2023 STATE OF DEEPFAKES

Realities, Threats, and Impact

Photo by Kirill Balobanov / Unplash

In an era of rapid technological progress, deepfakes have become a concern in digital development. This report is a milestone in an ongoing series that thoroughly examines the ins and outs of deepfake technology, revealing its evolving capabilities and imminent threats.

Through rigorous research, we aim to empower individuals and organizations with the knowledge to responsibly navigate the complex world of deepfakes. Our work stems from a comprehensive analysis of 95,820 deepfake videos, 85 dedicated channels across online platforms, and a meticulous review of over 100 websites linked to the deepfake ecosystem.

This 2023 state of deepfakes: realities, threats, and impact review is a powerful tool for fostering awareness of deepfake technology and protecting the masses from its negative effects. Join us as we uncover the layers of the state of deepfakes, revealing realities, highlighting threats, and exploring the technology's profound impact.

KEY FINDINGS

01.

The total number of deepfake videos online in 2023 is 95,820, representing a 550% increase over 2019.

02.

Deepfake pornography makes up 98% of all deepfake videos online.

03.

99% of the individuals targeted in deepfake pornography are women.

04.

South Korean singers and actresses constitute 53% of the individuals featured in deepfake pornography and are the most commonly targeted group.

05.

94% of those featured in deepfake pornography videos work in the entertainment industry.

06.

One in every three deepfake tools allow users to create deepfake pornography.

07.

It now takes less than 25 minutes and costs $0 to create a 60-second deepfake pornographic video of anyone using just one clear face image.

08.

48% of surveyed US men have seen deepfake pornography at least once.

09.

74% of deepfake pornography users don't feel guilty about it.

Our Most Helpful Articles

OVERVIEW OF CURRENT STATE

Deepfake technology represents a transformative advancement in artificial intelligence and digital manipulation. It refers to the use of deep learning algorithms to create compelling and often deceptive media content, such as videos, audio recordings, or images, that appear to feature real people saying or doing things they never did.

In recent years, deepfake technology has surged to unprecedented levels. What began as a niche and technically complex endeavor has rapidly evolved into a widespread phenomenon. However, as technology evolves, it drastically blurs the lines between creativity and trickery, shaping reality in astounding yet concerning ways.

In this section, we delve into the contemporary deepfake landscape. We provide a clear introduction and compelling statistics portraying the technology's extensive influence and prevalence.


total-percentage-of-deepfake-video-online

Deepfake pornography makes up 98% of all deepfake videos online.

An astonishing 98% of the deepfake videos found online are explicitly pornographic. This eye-opening figure highlights the distressing misuse of deepfake technology, particularly for generating explicit and non-consensual content.

This trend has profound implications for privacy and consent in the digital age. With the ability to copy and manipulate an individual's likeness without permission, deepfakes threaten personal integrity and digital well-being.

Total deepfake videos online in 2023

95,820
chart-total-deepfake-videos-online-in-2023

There are 550% more deepfake videos online in 2023 than in 2019.

In 2023, our research indicates that the digital realm will host a staggering 95,820 deepfake videos. This number reflects a 550% surge compared to measurements recorded in 2019.

Such exponential growth indicates a rapid proliferation of deepfake technology and its escalating integration into our online experience.

7-out-of-the-top-10-pornography-websites

7 out of the top 10 pornography websites host deepfakes

We evaluated the prevalence of deepfake pornography by observing the number of posted deepfake pornographic videos and the number of websites that hosted these videos.

Our research reveals that seven of the top ten pornographic websites host deepfake content.

Total video views across top 10 dedicated deepfake porn websites

303,640,207
total-video-views-across-top-10-dedicated-deepfake-porn-websites

The total number of video views across the top ten dedicated deepfake pornography websites is 303,640,207

Across the top ten dedicated deepfake pornography websites, the cumulative video views total 303,640,207. This substantial figure provides a clear glimpse into the extensive consumption and dissemination of deepfake-generated pornographic material.

The broad reach of these views sparks the need for conversations around societal attitudes toward such content. Ethical considerations also need to be focused on, particularly those surrounding deepfake pornography creation and distribution.


Total traffic across top 10 dedicated deepfake porn websites

34,836,914
total-traffic-across-top-10-dedicated-deepfake-porn-websites

The monthly traffic of the top ten dedicated deepfake pornography websites is 34,836,914

The monthly traffic generated across the leading ten dedicated deepfake pornography websites reached 34,836,914 this year. This considerable number shows the undeniable demand and the immense engagement these platforms command.

People flock to these websites in significant numbers, reflecting a widespread interest in the content they offer. The magnitude of this traffic makes it clear that societal curiosity and appetite for this controversial content category are on the rise.

Additionally, it indicates deepfake technology's growing influence on shaping online preferences for creators and consumers.

Number of deepfake porn

  • 2022 3,725
  • 2023 21,019
464%

Between 2022 & 2023, the amount of deepfake pornography created has increased by 464%

When comparing statistics from 2022 and 2023, we unraveled an alarming pattern. Unbelievably, the percentage increase in the production of deepfake pornographic videos within just one year was a startling 464%.

In 2022, the internet had about 3,725 deepfake pornography videos. However, in 2023, this number rose to 21,019. This drastic increase indicates the extent of the unchecked expansion of deepfake pornography.

Deepfake pornography videos by host website type

  • Dedicated deepfake pornography 90%
  • Mainstream pornography website 10%

Deepfake pornographic websites have a 90% market share

When we examine deepfake pornography by the type of hosting websites, we notice an alarming breakdown. Dedicated deepfake pornography platforms claim a significant share, encompassing a substantial 90%.

In contrast, mainstream pornography websites hold a smaller but still notable 10% share. This distinction indicates that deepfake pornography has carved out a unique space in the broader realm of online adult content.

TARGETED INDIVIDUALS

Below is a demographic breakdown of deepfake videos from top 10 dedicated deepfake pornography websites and 85 deepfake channels on Youtube, Vimeo, and Dailymotion. Deepfake content production is not uniform but selective, with specific individuals becoming prime targets based on gender, nationality, and profession.

This deliberate focus on particular attributes reflects the nuanced motivations and objectives behind creating and disseminating deepfake material.


Gender

Gender
gender-deepfake-pornography
gender-regular-deepfake

Gender plays a significant role in selecting individuals for deepfake content manipulation. In this section, we embark on a thorough exploration of the demographic makeup of deepfake videos.

The intersection of technology, creativity, and audience preferences becomes distinct as we uncover the role that gender plays in selecting individuals for deepfake content manipulation.

99% of deepfake pornographic content comprises female subjects

A glaring trend emerges when examining the demographic landscape of deepfake content. Our findings reveal that 99% of deepfake pornography features women as the primary subjects, while only 1% of the content features men.


Nationality

DEEPFAKE PORNOGRAPHY

  • South Korea 53%
  • America 20%
  • Japan 10%
  • England 6%
  • China 3%
  • India 2%
  • Taiwan 2%
  • Israel 1%
  • Other 4%
map-nationality

Nationality is another factor that influences the targeting of individuals by deepfake content. In some instances, a person's nationality may make them a more attractive target for malicious actors. This can be motivated by geopolitical tensions, cultural stereotypes, or simply the desire to create content that resonates with a specific audience.

This section discusses the countries most targeted by the technology and the extensive effects that it causes.

South Korea is the country most targeted by deepfake pornography

While deepfakes have been employed for various purposes, including entertainment, politics, and disinformation, certain nations are more susceptible to specific forms of deepfake content, particularly those of an explicit nature.

South Korea tops the list of these nations with a 53% rate of susceptibility to deepfake adult content. Other countries on the list include America at 20%, Japan at 10%, and England at 6%. Countries like China, India, Taiwan, and Israel all fall between the 3% to 1% mark, while the rest of the world has a combined 4% rate of being targeted for explicit deepfake content. Below is a table summarizing this information.

Country Susceptibility Rate to Deepfake Pornography
South Korea 53%
America 20%
Japan 10%
England 6%
China 3%
India 2%
Taiwan 2%
Israel 1%
Other 4%

Profession

Profession can also be a decisive factor in selecting targets for deepfake content. Public figures such as celebrities, politicians, and influencers often face a higher risk of being targeted due to their visibility and the potential impact of false or scandalous content on their careers and public standing.

The section below highlights the professions that are more likely to attract unconsented deepfake content generation.

94% of those featured in deepfake pornography videos work in the entertainment industry

The deepfake landscape reveals a striking dominance of individuals from the entertainment industry in adult-content videos. Our findings indicate that 94% of those featured in deepfake pornography videos are affiliated with the entertainment sector.

The list of highly featured professions in deepfake explicit creations includes singers, actresses, social media influencers, models, and athletes.

The table below summarizes this information statistically.

profession profession
Profession Susceptibility Rate to Deepfake Pornography
Singer 58%
Actress 33%
Social Media Influencer 3%
Model 2%
Athlete 2%
Other 2%

The prevalence of entertainment industry figures in deepfake adult content can be attributed to several factors. First, the high visibility of individuals in this sector makes them more likely targets, as they are readily recognizable to the general public.

Additionally, the availability of extensive visual and auditory material featuring entertainment professionals, such as movies, TV shows, and music videos, provides ample source material for deepfake creators.

Top 10 individuals most frequently targeted by deepfake pornography

Deepfake pornography is characterized by a select group of individuals who find themselves targeted for manipulative and often malicious purposes. These individuals, often public figures or celebrities, bear the brunt of deepfake creators' efforts.

This section delves into the profiles of the top ten most frequently targeted individuals in deepfake pornography. We shed light on the frequency of content that features.

However, due to the sensitive nature of the data, we have opted not to disclose the names of these individuals. Instead, we have provided information about their nationality and profession in the following table.

Nationality & Profession Video Features Views
South Korean Singer 1,595 5,611,500
South Korean Singer 1,238 3,865,000
South Korean Singer 923 2,305,500
South Korean Singer 844 3,395,500
South Korean Singer 803 1,365,200
South Korean Singer 761 3,294,000
South Korean Singer 751 1,995,800
Thai Singer 733 3,852,000
South Korean Singer 733 3,084,000
British Actress 713 3,190,100

Due to the sensitive nature of the data, we have opted not to disclose the names of these individuals. Instead, we have provided information about their nationality and profession.

The frequency with which these individuals are targeted underscores the need for comprehensive strategies to address deepfake threats. Such strategies should encompass legal measures, content detection technologies, and public awareness campaigns aimed at protecting these individuals' rights, privacy, and reputations.

Advancements in Deepfake Technology

In recent years, advancements in deepfake technology have ushered in a new era of creative possibilities and concerns.

Two significant factors have played pivotal roles in shaping the landscape of deepfakes: the rise of Generative Adversarial Networks (GANs) and the increasing availability of user-friendly tools, software, and communities for creating deepfakes.

This section will expound on both.

The rise of GANs

At the heart of the deepfake revolution lies the emergence and refinement of Generative Adversarial Networks (GANs).

Introduced in 2014, these neural network architectures have since become the driving force behind realistic data generation, including images, audio, and text used in deepfake technology. As a result, they have led to a wealth of academic research and publications. Scholars and researchers worldwide have recognized the potential and versatility of GANs, resulting in a substantial body of work exploring various aspects of these neural network models.

The table below summarizes the extent of academic research conducted in this field from 2014 to 2021.

chart-the-rise-of-gans

Number of academic report about GANs published per year

In 2014, GANs made their academic debut with just one research publication. This seminal paper by Ian Goodfellow and his collaborators introduced the concept of GANs, outlining their novel architecture and adversarial training framework.

In the year 2021, the field of GAN research will reach a remarkable milestone. A staggering 658 research publications focused on GANs emerged, marking an extraordinary increase compared to the humble beginnings of just a single publication in 2014.

The exponential growth in GAN-related research is a testament to the versatility and potential of this approach.

Availability of user-friendly tools, software, & communities for creating deepfakes

Another significant factor driving advancements in deepfake technology is the increasing accessibility of user-friendly tools, software, and online communities. In the early days of deepfakes, creating convincing manipulations required technical expertise.

However, as the technology matured, a wave of user-friendly applications and platforms emerged, democratizing the creation of deepfakes. User-friendly deepfake tools and software, often accompanied by intuitive graphical interfaces, have lowered the entry barrier for individuals interested in experimenting with deepfake technology.

These platforms offer a range of features, from simple face swapping to more complex video manipulations. Furthermore, the growth of online communities dedicated to deepfake creation and discussion has fostered collaboration, knowledge sharing, and the development of open-source projects. In the following sections, our research will reveal how these factors have contributed to the advancement of deepfake technology.

most-creation-communities-and-forums

The realm of deepfake technology has witnessed the emergence of a robust online community and forum ecosystem. As the data from our research shows, there are approximately 15 deepfake creation community websites and forums, boasting a collective membership of over 609,464 individuals. These communities serve as hubs for enthusiasts, researchers, and creators interested in exploring the possibilities and challenges presented by deepfake technology.

Notably, a significant portion of these deepfake creation communities and forums are hosted on platforms primarily associated with deepfake adult content. This represents a complex intersection of creativity and caution.

While they provide a space for innovation and artistic expression, they also underscore the need for vigilance to mitigate the potential for misuse, misinformation, and privacy violations.

42 user-friendly deepfake tools with roughly 10 million monthly searches

The number of user-friendly tools making deepfake content generation more accessible than ever is on the rise.

Our research indicates that approximately 42 such tools exist, catering to a broad spectrum of users. These tools collectively garner a staggering 10 million monthly searches, underscoring the widespread interest in and demand for user-friendly deepfake solutions.

User-friendly deepfake tools come in various forms, with approximately 40% being downloadable applications and the remaining 60% offered as web-based platforms.

This diversity in accessibility caters to the preferences and needs of users, whether they seek the convenience of online tools or the flexibility of downloadable software.

42-user-friendly-deepfake-tools

One in every three deepfake tools allow users to create deepfake pornography

One of the most significant aspects of deepfake technology is the creation of adult content. At least one in every three deepfake tools allows users to generate adult content using AI-driven manipulation techniques.

Most of these platforms, accounting for 92.3%, offer users the ability to create deepfake adult content for free, albeit with certain limitations. This approach to accessibility has contributed to the widespread use of such tools.

However, there's also a financial endgame. These platforms consistently try to monetize their services by setting up premium features and encouraging users to buy them. The following list explores some of the monetization strategies that these platforms explore.

Selling Extra Credits:

Many deepfake platforms offer users the option to purchase extra credits. These credits can be used to access premium features, enhance generated content quality, or unlock additional tools and resources. This strategy allows users to tailor their experience to their specific needs.

Premium Subscriptions:

Another common monetization approach involves offering premium subscriptions. Users can subscribe to these premium tiers to access higher-quality deepfake tools, advanced features, and priority customer support. This subscription model provides users with enhanced capabilities and a more comprehensive deepfake pornography creation experience.

Online Advertisements:

Many of these platforms generate revenue through online advertisements. Advertisements may appear within the user interface or alongside content generated by the tool.

frame-laptop

We conducted this test using an Intel Core i5 processor with 16 GB of RAM and an NVIDIA GeForce 2070 graphics card. The processing time can even be further reduced with more powerful hardware.

It now takes less than 25 minutes and costs $0 to create a 60-second deepfake pornographic video of anyone using just one clear face image.

Half an hour is all that deepfake creators need to generate a one-minute pornographic video.

Findings from our research show that it now takes less than 25 minutes and costs as little as $0 to create a 60-second deepfake explicit video of anyone using just one clear face image.

Among the myriad deepfake applications available, one Windows-based application stands out for its low entry barriers. This application leverages the user's hardware power, making it accessible to a wide range of users.

Using this app, it is possible to create a 60-second deepfake explicit video of anyone in under 25 minutes. Not to mention, it comes with daily free credits provided to users, allowing them to develop deepfake porn videos almost for free.

Deepfake Porn Survey:
Awareness, Attitudes, and Impact

The results of our survey, which involved 1522 American males, provide valuable insights into the awareness, attitudes, and impact of deepfake pornography. Through this study, we aimed to gain a comprehensive understanding of the prevalence of deepfake content consumption and the underlying motivations driving its popularity.

Our findings shed light on the evolving dynamics within the adult entertainment domain, uncovering the factors that draw individuals to deepfake content and how it has begun to reshape conventional pornography.

Additionally, we explore the growing interest in the creation of deepfake content, highlighting its transformative effect on the industry. This research offers a glimpse into the complex and evolving landscape of deepfake technology within adult entertainment, providing a foundation for further examination and discourse. Here are our findings.

48% of surveyed US men have seen deepfake pornography at least once

In our survey of US men, a notable 48% of respondents reported having viewed deepfake pornography at least once.

This statistic underscores the widespread presence of deepfake content within the adult entertainment landscape and highlights its growing influence on consumer preferences.

Technological curiosity is the most significant reason behind watching deepfake porn

The top three reasons cited by participants for engaging with deepfake pornography reveal interesting insights into the factors driving its consumption.

  • A significant 57% of respondents expressed their interest in deepfake technology, driven by a curiosity to explore the capabilities of this emerging field.

    57%
  • Celebrity attraction ranked second at 48%, indicating that many are drawn to deepfake content featuring their favorite stars.

    48%
  • Fantasy fulfillment, identified by 36% of participants, reflects a desire for personalized and tailored experiences within adult entertainment.

    36%

74% of deepfake pornography users don't feel guilty about it

Interestingly, 74% of users who have engaged with deepfake pornography did not report feeling guilty about their consumption.

This finding suggests that deepfake content has become an accepted and normalized part of adult entertainment preferences for a significant portion of viewers. These users cited the factors below as the reasons why they do not feel guilty for their actions.

Top 4 reasons why they don't feel guilty about it are

top-4-reasons-why

One out of five daily porn users has swapped regular porn for deepfake porn

Remarkably, one in five daily pornography users revealed that deepfake content has replaced traditional pornography in their viewing preferences. This finding shows the impact of deepfake technology on the adult entertainment industry and its appeal to a substantial portion of the audience.

20% of survey participants have considered learning to create deepfake pornography

The survey also revealed that 20% of participants have contemplated learning how to create deepfake pornography.

Furthermore, one in ten respondents admitted to attempting the creation of deepfake pornography featuring either celebrities or individuals known to them. These statistics highlight the growing interest in deepfake content creation, with some individuals actively exploring this technology for various purposes.

73% of survey participants would want to report to the authorities if someone close to them became a victim of deepfake porn.

These reactions underline the seriousness with which many individuals view the potential harm caused by deepfake technology and the importance of safeguarding individuals from its misuse.

73%

of respondents expressed the desire to report such incidents to the authorities.

68%

indicated that they would feel shocked and outraged by the violation of someone's privacy and consent in the creation of deepfake pornographic content.

When we asked, "If someone close to you became a victim of deepfake porn, what is your reaction?"

Concluding Remarks

The prevalence of deepfake videos is on an upward trajectory, with a substantial portion featuring explicit content. Deepfake pornography has gained a global foothold and commands a considerable viewership on dedicated websites, most of which have women as the primary subjects.

Regrettably, this content continues to flourish, courtesy of communities, computer programs, and services. These factors contribute significantly to the accessibility and use of deepfake creation tools.

It is worth noting that there's a noticeable divide between the broader public's perception of deepfakes as unethical and wrong and the deepfake pornography user community.

These findings emphasize the multifaceted nature of deepfake technology and its impact on various domains. As a result, there is a pressing need for responsible usage, ethical considerations, and regulatory safeguards to address the evolving challenges that deepfake content presents.

APPENDIX

Data Sensitivity

METHODOLOGY

Conducting open-source research on sensitive data demands careful consideration, which we take very seriously. In the interest of ethical practice and responsible disclosure, we have made the decision to redact the names and URLs of specific sources, especially those associated with deepfake pornography, in order to avoid inadvertently endorsing or promoting tools, services, or websites with legal ambiguities or explicit unethical applications.

Forums and creation communities

METHODOLOGY

Our research included Google searches for terms such as 'deepfake forums' and 'deepfake communities,' as well as searches on websites associated with deepfake pornography and forum-based platforms. We relied on publicly accessible data from the respective websites and channels to determine the number of registered users across 15 identified deepfake forums and creation communities.

Deepfake Pornography's Online Presence

METHODOLOGY

Conducting open-source research on sensitive data demands careful consideration, which we take very seriously. In the interest of ethical practice and responsible disclosure, we have made the decision to redact the names and URLs of specific sources, especially those associated with deepfake pornography, in order to avoid inadvertently endorsing or promoting tools, services, or websites with legal ambiguities or explicit unethical applications.

Academic report about GANs published per year

METHODOLOGY

We conducted our search using arXiv's search engine, employing the keywords 'GAN' or 'generative adversarial network.' This allowed us to identify papers containing these keywords in their titles or abstracts for each year.

Tools and software for creating deepfakes

METHODOLOGY

We acquire data about deepfake applications and software by employing specific keywords associated with deepfake tools and services. Using this approach, we also collected details like release years, category, costs, the inclusion of mature content support, and availability of trial periods for the applications. To enhance our analysis, we've also utilized Ahrefs to gauge domain ratings and traffic metrics for these applications.

Deepfake Porn Survey: Awareness, Attitudes, and Impact

METHODOLOGY

To gather the information presented earlier, we conducted a survey with 1,522 male participants from the United States, all of whom had viewed pornography at least once in the past six months. We included attention-check questions to ensure thoughtful responses, excluding data that didn't meet our criteria from our analysis. It's important to recognize that relying on participants' recollections may introduce memory distortions and overstatements, potentially influencing the responses. It's essential to note that the survey outcomes do not reflect our personal standpoint.

The collection of quantitative data covered the period from July 15th to August 29th, 2023