How to Spot Deepfakes and AI Videos

How to slow down around AI videos, deepfakes, fake celebrity clips, altered video calls, and shocking viral footage.

What to know first

Deepfakes can be difficult to identify by sight alone. Look for technical clues, but rely more on source verification, context, and whether the video is trying to make you act quickly.

AI videos and deepfakes can imitate people, voices, facial movements, scenes, and events. Some are jokes or labeled creative work. Others are used for scams, harassment, misinformation, fake endorsements, or business fraud.

The more emotional or consequential the video is, the more you should slow down and find independent confirmation.

Practical red flags

  • A celebrity, executive, politician, friend, or family member makes an unusual request or endorsement.
  • The video appears only as a short clip without a full source, date, location, or original account.
  • Mouth movement, blinking, facial edges, skin texture, lighting, or background motion looks inconsistent.
  • The voice and face do not fully match the environment or emotional moment.
  • The clip asks for money, investment, donations, login codes, payment changes, or urgent sharing.
  • Comments or captions push secrecy, outrage, fear, or immediate action.

What to do next

  1. Search for the full video or original post.
  2. Check official accounts, press pages, or trusted news coverage for the same claim.
  3. For business requests, verify payment or access changes through a separate known channel.
  4. For family emergencies, call a known number or another trusted person.
  5. Do not share humiliating, explicit, or reputation-damaging content of a real person.
  6. Report harmful deepfakes to the platform and consider legal or school reporting paths when minors are involved.

Video clues can help, but they are not enough

Look for mismatched shadows, strange face edges, unnatural blinking, inconsistent background movement, sudden quality changes, or audio that does not match the room. These signs may suggest manipulation.

But video compression, poor lighting, filters, translation dubbing, and editing can also look strange. Treat these signs as reasons to verify.

Fake endorsements and fake authority

Deepfake clips are often used to make a trusted person appear to endorse a product, investment, charity, political claim, or urgent instruction. The risk is not only the technology. It is the borrowed trust.

If a famous person, local leader, pastor, teacher, doctor, banker, or company executive seems to ask for money or action in a surprising way, verify outside the video.

Do not spread harmful clips while investigating

If a video appears to show violence, explicit content, a minor, a private person, or a serious accusation, do not share it to ask whether it is real. Sharing can cause harm even if the content is fake.

Save evidence when appropriate, report through the platform, and contact a trusted adult, school, employer, attorney, or law enforcement when the situation involves threats, extortion, minors, or non-consensual intimate imagery.

Trusted resources

These references support the guidance above. They do not create certainty about whether specific content is AI-generated.

Verify the person through a separate path.

A convincing face or voice is not enough when money, access, safety, or reputation is involved.

Make a verification plan