Deepfake Identity Fraud Poised to Increase Nearly 500 Percent in 2026

Deepfake-enabled fraud is having a moment. Across four main artificial intelligence (AI) attack types, 2026 is on track for a 495 percent increase in deepfake identity fraud over 2025, according to the Identity Fraud Index report from identity management company Shufti.

The report tracked global fraud attempt data and it projected a sixfold jump in deepfake fraud this year compared to 2025, with swift growth in document deepfakes, as well as strong use of synthetic identity fraud.

Deepfakes use AI to create synthetic faces, videos, and identity documents that can pass automated identity checks. The growth of generative AI tools makes this type of fraud more accessible—it no longer takes hours of work and in-depth expertise to craft a fake identity; it takes one image or a text prompt in an AI tool. This has enabled AI fraud to scale rapidly.

Shufti tracked four primary areas of deepfake fraud:

• Synthetic identity: Entirely synthetic faces of people who do not exist, which comprised 42.3 percent of 2025 deepfake fraud. This fraud type is expected to grow around 73 percent this year.
  
  
• Live video deepfakes: Manipulated live video stream during verification, which comprised 28.1 percent of 2025 deepfake fraud.
  
  
• Face swaps: A victim’s face mapped onto an attacker’s head in real time, which comprised 17.6 percent of 2025 deepfake fraud.
  
  
• Document deepfakes: AI-produced documents and media submitted as genuine. Comprised 11.9 percent of 2025 deepfake fraud. It is expected to grow 3,892 percent in 2026—increasing 40 times over 2025 levels.
  
  

The most effective deepfake identity fraud attacks combine different techniques, such as blending a presentation attack with a synthetic identity or using an injection attack to skip the use of a camera to pipe an AI-generated synthetic identity straight into a verification tool.

Deepfake videos and images have gotten so effective and seamless that human review is no longer reliable. In a 2025 study published in Scientific Reports, researchers found that study participants could correctly identify an AI-generated voice only around 60 percent of the time, and they identified an AI-generated voice to be the same as its real counterpart about 80 percent of the time. Video used to be more of a giveaway, when audio and mouth movements didn’t match up so seamlessly, but advancements in generative AI have closed that gap.

This difficulty makes layered fraud detection essential, the Shufti report said. Those layers can include verifying the integrity of the image capture (such as confirming the video originates from a genuine smartphone lens), liveness cues and active challenges that prerecorded deepfakes cannot improvise, and media forensics tools that check for unnatural pixel blends where a swapped face meets a real jawline.

For more insights on managing deepfake threats, revisit our series on Deepfakes and Fraud.