How to Spot a Deepfake Photo or Video

You've probably seen a photo online that looked real but felt... off. Maybe it was a celebrity saying something they'd never say. Or a politician in a situation that seemed unlikely. These are often deepfakes, and they're getting more convincing by the month.

A deepfake is a photo, video, or audio clip that's been created or altered using AI to look and sound like a real person. The technology behind it is impressive in a technical sense, but it's also being used to spread misinformation, run scams, and cause confusion.

The good news? You can train yourself to spot them. Here's how.

What to Look For in Fake Photos

AI-generated images have improved hugely, but they still make mistakes. Some of the most common tells:

Hands and fingers. This has been AI's weak spot for a long time. Look for extra fingers, fingers that merge together, or hands that look twisted or blurry. It's getting better, but hands are still the first place to check.

Teeth. AI often struggles with teeth. They might look too uniform, too blurry, or there might be an odd number. Real teeth are a bit irregular. AI teeth sometimes look like a smooth white strip.

Ears and jewellery. Earrings that don't match, ears that are different shapes, or glasses frames that melt into the skin. Small details like these often give it away.

Background details. Look at what's behind the person. Are there words or signs that look garbled? Shelves with objects that don't quite make sense? Straight lines that bend or wobble? AI is great at faces but often botches the background.

Skin texture. AI-generated faces sometimes look too smooth, almost waxy. Real skin has pores, wrinkles, and imperfections. If someone's face looks unnaturally perfect, that's a red flag.

Hair edges. Where hair meets the background or forehead, AI can struggle. Look for hair that seems to blur into the background or individual strands that don't behave naturally.

What to Look For in Fake Videos

Videos are trickier because you're watching something move, and your brain tends to fill in gaps. Still, there are signs:

• Lip sync. Watch the mouth carefully. Does it match the words exactly? Even slight mismatches between lip movement and audio can indicate a deepfake.
• Blinking. Early deepfakes had a problem with blinking. The person wouldn't blink at all, or would blink oddly. Newer ones have fixed this, but it's still worth watching for.
• Edge flickering. Around the jawline, hairline, and edges of the face, you might see a slight shimmer or flicker. This is where the AI-generated face is being blended onto the original video.
• Lighting inconsistencies. If the lighting on the face doesn't match the lighting in the room, something's off. Shadows should behave consistently.

The "Zoom In" Test

If you're looking at a photo and something feels wrong, zoom in. AI-generated images often fall apart at close range. Details that looked fine from a distance become obviously strange when you look closely.

Pay particular attention to where different elements meet: skin and clothing, hair and background, teeth and lips. These transition areas are where AI most often slips up.

Why Deepfakes Matter

This isn't just a curiosity. Deepfakes are being used for:

• Scams. Fake videos of trusted figures (like Martin Lewis or Rishi Sunak) promoting dodgy investments have appeared on social media. They look real enough to fool people into parting with money.
• Misinformation. Fake videos of politicians saying things they never said, designed to influence opinions or stir up anger.
• Personal attacks. Fake intimate images of real people, which is now a criminal offence in the UK.

Being able to spot deepfakes isn't just a fun skill. It's a genuine form of self-defence against manipulation.

What to Do if You Spot One

Don't share it, even to say "look at this fake!" Sharing deepfakes, even to debunk them, helps them spread.

If you see a deepfake being used to promote a scam, report it to the platform (Facebook, YouTube, Twitter, etc.) using their reporting tools. You can also report scam content to Action Fraud on 0300 123 2040.

If someone sends you a video or photo that seems too good (or too shocking) to be true, check a fact-checking site like Full Fact (fullfact.org) or the BBC Verify team before believing it.

You Don't Need to Be an Expert

The single best defence against deepfakes isn't technical knowledge. It's scepticism. If something looks surprising, shocking, or too perfect, pause. Ask yourself: where did this come from? Who benefits from me believing it? Can I find this story from a trusted news source?

That moment of pause is worth more than any detection tool.

Want to know more about AI-powered scams? Read our guide on how to spot a fake voice call. And for a broader look at staying safe, see our AI safety guide.