What the terms mean
Synthetic media is a broad label for images, audio, video or text that was generated or heavily altered by software, usually using artificial intelligence. Deepfake is a narrower and more charged word. It usually refers to synthetic media that makes a real person appear to say or do something they did not, most often in video or audio.
Not all synthetic media is deceptive. Film studios have used digital effects for decades, voice tools can help people who have lost the ability to speak, and many artists use generated imagery openly. The concern is about material designed to mislead.
How fake video and audio are made
The details vary by tool, but the general idea is that software learns patterns from many examples and then produces new material that matches those patterns.
- Face swapping. A program is trained on many images of two faces. It then renders one person’s expressions and movements with the other person’s face.
- Lip-sync and reenactment. Software changes a person’s mouth movements, or their whole head motion, to match new audio.
- Voice cloning. A system learns the characteristic sound of a voice from recordings and then speaks new text in a similar voice. Depending on the tool, a fairly small sample may be enough for a rough imitation, although quality varies.
- Fully generated images and video. Some models create pictures or clips from a text description, with no real footage behind them.
Many of these tools rely on neural networks, which are systems loosely inspired by the brain that learn from large amounts of data. Newer tools are often described as generative AI. The technology has improved quickly, so advice that relied on spotting specific flaws can go out of date.
Why they can be convincing
People tend to trust what they see and hear, and for most of history a recording was strong evidence. Fakes exploit that habit. Several other factors help them spread:
- Short, low-resolution clips hide flaws, and compression on social platforms blurs small errors.
- Content that provokes anger or surprise is shared quickly, often before anyone checks it.
- A fake only needs to be believed long enough to cause damage, such as a rumor or a scam.
- Real recordings can also be falsely dismissed as fake, which muddies the water further.
Where harm can occur
Without going into detail, the main categories of misuse include:
- Fraud. A cloned voice or fake video call may be used to pressure someone into sending money or information, for example by pretending to be a relative or a manager.
- Harassment and non-consensual imagery. Fake explicit images of real people are a serious abuse. Many places have been adding or updating laws about this, and platforms often have reporting tools.
- Political and public deception. A fabricated statement by a public figure may be circulated shortly before an event.
- Reputation damage to private individuals or businesses.
Signs that may suggest a fake
No single sign is reliable, and good fakes may show none of them. Still, some things are worth noticing:
- Odd details in faces or hands, such as blurry edges around hair, unnatural blinking, or teeth that look uneven.
- Lighting or shadows that do not match the scene.
- Audio that sounds flat, has strange pacing or does not quite fit the mouth movements.
- Text in the background that looks garbled.
- A person saying something far out of character, or a story that seems designed to inflame.
Treat these as prompts to look harder, not as proof either way. Fakes can pass, and real footage can look strange because of poor lighting or compression.
Questioning media in a practical way
Habits based on context tend to hold up better than inspecting pixels.
- Pause before sharing. Strong emotion is a cue to slow down.
- Ask where it came from. Is there an original source, a date and a place? Clips with no origin deserve more doubt. Judging the reliability of a source applies to video as much as to articles.
- Look for other coverage. If something dramatic really happened, established outlets and other independent sources would typically report it.
- Search for the original. Searching for key frames or descriptions can show whether a clip is old, edited or taken from another event.
- Check the full context. A real clip with the wrong caption is also a common trick.
- Be careful with urgent requests. If a call or video message from someone you know asks for money or secrets, hang up and contact them through a number you already have. Some families agree on a code word for emergencies.
- Remember that detection tools are imperfect. Programs that claim to detect fakes can be wrong in both directions, so treat their results as one clue.
These are the same habits that help with online misinformation more generally, and fake-media tricks often arrive through phishing emails or similar messages.
Labels and the larger effort
Some platforms and tool makers ask creators to label AI-generated content, and some are experimenting with embedding information about how a file was made. These efforts are uneven and can be removed or ignored, so they are a helpful signal rather than a guarantee. Laws and rules about synthetic media are also changing and differ by country.
A balanced view
The existence of convincing fakes does not mean nothing can be trusted. It means trust should rest on provenance and corroboration, not on a single clip. Staying curious, checking before sharing and keeping some humility about what you can judge by eye go a long way.