What a deepfake is
A deepfake is a video, image or audio recording that has been created or altered by software so that a person appears to say or do something they did not. The word combines “deep learning”, a type of computer technique, with “fake”. In everyday use it covers several things: swapping one person’s face onto another’s body, making a person’s mouth match different words, copying a voice, or generating a lifelike person who never existed.
Not every edited clip is a deepfake. A video trimmed to remove context, slowed down, or paired with a false caption is misleading but uses ordinary editing. The difference matters because the checks differ. The article on recycled images covers the older tricks of misleading captions; this one focuses on synthetic video and audio.
How they are made, in general terms
Without going into methods, the basic idea is that software studies many examples of a person’s face or voice and learns to reproduce them. With enough material it can then produce new footage that looks or sounds similar. Tools have become easier to use, and quality ranges from obvious glitches to results that fool many viewers at first glance.
Because public figures have a lot of recorded material, they are easier to imitate. But private individuals can be targeted too, using pictures and clips from their own social media accounts, which is one reason to think about what you post and who can see it. Your digital footprint and your privacy settings influence how much material is available.
Legitimate uses
The technology is not wrong in itself. Film studios use it for visual effects and for restoring or de-aging actors with permission. Companies use it to translate videos so the speaker seems to talk in another language. It can help people who have lost their voice, and it is used in satire and art, when labeled as such. The key difference is whether the people involved agreed and whether viewers are told.
Why misleading uses matter
Deception about events and statements. A fabricated clip of a leader or official saying something alarming can spread before anyone checks it. Examples of this concern include timing around votes, covered in the article on election misinformation.
Fraud. Cloned voices and faces can support scams, such as a caller who sounds like a relative or a video call that appears to show a boss. See the separate article on voice cloning for that risk.
Harassment and abuse. Fabricated clips or images of real people can be used to humiliate, threaten or blackmail them. This includes fabricated intimate images, which are a serious form of abuse even though they are fake. Many places have laws on this, and the rules vary, so ask local police or a legal advice service about your situation. If a child is involved, contact your national child-protection hotline or local police.
The “liar’s dividend”. Once people know fakes exist, a real recording can be dismissed as fake. That can let wrongdoers deny genuine evidence, and it can make people trust nothing at all. Neither extreme is healthy.
Lasting damage. A debunking rarely travels as far as the original. Even a clip shown to be false may leave an impression, which is why prevention and caution matter.
Cues, with caveats
People sometimes point to unnatural blinking, mismatched lip movement, blurred edges around the face, odd lighting or a flat voice. Some older fakes show these, but newer ones may not, and ordinary low-quality video can look strange too. Treat such cues as a reason to investigate, not as proof.
Stronger evidence comes from context:
- Who published the clip first, and do they have a track record?
- Do established news outlets or the person’s own verified channels report it?
- Is the statement out of character, or tailored to inflame?
- Is there a complete, longer version of the footage?
- Do official statements confirm or deny it?
What to do if you see one
- Do not share it. Even sharing to criticize it adds to its reach.
- Check the origin and look for corroboration from independent reliable sources.
- Use reporting tools. Most large platforms have a way to flag manipulated or abusive content.
- Tell the person affected, if you know them, and offer support.
- Save evidence such as links and screenshots before the post disappears, in case it is needed later.
If you are the target
Being the subject of a fake can be upsetting. You are not at fault. Keep records, report the content to the platform, tell someone you trust, and consider contacting local police or a legal advice service. If it is affecting your wellbeing, a qualified counselor or your doctor can help. If you feel in danger, contact local emergency services.
A balanced view
Synthetic video is a real concern, but panic is unhelpful. Most misleading content online is still simple: real footage with false claims attached. Staying calm, checking the origin and being willing to wait protect you from both. As the tools improve, trust will rely less on how a clip looks and more on who stands behind it.