Is this picture or video real? Checking for AI and edits

Why it is harder than before

AI image generation tools can produce a realistic photo from a short text description. Other tools can swap faces, change what someone appears to say in a video or remove things from a scene. A deepfake is a fake video or audio recording of a real person.

A few years ago, fakes often had obvious mistakes. Today many of those mistakes have been fixed. Tips you read online last year may no longer work.

No single trick is reliable

You may have heard that AI pictures always get hands wrong. You may also have heard that people in fake videos never blink. These were sometimes true for older tools. Newer tools often avoid them.

Real photos can also look strange. Poor lighting, a cheap camera and the compression WhatsApp applies to images can all make a genuine picture look odd.

The same caution applies to "AI detector" websites and apps. They can be one useful signal, but they make mistakes in both directions. A real photo can be flagged as fake. A fake can pass as real.

So treat each check as a clue. Several clues together tell you more than any one of them.

Clues to look at

In pictures

• Hands, ears, teeth and jewellery that look melted, merged or uneven
• Writing on signs, shirts or banners that is blurred or not real words
• Backgrounds where lines bend, buildings warp or people blend together
• Skin that looks too smooth, almost like plastic
• Shadows and light that seem to come from different directions

In videos

• Lips that do not quite match the words
• Edges of the face that blur or flicker when the head turns
• A voice that sounds flat, with odd pauses or the wrong accent
• Very short clips with no wider context

Around the picture or video

• Who first posted it and when
• Whether the account is new or has a history of sensational posts
• Whether trusted news outlets show the same picture or video

Simple checks you can do

1. Search with the image. Many search apps let you search using a picture instead of words. This is often called reverse image search. It can show whether the picture appeared earlier, perhaps with a different story.
2. Look for the original. Find the longest, clearest version. Fakes often circulate as cropped, blurry copies.
3. Check the source. A video said to come from a news channel should appear on that channel's own verified accounts.
4. Look for labels. Some platforms and AI tools add visible labels or hidden marks to AI content. This is called watermarking. A missing label does not mean the content is real.
5. Ask someone else. A second pair of eyes often spots what you missed.

What to do with doubtful media

• Do not forward it while you are unsure. Doubt is a good reason to stop.
• Do not use it as evidence in an argument, a complaint or a news story until you have confirmed it.
• Report it on the platform if it seems designed to deceive, harass or damage someone's reputation.
• If it shows you or a family member in a fake or compromising way, keep screenshots and links. Then report the crime: call the National Cyber Crime Investigation Agency (NCCIA) cybercrime helpline on 1799.

If someone has made a fake image of you, the fault is theirs, not yours. Do not pay anyone who threatens to share it.

Real pictures can mislead too

A genuine photo with a false caption is still misleading. Ask two questions. Is the picture real? Is the story attached to it true?

For forwarded posts and rumours, see spotting AI fake news.