Deepfake
Last updated June 14, 2026
What is Deepfake in simple terms?
In simple terms, a deepfake is a fake video, photo, or voice clip made by AI that looks and sounds like a real person — putting words in their mouth they never said. A convincing digital impersonation.
What is Deepfake?
A deepfake is a piece of synthetic media — typically a video, image, or audio clip — in which AI has been used to convincingly replace or fabricate a real person's face, voice, or actions, making them appear to say or do something they never actually said or did.
A deepfake is media that uses AI to make a real person appear to do or say something they never did. The classic case is a video where one person's face is swapped onto another's body so seamlessly that it looks genuine, but the same idea covers cloned voices and entirely fabricated clips. The name is a blend of "deep learning" — the branch of AI that powers it — and "fake." What sets a deepfake apart from old-fashioned video editing is that the AI learns what a specific person looks or sounds like from real footage, then generates new, believable frames or audio of that person doing things that never happened.
The technology behind it grew out of a method where two AI systems train against each other: one generates fake images while the other tries to spot the fakes, and they push each other until the forgeries are good enough to fool the detector — and often people too. As the tools have spread, making a basic deepfake no longer requires special skills or expensive equipment; consumer apps can do a passable job, while the most sophisticated results still take serious effort. The realism has climbed fast enough that "seeing is believing" is no longer a safe assumption for video or audio.
This is why deepfakes sit squarely in AI safety rather than being treated as a neat party trick. They're used to spread misinformation, impersonate public figures, defraud people with cloned voices of a "boss" or "family member," and create harmful fake imagery of real individuals without consent. There are legitimate, consenting uses too — film dubbing, satire, accessibility, recreating a performer with permission — but the headline concern is deception. The hard problem is that detection is a moving target: as detectors improve, so do the fakes, which is why the response also leans on labeling, provenance records, and plain public awareness that a convincing clip is not automatically a real one.
Real-world example of Deepfake
A finance worker at a company gets a video call from someone who looks and sounds exactly like the chief financial officer, asking them to urgently transfer funds for a confidential deal. Everything seems right — the face, the voice, the mannerisms — so the request feels legitimate and the pressure to act fast is real. But the "executive" on the call is a deepfake, generated from public footage and recordings of the actual person, driven in real time by a scammer. The worker, with no reason to doubt their own eyes and ears, follows the instructions. The reason this kind of fraud works is precisely the reason deepfakes are dangerous: our instinct to trust a familiar face and voice was built for a world where you couldn't manufacture them.
Related terms
Frequently asked questions about Deepfake
What is the difference between a deepfake and synthetic media?
Synthetic media is the broad umbrella: any content — image, audio, video, text — that AI generated or substantially altered, including harmless and creative uses. A deepfake is a specific, narrower kind of synthetic media: one that depicts a *real, identifiable person* doing or saying something they didn't, usually with the aim of being mistaken for genuine. So every deepfake is synthetic media, but most synthetic media isn't a deepfake — an AI-generated landscape painting or a fictional character's voice is synthetic media without impersonating anyone real. The word "deepfake" carries the implication of a convincing, often deceptive impersonation. **2. Mechanism — How does a deepfake work?**
How does a deepfake work?
An AI is trained on real images, video, or audio of a target person until it learns the detailed patterns of how they look, move, or sound. It can then generate new frames or audio of that person that never happened — swapping their face into a video, animating a still photo, or cloning their voice to read any script. Many systems were built on an approach where one AI generates the fake and a second AI judges whether it's real, the two improving each other until the result is convincing. The output is rendered frame by frame or sample by sample so the fabricated person matches a target performance or script. **3. Application — What is a deepfake used for?**
What is a deepfake used for?
The uses split into harmful and legitimate. The harmful and most discussed ones are misinformation, impersonating public figures, voice-cloning fraud, and creating non-consensual fake imagery of real people. Legitimate, consenting uses do exist: dubbing films into other languages with matching lip movements, recreating or de-aging actors with permission, satire and art, and accessibility tools. The dividing line is consent and deception — the same technique that lets a studio ethically revoice a film also lets a scammer fake an executive on a call, which is why deepfakes are treated primarily as a safety and trust problem.