Deepfake technology can put anyone's face into any video, but the targeting isn't random. Across every dataset that tracks it — hotline reports, platform takedown data, academic model studies, and law-enforcement statistics — the pattern is strikingly consistent by gender, profession, age, and geography. Here's what the numbers actually show about who ends up depicted.
1. The Scale and Gender Skew of Deepfake Pornography
The starting point for any discussion of targeting is how overwhelmingly the underlying content skews toward non-consensual sexual imagery of women.
Between 90% and 95% of all deepfakes circulating online are non-consensual pornographic depictions, and roughly 90% of the individuals shown in them are women.[1] A separate count of deepfake video files found 95,820 online in 2023 alone — a 550% increase since 2019 — with 98% of all deepfake videos being pornographic in nature and 99% of the people depicted in that pornography being women.[2]
| Share of all online deepfakes that are non-consensual pornography | 90-95% | UN Women |
| Share of NCII deepfake subjects who are women | ~90% | UN Women |
| Deepfake videos found online (2023) | 95,820 | Security Hero |
| Growth in deepfake video count since 2019 | 550% | Security Hero |
| Share of deepfake videos that are pornographic | 98% | Security Hero |
| Share of deepfake pornography targets who are women | 99% | Security Hero |
99%
of people depicted in deepfake pornography are women, according to an analysis of the leading deepfake-hosting sites.
Home Security Heroes, State of Deepfakes
“
The overwhelming majority of deepfakes online are sexually explicit, non-consensual, and made of women — this isn't a fringe use of the technology, it's the dominant one.
”— UN Women, AI-powered online abuse
2. Who Within the Entertainment Industry Gets Targeted
Within the pool of deepfake pornography targets, entertainment-industry figures make up the large majority — but not evenly across roles.
94% of deepfake pornography targets work in the entertainment industry, and within that group, singers account for 58%, actresses for 33%, social media influencers for 3%, models for 2%, and athletes for 2%.[2] By country of origin, targets are most often South Korean (53%), followed by American (20%), Japanese (10%), and English (6%).[2] Separately, 48% of surveyed US men report having viewed deepfake pornography at least once.[2]
Deepfake pornography targets by entertainment role
| Metric | Value | Source |
|---|---|---|
| Deepfake targets working in entertainment | 94% | Security Hero[2] |
| Top country of origin for targets (South Korea) | 53% | Security Hero[2] |
| US men who've viewed deepfake pornography at least once | 48% | Security Hero[2] |
58%
of deepfake pornography targets within the entertainment industry are singers — the single largest professional group affected.
Home Security Heroes, State of Deepfakes
“
Fame itself is the risk factor. The more publicly available reference footage someone has, the easier they are to target — which is exactly why entertainers dominate the victim pool.
”— Home Security Heroes, State of Deepfakes
3. Women in Public Office and Politics
Deepfake targeting extends well beyond entertainers into public office, where the gender skew is even more pronounced.
An analysis of 11 well-known deepfake websites found more than 35,000 mentions of non-consensual intimate imagery depicting 26 members of the 118th US Congress — 25 women and 1 man — meaning women members were roughly 70 times more likely to be victimized than their male colleagues.[3] The Markup independently confirmed the scope of that search.[4] The 19th News's reporting on the same findings underscored the same lopsided split: 25 of the 26 targeted members were women.[19] Reporting on deepfake pornography targeting elected officials shows this isn't limited to the US — cases involving women politicians have surfaced across multiple countries.[18]
| Metric | Value | Source |
|---|---|---|
| NCII mentions found depicting Congress members | 35,000+ | American Sunlight Project[3] |
| Congress members depicted | 26 (25 women, 1 man) | American Sunlight Project[3] |
| Relative likelihood women in Congress are targeted vs. men | 70x | American Sunlight Project[3] |
| Deepfake sites searched for the analysis | 11 | The Markup[4] |
| Share of targeted Congress members who were women | 25 of 26 | The 19th News[19] |
70x
more likely — women members of the 118th Congress were victimized by deepfake NCII compared to their male colleagues.
American Sunlight Project
“
Nearly every targeted member of Congress found in this analysis was a woman — a pattern that mirrors what researchers see in the wider deepfake ecosystem, not an anomaly specific to politics.
”— The 19th News
4. Minors and the AI-CSAM Surge
Age is another axis where the data shows sharply rising, disproportionate targeting.
NCMEC's CyberTipline received about 4,700 reports involving generative-AI child sexual exploitation content in 2023, rising to roughly 67,000 in 2024, and then to somewhere between 400,000 and 1.5 million in 2025 depending on whether one platform's bulk hash-hit submissions are included.[5] Across 2024 and 2025, NCMEC identified more than 275 direct victims of generative-AI CSAM, and its staff categorized more than 158,000 images and videos in CyberTipline reports as AI-generated between January 2023 and December 2025.[6] Among teens aged 13-17, roughly 1 in 17 report having personally been targeted by a deepfake nude, with LGBTQ+ teens (53%) and girls (41%) the demographic groups most likely to report the experience.[7] Separately, 1 in 10 minors say a peer has used AI to generate nude images of another minor.[8] WeProtect Global Alliance found NCMEC's AI-generated CSAM reports jumped from 6,835 in the first half of 2024 to 440,419 in the first half of 2025 — a 6,345% increase.[9]
| Metric | Value | Source |
|---|---|---|
| NCMEC generative-AI CSEC reports, 2023 | 4,700 | NCMEC[5] |
| NCMEC generative-AI CSEC reports, 2024 | 67,000 | NCMEC[5] |
| NCMEC generative-AI CSEC reports, 2025 | 400,000-1.5M+ | NCMEC[5] |
| Direct victims of generative-AI CSAM identified, 2024-2025 | 275+ | NCMEC[6] |
| AI-generated images/videos categorized by NCMEC, Jan 2023-Dec 2025 | 158,000+ | NCMEC[6] |
| Teens (13-17) personally targeted by a deepfake nude | 1 in 17 | Thorn[7] |
| Minors who say a peer used AI to generate nudes of another minor | 1 in 10 | Thorn[8] |
| Growth in AI-CSAM reports, H1 2024 to H1 2025 | 6,835 to 440,419 (6,345%) | WeProtect Global Alliance[9] |
1 in 17
teens (13-17) report having personally been targeted by a deepfake nude, with LGBTQ+ teens and girls most likely to report the experience.
Thorn
“
This isn't a hypothetical harm for teenagers — it's already happened to a meaningful share of them, and the demographic pattern tracks the same lines seen in adult targeting.
”— Thorn, Deepfake Nudes and Young People
5. Sextortion, Fraud, and Emerging Scam Patterns
Beyond pornography and CSAM, deepfake and AI-generated content increasingly shows up in financial fraud and extortion schemes.
Reports of financial sextortion to NCMEC's CyberTipline rose from 139 in 2021 to 10,731 in 2022 and 26,718 in 2023.[10] 72% of Americans say they've seen a fake or AI-generated celebrity endorsement online.[11] More broadly, 1 in 8 US adults report having been threatened with or subjected to non-consensual intimate imagery sharing, with women 1.7 times as likely as men to be targeted.[12] In identity-verification fraud specifically, deepfakes accounted for 24% of all fraud attempts against motion-based biometric checks in 2024.[15] Scammers have also begun building AI deepfake videos and voice clones that impersonate OnlyFans creators themselves, using the fabricated persona to defraud fans directly.[21]
| Metric | Value | Source |
|---|---|---|
| Financial sextortion reports to NCMEC, 2021 | 139 | WeProtect Global Alliance[10] |
| Financial sextortion reports to NCMEC, 2022 | 10,731 | WeProtect Global Alliance[10] |
| Financial sextortion reports to NCMEC, 2023 | 26,718 | WeProtect Global Alliance[10] |
| Americans who've seen a fake AI celebrity endorsement | 72% | McAfee[11] |
| US adults affected by NCII threats/sharing | 1 in 8 | Cyber Civil Rights Initiative[12] |
| Relative NCII risk for women vs. men | 1.7x | Cyber Civil Rights Initiative[12] |
| Deepfakes' share of fraud attempts against biometric checks (2024) | 24% | Entrust[15] |
24%
of fraud attempts against motion-based biometric identity checks in 2024 involved a deepfake.
Entrust, Identity Fraud Report
“
Fans aren't just at risk of seeing fabricated content of a creator — some are being defrauded by a fake version of that creator, voice and video included.
”— Malwarebytes
6. The Model Supply Chain Behind the Content
Targeting patterns also show up upstream, in the tools used to generate deepfakes in the first place.
Researchers at the Oxford Internet Institute found roughly 34,000 to 35,000 publicly downloadable deepfake image-generator models on a single platform, Civitai, with about 15 million cumulative downloads since late 2022 — and 96% of the most-downloaded models were built to target identifiable women.[13] A peer-reviewed companion analysis of the same dataset found the most popular models disproportionately targeted individuals from China, South Korea, Japan, the UK, and the US.[14]
| Metric | Value | Source |
|---|---|---|
| Publicly downloadable deepfake models found (Civitai) | ~34,000-35,000 | Oxford Internet Institute[13] |
| Cumulative downloads since late 2022 | ~15 million | Oxford Internet Institute[13] |
| Share of top models targeting identifiable women | 96% | Oxford Internet Institute[13] |
| Geographic skew of top models' targets | China, South Korea, Japan, UK, US | Oxford Internet Institute / ACM FAccT 2025[14] |
96%
of the most-downloaded, publicly available deepfake image-generator models are built to target identifiable women.
Oxford Internet Institute, University of Oxford
“
These aren't obscure tools. Millions of downloads, overwhelmingly aimed at real, identifiable women, show this is a mainstream part of the deepfake supply chain, not a fringe corner of it.
”— Oxford Internet Institute
7. South Korea's Deepfake Crisis and the Global Legal Gap
South Korea offers the most detailed national-level data on age and scale within a single country, and illustrates a legal gap that extends well beyond its borders.
Of 527 reported South Korean deepfake sex-crime victims between 2021 and 2023, 59.8% (315) were teenagers.[16] Reported cases in the country rose from 156 in 2021 to 297 through July 2024 alone.[16] One of the largest Telegram channels distributing deepfake images of women in the country reportedly had more than 220,000 subscribers.[17] The United Nations has separately framed this pattern of AI-driven abuse as a form of violence against women that still lacks adequate legal protection in most jurisdictions worldwide.[20]
| Metric | Value | Source |
|---|---|---|
| Share of South Korean deepfake sex-crime victims (2021-2023) who were teenagers | 59.8% (315 of 527) | UPI, citing Korea National Police Agency[16] |
| Reported deepfake sex-crime cases, South Korea, 2021 | 156 | UPI[16] |
| Reported deepfake sex-crime cases, South Korea, through July 2024 | 297 | UPI[16] |
| Subscribers to one major Telegram deepfake-sharing channel | 220,000+ | NPR[17] |
59.8%
of reported South Korean deepfake sex-crime victims between 2021 and 2023 were teenagers.
UPI, citing South Korea National Police Agency
“
AI-driven deepfake abuse is spreading faster than the legal frameworks meant to address it — leaving most victims worldwide without a clear path to protection.
”— UN News
Key Numbers at a Glance
| Metric | Value |
|---|---|
| Share of all deepfakes that are non-consensual pornography | 90-95% |
| Share of deepfake pornography targets who are women | 99% |
| Deepfake pornography targets who are singers | 58% |
| Congress members depicted in deepfake NCII (2024 analysis) | 26 (25 women, 1 man) |
| Women in Congress: relative likelihood of being targeted vs. men | 70x |
| Teens (13-17) personally targeted by a deepfake nude | 1 in 17 |
| Growth in NCMEC AI-CSAM reports, H1 2024 to H1 2025 | 6,345% |
| US adults affected by NCII threats/sharing | 1 in 8 |
| Deepfakes' share of biometric-verification fraud attempts (2024) | 24% |
| Share of top deepfake-generator models targeting identifiable women | 96% |
| South Korean deepfake sex-crime victims (2021-2023) who were teenagers | 59.8% |
Sources
Sources
- 1. UN Women, AI-powered online abuse: how AI is amplifying violence against women and what can stop it ↩
- 2. Home Security Heroes, State of Deepfakes ↩
- 3. American Sunlight Project, Deepfake Pornography Report ↩
- 4. The Markup, 1 in 6 Congresswomen Targeted by AI-Generated Sexually Explicit Deepfakes ↩
- 5. NCMEC, The Work Never Stops: First Look at NCMEC's 2025 Data ↩
- 6. NCMEC, Generative AI ↩
- 7. Thorn, Deepfake Nudes and Young People ↩
- 8. Thorn, 1 in 10 Minors Say Peers Have Used AI to Generate Nudes of Other Kids ↩
- 9. WeProtect Global Alliance, Global Threat Assessment 2025 ↩
- 10. WeProtect Global Alliance, Sextortion ↩
- 11. McAfee, The Stars Scammers Love Most ↩
- 12. Cyber Civil Rights Initiative, FAQs ↩
- 13. Oxford Internet Institute, Dramatic Rise in Publicly Downloadable Deepfake Image Generators ↩
- 14. Oxford Internet Institute / ACM FAccT 2025 (arXiv preprint) ↩
- 15. Entrust, Identity Fraud Report ↩
- 16. UPI, South Korea Deepfake Sex Crime Digital Pornography Scandal ↩
- 17. NPR, South Korea Deepfake ↩
- 18. France24 (AFP), A Form of Violence Across the Globe: Deepfake Porn Targets Women Politicians ↩
- 19. The 19th News, AI Sexually Explicit Deepfakes Target Women in Congress ↩
- 20. UN News, AI Deepfake Abuse Story ↩
- 21. Malwarebytes, Scammers Target OnlyFans Users With Deepfakes ↩
