A deepfake is a video, audio clip or image that has been generated or altered with AI and makes a real person appear to say something they never said or do something they never did. There is no single definitive sign; you need to weigh up inconsistencies in the face and the lighting, how well the voice matches the lips, where the content came from and the emotion it is trying to stir in you. The advice below is based on MIT Media Lab's Detect Fakes research project, the Government of Western Australia's cyber security guide and YouTube's help page; the information is current as of October 2026.
What is a deepfake, and where might you come across one?
Deepfake technology can place a person's face into another video, match their lip movements to a new voice or imitate that person's voice from a short audio recording. In everyday life you are most likely to come across it in these forms:
- An investment or product advert featuring a well-known face that promises "guaranteed returns"
- A phone call or voice message in the voice of someone close to you, urgently asking for money
- Altered footage of a real event, or a realistic video of a scene that never happened
- A fake photo or video created to humiliate someone
What signs should you look for in a video?
Researchers at MIT Media Lab note that sophisticated deepfakes are almost always done on the face, and list the points to watch for as follows:
- Skin: Are the cheeks and forehead too smooth or too wrinkled? Does the age of the skin match the hair and eyes?
- Eyes and eyebrows: Are the shadows where they should be?
- Glasses: Is there glare on the lenses, and does the angle of the glare change as the person moves?
- Beard, moustache and moles: Do they look natural, and do their edges sit properly on the face?
- Blinking: Does the person blink too little or too often?
- Lips: Do the lip movements match the speech naturally?
The Government of Western Australia's guide adds lighting to this: in real videos the light and shadows match the scene, whereas in deepfakes the light on the face often does not match the background or surrounding objects. Even so, a high-quality deepfake may show none of these signs; so looking at the image alone is not enough.
How can you spot a fake voice or call?
The Western Australia guide lists a voice sounding "a bit off" and mouth movements not matching the words being said among the important signs. Because there is no picture on the phone, the real warning sign is often the content of the conversation:
- A request that rushes you, asking for money or a code "right now"
- A conversation in which you are asked not to tell anyone
- A familiar voice calling from an unusual number
On a call like this, hang up and call the person yourself on a number you know. Agreeing on a code word in advance that only your family knows also makes it easier to tell such calls apart.
How do you check suspicious content step by step?
- Stop before sharing. According to the Western Australia guide, deepfakes are designed to stir strong emotions; if the content has really surprised or angered you, that is a warning.
- Look at the source: who shared the content first, how long has the account existed, and what has it shared before?
- Search for the same news or footage in other reliable sources. If an important event is real, it will appear in more than one reliable place.
- Check the image, or a frame taken from the video, with a reverse image search; an older or different version of the same image may turn up.
- Go through the face, lighting and voice signs above one by one.
- Look at the platform's AI label and the description field.
- If you are still not sure, don't share it; not spreading content that may be false is the safest choice.
What do platform labels mean?
Some platforms mark AI-generated content. YouTube, for example, requires creators to disclose content that makes a real person appear to say something they didn't say, alters footage of a real event or creates a realistic scene that didn't happen. An AI label is added to a disclosed video; for photorealistic content this label may also be shown on the video player itself. These labels are helpful, but the absence of a label does not prove that content is real; some accounts break the rules or never disclose.
What should you do when you come across a deepfake?
- Report the content using the complaint or report option on the platform where you found it.
- If you sent money or gave your card details after a fake call or message, call your bank immediately; report the suspected fraud to the security authorities.
- If your own image or voice was used without permission, keep evidence such as screenshots and links; for legal steps, consult a lawyer or the relevant authority.
- Use strong passwords and multi-factor authentication to protect your accounts. The Western Australia guide recommends this as a measure that makes it harder for your data to be used without permission to create deepfakes.
For other guides on using AI safely in everyday life, see the AI in Everyday Life section.
Frequently asked questions
What does deepfake mean?
A deepfake is a video, audio clip or image generated or altered with AI that makes a real person appear to say something they didn't say or do something they didn't do.
How can you spot a deepfake video?
Focus on the face: is the skin too smooth or out of keeping with the person's age, are the shadows in the right place, does the glare on glasses change with movement, are the blinking and lip movements natural? Also check whether the light on the face matches the background.
How can you spot a voice imitated with AI?
The voice may sound a little odd or flat, but the main warning is usually the content of the conversation: rushing you, asking for money or a code, asking you to keep it secret. Hang up and call the person yourself on a number you know.
Can deepfake detection apps be trusted?
No tool gives a definitive result on its own; high-quality deepfakes can get past some checks. Weigh up the tool's result together with a source check, a search of reliable sources and your own observation.
What does the AI label on YouTube mean?
YouTube requires creators to disclose content that looks realistic and has been generated or meaningfully altered with AI, and adds a label to these videos. The absence of a label does not prove the content is real.
What should I do if I see a deepfake?
Don't share it, and report the content on the platform where you found it. If you sent money after a fake call, call your bank immediately and report the suspected fraud to the security authorities; if your own image was used, keep the evidence and consult a lawyer.
Editorial desk
Stories signed by the Cultvia editorial desk follow our editorial policy; every story lists its sources at the end and errors are fixed under our corrections policy.
Last updated:
What changed?
- : First published
Your follow list is kept only on this device; when the page is updated you'll see it in Following.
Get email updates
Sources
- Detect DeepFakes: How to counteract misinformation created by AI — MIT Media Lab
- How to identify Deepfakes Cheat Sheet — Batı Avustralya Hükümeti (Government of Western Australia)
- Üretken yapay zeka içeriklerinin kullanımını açıklama — YouTube Yardım

