What is a deepfake? A plain-English definition for parents

What does deepfake mean?

Ofcom, the UK's communications regulator, defines deepfakes as audio-visual content that has been generated or manipulated using AI, and that misrepresents someone or something. That last part is what separates a deepfake from an ordinary filter. A cat with bunny ears is edited. A video of a head teacher announcing that school is cancelled tomorrow, which the head teacher never said, is a deepfake.

The word comes from "deep learning", the kind of AI used to make them, and "fake".

What kinds of deepfake will children meet?

Mostly three kinds, and only one of them is the scary kind.

• Funny or silly ones. A celebrity singing a nursery rhyme, a teacher's face on a dancing cartoon. Children make these with apps and share them for a laugh.
• Misleading ones. A clip of a politician, a fake news report, a "photo" of a shark swimming down a flooded street. These are made to be believed.
• Harmful ones. A fake voice message from "Mum" asking for money, or a fake nude image used to bully or blackmail someone. These are made to hurt.

Most children will see plenty of the first kind and some of the second. Ofcom's polling found almost half of children aged 8 to 15 believed they had come across a deepfake in the previous six months.

Can you still tell a deepfake by looking?

Less and less. A few years ago the advice was to count fingers, look for blurred earrings and check whether the person blinked. The tools have improved, and those tells are no longer reliable on their own.

What still works is thinking about where something came from. Who posted it first? Is any trusted news source reporting the same thing? Would this person really say that? Could that really happen? Those questions work on a fake made last year and on one made next year.

How do I explain a deepfake to my child?

Start with something they already understand. A deepfake is like a puppet that looks exactly like a real person. Someone else is pulling the strings, choosing what it says and does.

In our picture book The Shark on the Screen, five-year-old Ash sees a photo of a shark in a party hat in his local canal. He does not need to know how the picture was made. He needs one question, and the book gives him one: could that really happen? Then he goes and asks a grown-up. That is the whole skill, and it scales up with them.

For older children, go further. Show them a deepfake you have found together and talk about who made it and why. Was it to make people laugh, to sell something, or to make someone look bad?

What should we do if a deepfake involves our child?

If someone has made a fake image or video of your child, do not share it further, even to show people how bad it is. Keep a record of where it appeared, report it to the platform, and tell the school if other pupils are involved. If the image is sexual, treat it as a crime whether it is real or AI-made (in the UK, a fake sexual image of a child is illegal in the same way as a real one): report it to the police, and use NCMEC's Take It Down or, in the UK, Childline and the IWF's Report Remove to get it removed.

Try this at home or in class

Our free Is This Photo Real? game shows children real and AI-made pictures side by side. Play a round together, and before each answer ask them to say why, not just which. The reasons matter more than the score.