia (27% of incidents) and Europe (21%). 

However, almost two-thirds of incidents crossed borders. Globalization has led to sophisticated cross-border criminal networks that aren’t restricted to single nations. 

48% of deepfake incidents in Q3 2025 used celebrities (Resemble)

Deepfake scammers often use the likeness of celebrities, because people are more likely to trust a figure they (think they) recognize and know. Celebrities were the most popular category of deepfake incidents tracked by Resemble, followed by corporate deepfakes, political, and minors. 

Women are targeted by deepfakes 4.5x more often than men (Resemble)

This is based on 2,031 verified incidents tracked by Resemble in Q3 2025. Female targets made up 34.6% of cases, while men made up just 7.7% of cases. Half of all incidents targeted a non-person (including businesses).

In the previous quarter, 84% of deepfake attempts targeted women.

48.3% of deepfake attacks target businesses (Resemble)

Businesses are one of the most common targets of deepfake attacks. That may be because they usually have more to steal than individuals. And employees may be more likely to transfer company funds than their own personal money.

The most commonly exploited star is Taylor Swift (McAfee)

McAfee ranked the celebrities and influencers whose likeness is most often exploited by deepfake scammers around the world. Taylor Swift topped the list overall, while streamer Pokimane was the most commonly exploited influencer.

1.1% of deepfake incidents in Q3 2025 involved Will Smith (Resemble)

Will Smith was a surprisingly popular likeness. After him, the most commonly exploited people were all politicians: Barack Obama, Donald Trump, Marco Rubio, and Alexandria Ocasio-Cortez.

Malicious actors aren’t always looking for financial gain. In July 2025, attackers used deepfakes of Marco Rubio to try to influence government officials. 

Ability to Spot Deepfakes

Just 0.1% of people correctly identified deepfakes (iProov)

To be clear, this figure refers to the people who correctly identified all images and videos in a test. But they were primed on deepfakes, which means they were likely more accurate than the population at large. 

Just 29% of people feel confident in their ability to identify a deepfake (McAfee)

While studies usually find that people are overconfident in their ability to identify deepfakes, this McAfee study provided a somewhat different picture. Still, only 21% of people actually admit to having low confidence in their ability to correctly identify deepfakes, leaving a majority who aren’t confident or willing to admit to being unconfident. 

People correctly identify an AI-generated voice 60% of the time (Nature)

This 2025 study asked participants to distinguish between a real speaker and a deepfake clone of a real speaker.

People identified deepfake audio 73% of the time (UFL)

In this study, people were more successful at identifying deepfake audio. They were not trained about deepfakes before being tested.

Interestingly, people often guessed incorrectly due to underestimating how sophisticated deepfakes could be.

They can identify deepfake images 86% of the time (ACM)

In this study, participants were relatively successful at identifying deepfake images. Their accuracy improved over time, as they were shown more and more images. 

68% of deepfakes are nearly indistinguishable from real media (Variety)

As AI tools become more sophisticated, it will become more and more difficult for people to distinguish between synthetic and genuine media without the use of other tools.

Accuracy rates of deepfake voice detection software can be as low as 38.2% (University of Waterloo via Pindrop)

Various organizations have developed technologies to detect synthetic voices. However, there are various tricks that malicious actors can use to bypass these defenses, including replacing silences with actual recorded silences and adding various audio effects. 

The University of Waterloo tested various systems with different kinds of attacks. In many cases, the defense systems had a roughly 50-50 accuracy rate. But in some cases, accuracy rates were as low as 38.2%.

People are correct about their ability to identify deepfake videos just 18.75% of the time (UVA)

This study found that people often overestimate their ability to distinguish between deepfake videos and genuine ones. Participants were shown 16 videos, some real and some fake. In almost all cases, their confidence in their ability to identify which video was real was significantly higher than their actual accuracy rate. In fact, their accuracy rate was roughly 50-50 — effectively no better than randomly guessing — for 11 out of the 16 videos. 

Notably, participants also displayed a bias toward authenticity. Despite being told beforehand that half of the videos were real and half fake, participants guessed the video was authentic 67.4% of the time. 

People’s confidence in their ability to accurately detect a deepfake declines by 17.3% after they are exposed to one (Sage)

In this study, participants were asked to identify real and fake images. When they were exposed to a deepfake, their confidence in their ability to correctly distinguish between the two significantly declined. 

Strangely, people’s confidence also declined when they were exposed to a genuine photo. Perhaps this is another sign that people tend to overestimate their ability to identify deepfakes; the number of people who had to lower their confidence seems to outweigh the number who became more confident after being tested. 

But people still rate their deepfake detection skills at 60%, regardless of the accuracy of their answers (iProov)

This tendency toward overconfidence, which iProov calls a “false sense of security,” is particularly prevalent among younger people.

People are 36% less likely to correctly identify deepfake videos compared to images (iProov)

Deepfake videos are apparently more difficult to distinguish than images. That might be due to our tendency to underestimate the sophistication of deepfake technology, which can now generate extremely realistic videos.

How Organizations Defend Against Deepfakes

38% of executives say their organization has trained employees on the risks of deepfakes (Business.com)

Training is the most popular defense method against deepfakes. It’s followed by communicating with stakeholders (35% of executives say their company has done it), establishing anti-deepfake protocols (32%), and utilizing technology (31%).

13% of companies have extensive anti-deepfake protocols (Business.com)

That’s a surprisingly small proportion, considering how common (and costly) deepfake attacks are becoming.

How Individuals Defend Against Deepfakes

25% of people verify information with alternative sources (iProov)

As deepfakes become more realistic, turning to alternative sources becomes all the more necessary when determining if information is real or fake. Yet three-quarters of people fail to do that.

And just 11% conduct critical source analysis (iProov)

Even fewer people critically analyse the source of potentially fake information. This is another crucial technique for determining if something is genuine; when anyone can create realistic images, video, and audio, verifying trustworthy sources is extremely important.

29% of people take no action when they encounter a deepfake (iProov)

That’s probably a mix of laziness and a lack of knowledge. To the latter point, 48% of people admit they don’t know how to report deepfakes. 

Social media platforms should have a process for reporting deepfakes. You can also report deepfakes to law enforcement, either at the local or national level, such as with the FBI’s Cyber Watch. 

Legislation Addressing Deepfakes

46 states have enacted deepfake legislation (Programs.com)

169 laws addressing deepfakes have been passed since 2022. In 2025 alone, state lawmakers introduced almost 150 bills related to deepfakes. They addressed a range of issues, from political misinformation to impersonation of real people. 

The state that has passed the most of these laws is California, with 19. It’s followed by Texas (11), Illinois (8), and Tennessee (8). Meanwhile, Alaska, Ohio, Missouri, and Wyoming haven’t passed any laws addressing deepfakes since 2022. 

146 state bills addressing deepfakes were introduced in 2025 (Programs.com)

Lawmakers are fighting to keep up with the rapid evolution and spread of deepfake technology. However, many bills that are introduced aren’t enacted into law. For example, in 2024, state lawmakers introduced almost 350 bills, of which only 79 became law. 

Creating Deepfakes Statistics

The price-per-minute to buy a deepfake video ranges from $300 to $20,000 (Kaspersky via Security.org)

The cost of creating a deepfake video depends on various factors, including the length, the amount of data available to train the model, and how seamless you want the product to be.

But in 91% of cases, creating a convincing deepfake takes just $50 and 3.2 hours (Resemble)

With the proliferation of powerful AI tools, malicious actors can create reasonably realistic deepfakes with only a little time and money.

3 seconds of audio is all that’s needed to create a deepfake clone with an 85% voice match (McAfee)

Malicious actors only need a tiny amount of data to create a reasonably convincing clone of a person’s voice.

Vulnerability to Deepfakes

22% of people have never heard of deepfakes (iProov)

Unsurprisingly, older generations are less likely to be aware of deepfakes. 30% of people aged 55-64 and 39% of people aged 65 and over have never heard of them.

People think an AI-generated voice sounds the same as a real voice 80% of the time (Nature)

In this 2025 study, researchers used a deepfake creation tool called ElevenLabs. It was used by a group of people to create a robocall of Joe Biden, which they used to try to suppress the vote in New Hampshire’s presidential primary election. 

Hundreds of speakers were asked to record their voices, which were used to create clones of each of them. Participants then had to distinguish between real recordings and cloned versions. They struggled to do so.

25% of company executives have little or no familiarity with deepfake technology (Security.org)

Despite the clear threat posed by deepfakes, a significant proportion of people (including company leaders) are unfamiliar with the technology behind them. 

Still, a majority of executives are somewhat or moderately familiar with deepfake technology.

53% of US adults share their voice data in recorded notes or online at least once per week (McAfee)

The more personal data you make available, the easier it is for malicious actors to create a convincing deepfake clone of your voice. Still, the majority of people record or share their voice data often. 

45% of people would respond to a voicemail or voice message from a friend or loved one in need of money (McAfee)

Deepfake scams prey on our willingness to trust and desire to help those close to us. A little under half of all adults say they would respond to a message from a loved one asking for money, which is why deepfake scams so often use this exact tactic.

31% of business leaders think deepfakes haven’t increased their fraud risk (Business.com)

This might be surprising, considering the increasing number of high-profile deepfake fraud cases. 

For example, in 2024, an employee at a UK engineering firm transferred $25 million after a video call with senior management. But that call had been faked; the senior officials were all deepfake clones, created by criminals who successfully stole the money. 

61% of executives say their business has no protocols to address deepfake risks (Business.com)

Developing anti-deepfake protocols takes time and costs money. But the potential costs of a successful deepfake attack are likely far higher.

Even more companies — 80% — lack protocols for handling deepfake attacks, specifically.

43% of finance professionals who’ve been targeted by deepfake fraud attempts admit to falling victim (Medius)

That’s a concerningly high rate of success, considering finance professionals should theoretically be aware of the dangers of fraud. Perhaps it’s a sign that more organizations should devote resources to training employees on the dangers of deepfake attacks.

87% of professionals say they would make a payment if told to by their CEO or CFO (Medius)

Of course, without some kind of additional verification (like a verbal password), it’s increasingly difficult for employees to know whether they are speaking to their actual CEO or a deepfake clone. Scammers often exploit our tendency to obey figures of authority, especially ones we are familiar with.

Miscellaneous Deepfakes Statistics

Cryptocurrencies are involved in 60% of all deepfake fraud incidents (Entrust)

Crypto is commonly used by scammers because it’s hard to trace and often impossible to reverse. The FTC warns that crypto payments “typically are not reversible” and “do not come with legal protections.”

74% of people worry about the societal impact of deepfakes (iProov)

The top concern is fake news and misinformation, which 68% of people worry about. This concern is particularly common in older generations; 82% of people aged 55 and over are worried about the spread of disinformation.

49% of people trust social media less after learning about deepfakes (iProov)

The spread of deepfakes on social media seems to be lowering trust in online platforms. That’s concerning, as the majority of Americans now get news from social media platforms. In fact, more people use social media to get news than TV and news sites. 

However, Americans still trust news organizations far more than social media sites. 70% of US adults trust local news organizations and 56% trust national news organizations, while just 37% trust social media sites.