A deepfake is a synthetic image, video, or audio clip created by artificial intelligence to appear authentic. The technology is increasingly used in romance scams, corporate fraud, and disinformation — and human eyes can no longer reliably detect it.
Deepfake (noun) — A synthetic image, video, or audio clip generated by artificial intelligence (AI) to convincingly depict a real person saying, doing, or appearing somewhere they did not. The term combines "deep learning" (the AI technique) with "fake." Deepfakes can create entirely fictional people who look photographically real, or place real people in fabricated scenarios.
Deepfakes exist in three primary forms, each with distinct creation methods and use cases:
AI-generated still photos of people who don't exist, or real people placed in fabricated scenarios. Created by tools including Midjourney, DALL-E, Stable Diffusion, Adobe Firefly, and Flux. Most commonly used in romance scam profile photos.
AI-generated or AI-manipulated video that shows a person doing or saying something they didn't. Created by tools including Runway, Sora (OpenAI), Kling, and Pika. Used in corporate fraud, political disinformation, and non-consensual intimate imagery.
AI-cloned voice audio that sounds like a real person. Modern voice cloning requires as little as 3 seconds of audio. 70% of people cannot distinguish an AI-cloned voice from a real one (McAfee 2024). Used in phone scams impersonating family members, executives, and public figures.
Deepfake images are generated using AI models trained on massive datasets of real photographs. When you prompt an AI image generator — "a professional headshot of a woman in her 30s" — the model synthesizes a photorealistic face that matches no real person. The result is a photo-quality image of someone who does not exist.
The most widely used AI image generators, in rough order of detection difficulty (hardest to easiest):
Open-source detection tools struggle to keep pace with advances in these generators. Commercial AI detection models — like Hive AI, which powers Faux Spy — achieve approximately 94% accuracy across all major generators by analyzing deeper pixel-level patterns that open-source tools miss.
Detection accuracy figures represent independent researcher benchmarks (2024–2026). Commercial model accuracy reflects Faux Spy's internal testing.
Check if a photo is a deepfake — right now, for free
🕵️ Add to Chrome — Free10 checks/day free · No account required · Works on dating apps, LinkedIn, Facebook, and any website
The most common use of deepfake images is in romance scams. Scammers use AI image generators to create realistic profile photos of attractive, fictional people — then build trust over weeks or months before requesting money. Because the face is entirely synthetic, reverse image search tools find no matches. The FBI confirmed increasing use of AI-generated images in romance scam operations in its 2024 IC3 report.
Romance scam losses reached $672 million in 2024 (FBI IC3) — and the Consumer Federation of America estimates the true total is closer to $4.7 billion annually, with roughly 87% of victims never reporting the crime.
Deepfake video calls impersonating executives have resulted in major corporate losses. In a landmark 2024 case, an employee at a Hong Kong firm was deceived by a deepfaked video conference call featuring an AI-generated CFO and multiple colleagues, resulting in a $25 million wire transfer. 90% of US companies have experienced deepfake fraud (Medius 2024).
AI voice cloning requires as little as 3 seconds of audio. Scammers use cloned voices to impersonate family members in distress ("grandparent scams"), executives authorizing transfers, or government officials demanding payment. 70% of people cannot distinguish an AI-cloned voice from a real one (McAfee 2024).
77% of US voters encountered deepfake political content during the 2024 presidential election. These include AI-generated videos of candidates saying things they didn't say, fake audio recordings, and AI-generated images misrepresenting real events.
Sources: FBI IC3 2024 Annual Report • Medius Group 2024 • McAfee Consumer Survey 2024 • 2024 Election Integrity Research
68% of deepfakes are now nearly indistinguishable from real images by human perception (2024 benchmark research). The tells people look for — unnatural eyes, weird hands, strange backgrounds — have largely been eliminated by modern AI generators. Relying on visual inspection is no longer reliable, even for trained experts.
AI detection tools analyze patterns in the underlying image data that the human eye cannot perceive: statistical regularities in pixel distributions, artifacts from the generative model's upsampling process, inconsistencies in noise patterns between the subject and background, and frequency-domain anomalies invisible in the visible-light spectrum. These signals are present even in high-quality deepfakes and remain detectable by well-trained AI models.
Free users get 10 checks per day. Pro users ($9.99/month or $99/year) get unlimited checks plus deepfake manipulation detection.
A deepfake is a synthetic image, video, or audio clip created by artificial intelligence to convincingly appear real. The term combines "deep learning" (the AI technique used) and "fake." Deepfakes can show real people saying things they never said, create entirely fictional people who look photographically real, or clone real voices from a few seconds of audio.
Still-image deepfakes are created with AI image generators: Midjourney, DALL-E (via ChatGPT), Stable Diffusion, Adobe Firefly, and Flux are the most widely used. Video deepfakes use tools like Runway, Sora, and Kling. Audio deepfakes use voice cloning models that replicate someone's voice from as little as 3 seconds of audio. All of these tools are now publicly available, most for free.
Creating a deepfake is not inherently illegal, but using one to commit fraud, impersonate someone without consent, create non-consensual intimate imagery, or interfere with elections is illegal. As of 2025, 47 US states have enacted or are advancing deepfake-specific legislation. The US DEFIANCE Act (2024) makes non-consensual deepfake intimate imagery illegal at the federal level.
No — not reliably. 68% of deepfakes are nearly indistinguishable from real images by human perception (2024 benchmark data). Modern AI generators have eliminated most of the visual tells people were trained to spot. AI detection tools like Faux Spy are significantly more accurate (~94%) because they analyze pixel-level patterns invisible to the human eye.
Romance scammers use AI image generators to create fake profile photos of people who do not exist. Because the face is entirely synthetic, reverse image search finds no matches. 1 in 4 Americans encountered an AI-generated photo on a dating app in 2026 (McAfee). 55% of malicious clones on Tinder now use AI-generated images. The FBI confirmed this trend in its 2024 IC3 annual report.
An AI-generated image is any image created from scratch by AI, including entirely fictional people and scenes. A deepfake more specifically refers to AI content that impersonates a real person — placing them in situations they weren't in, or creating content in their likeness. In common usage, the terms are often used interchangeably when referring to fake dating profile photos using AI-generated faces.
1 in 4 Americans encountered an AI-generated or AI-modified photo on a dating app in 2026 (McAfee). 55% of malicious dating app clones now use AI-generated profile images. The FBI's 2024 IC3 report confirmed that romance scammers are actively using AI generators to create undetectable fake profile photos, leading to $672 million in reported losses in 2024 alone.
Deepfakes are created by a wide range of actors: romance scammers using free AI tools to create fake dating profiles; organized crime groups running pig butchering and investment fraud; individual bad actors creating non-consensual intimate imagery; state-sponsored actors generating political disinformation; and legitimate creators using the same tools for entertainment, film effects, and accessibility features. The technology itself is neutral — the intent behind the use is what matters legally and ethically.
Faux Spy is a free Chrome and Firefox extension powered by Hive AI. Right-click any photo on any website and get an instant Real vs. AI-Generated verdict with a confidence score — no uploading required.
🕵️ Add to Chrome — Free 🦊 Add to Firefox10 checks/day free · No account required · ~94% accuracy