AI or human game for kids: free real-or-AI quizzes, plus a no-tech version

The best free AI or human games for kids are Which Face Is Real, where you pick the real photo from a pair, and the Real or Not quiz, where you judge 15 pictures one at a time; for writing, the best game is one you make at home by reading two short passages aloud and guessing which a computer wrote. All three are free, and the two websites need no account. We opened and played each site on 18 October 2026. This guide was written by Peabody (peabody.kids), a free reading and learning site that is free through the end of 2026. None of these sites is ours, and nobody paid to be listed.

One thing to say first: these games are fun, but the point isn't to become an expert spotter. Even grown-ups do only a little better than guessing. The lasting lesson is to ask where something came from.

Free real or AI games for kids: the quick table

Ages are our suggestions, except the two lessons, whose grades come from their makers.

Which Face Is Real: the spot-the-AI face game

What it is. Two faces side by side, and one line: "Click on the person who is real." Click one and the site tells you if you were right. It was made by Jevin West and Carl Bergstrom at the University of Washington. The site says the fake faces were made with StyleGAN software (via thispersondoesnotexist.com) and the real ones come from a public collection of photos called FFHQ.

What kids learn. Its Learn page is the best part. It teaches tells like shiny "water splotch" blobs where hair meets the background, strange or melted backgrounds, glasses or earrings that don't match on each side, hair that looks painted on, and odd teeth. It also says what makes a photo look real: matching glasses and earrings, believable people at the edges, and readable text in the background.

The honest catch. The site dates from 2019, and its own Learn page warns that the splotch tell belongs to that one program and "we can't expect future algorithms to have this problem." So it's a great game for learning to look closely, and a poor guide to today's pictures.

For parents. The faces are ordinary adults. The site's menu links to its makers' course and book, whose title includes a swear word, so you may want to keep younger children on the game itself.

Real or Not: Microsoft's quiz for real or AI pictures

What it is. Press Start Game and you see one picture at a time, for 15 rounds, and choose Real or Artificial. It was built by Microsoft's AI for Good Lab for a study. Their paper says the game draws on 350 real photos and about 700 AI-made ones, and reports about 287,000 guesses from more than 12,500 people.

What kids learn. The study's headline is the lesson: people were right only 62% of the time, a little better than a coin toss. They did best on portraits of people and worst on landscapes and city scenes. Children who score badly can hear that this is normal.

For parents. Our round showed everyday scenes (two people walking towards a domed building). We can't see all thousand-odd pictures in the set, so play a round yourself before handing it over.

The no-tech AI or human game you can play tonight

You need nothing but this page. Read each pair aloud, in the same voice, and let your child guess which one a computer wrote. Then talk about why.

Being straight with you about these examples: all eight passages, the "person" ones too, were written with an AI model for this post, shaping one in each pair to sound like a child and one to sound the way chatbots usually write. So they're practice, not a real test. The real version is the one you make yourselves (see below).

Pair 1: My dog

• A. Biscuit ate my sock again. Not even a clean one. Mum says he's a menace but she still lets him sleep on her bed.
• B. Dogs are wonderful companions who bring joy to families everywhere. With their loyal hearts and playful energy, they remind us every day of the true meaning of friendship.

Pair 2: A rainy day

• A. Rainy days can be a perfect opportunity to slow down and enjoy the simple things in life, like reading a good book, building a cosy blanket fort, or sharing a warm drink with family.
• B. It rained so hard the drain on our road made a lake. My brother put his welly in it to see how deep it was. Very deep. He had to walk home with one wet sock.

Pair 3: Pizza

• A. I like pizza but only cheese. Pepperoni is spicy and the little circles curl up and hold grease like tiny bowls. Gross.
• B. Pizza is a delicious and versatile food enjoyed by people of all ages. Whether you prefer classic margherita or creative toppings, there is a pizza for everyone!

Pair 4: The beach (the tricky one)

• A. We went to the beach. It was fun. We swam and ate sandwiches and then we went home. It was a good day.
• B. At the beach I found a crab in a crisp packet. I named him Gary and let him go by the rocks, but a seagull was watching, so I'm not sure Gary made it.

Answers. The "computer" passages are 1B, 2A and 3A. Pair 4 is a trap: neither has the chatbot sound. 4A is flat in the way a tired person writes, and 4B is full of the odd, specific detail people include. An AI could have been asked to write either one. That's the point of the pair: the tells aren't reliable.

How to make your own

1. Pick a topic your child knows well: their pet, their school, last weekend.
2. You write three or four sentences about it in your own voice. Or use something your child wrote.
3. On your own device, ask a chatbot for a short paragraph on the same topic. Keep your child's name and details out of it.
4. Read both aloud. Kids guess, then explain why.
5. Next round, ask the chatbot to "write like a nine-year-old" and see if that fools them. It often does.

A twist for readers: Peabody's own stories in the library are written by AI at a child's reading level (the classics there aren't), and Peabody says so. Read one together and ask, "Could you tell?"

How to spot AI writing and pictures: the tells, and why they fail

Writing that often sounds like a chatbot:

• Smooth, general and upbeat, with no odd details.
• Lists of three ("reading a good book, building a fort, or sharing a drink").
• A tidy lesson at the end ("the true meaning of friendship").
• No mistakes, no jokes that only make sense to one family.

Pictures that often give AI away (mostly from Which Face Is Real's Learn page):

• Smudgy blobs or glow where hair meets the background.
• Backgrounds that melt, bend or make no sense.
• Earrings, glasses or collars that don't match on each side.
• Text in the picture that's gibberish.

Why none of this is reliable. AI can be told to write with spelling mistakes and silly details, and people write bland paragraphs all the time. New image programs fix old tells, which is exactly what Which Face Is Real's own page warns about. And in Microsoft's study, people were right about 62% of the time.

What about AI detectors? Don't trust them with a child's homework. Stanford researchers found that detectors flagged more than half (61%) of essays written by non-native English speakers as AI-made, and the study's senior author called the detectors "unreliable and easily gamed." That study is from 2023, and detectors have changed since, but nobody has shown that any free detector can be trusted to judge one child's work.

The skill that lasts: ask where it came from. Who posted it? Can you find the same thing from a source you trust? That works whether a picture is AI, edited, or real but misleading.

What age is right for spot-the-AI games?

• Under 7: skip the websites. Talk about real and pretend with picture books ("Is this a photo or a drawing? How do you know?").
• 7 to 9: the home game, and Which Face Is Real with you beside them.
• 10 to 12: Real or Not, Which Face Is Real's Learn page, and making their own pairs with you.
• 11 and up (grade 6+): the News Literacy Project lesson and AI Unlocked, often through school.

For how AI learns in words a younger child can follow, read AI explained for 7-year-olds together. If your child is starting to use chatbots, see is ChatGPT safe for kids?. For gentler AI play, try the Google AI games, where a computer guesses your doodles.

Where Peabody fits

Peabody doesn't have a spot-the-AI game. Its own stories are written by AI and pass safety screens before a child sees them; its games use questions made and marked by code, not AI. Using a Peabody story as the "AI" half of the home game is a fair way to show a child that AI writing can be perfectly readable, and that this is exactly why the label matters.

In a homeschool

In a homeschool this makes a good media-literacy thread for a month: one 15-minute round a week, the no-tech game first, then Which Face Is Real or Real or Not for older children. Keep a list on the fridge of the clues your child used and whether they held up. Peabody can supply the "AI" half of the home game: once a week, make a short story together on the free AI story generator (about 10 minutes), read it beside a paragraph you wrote, and ask which was which. Then read something from the library to finish. The honest limit: Peabody has no spot-the-AI game and doesn't teach media literacy itself. See AI in homeschooling and the homeschool hub.

Learn more elsewhere

• Which Face Is Real: Learn: the tells for fake faces, with the makers' own warning that they date.
• Stanford HAI: AI detectors biased against non-native English writers: why detectors shouldn't decide anything about a child's work.
• How good are humans at detecting AI-generated images?: the Microsoft study behind the Real or Not quiz.
• News Literacy Project: teaching about AI: the "AI or not?" lesson, quizzes and the Two Truths and AI videos.

For about seventy more free AI-learning sites, sorted by age, see where to learn about AI.

Questions parents ask

What is a good AI or human game for kids?

For pictures, Which Face Is Real (pick the real face from two) and the Real or Not quiz (15 pictures, real or AI). Both are free and need no account. For writing, make your own: read your child one paragraph you wrote and one a chatbot wrote, and let them guess.

Is Which Face Is Real suitable for children?

The game shows ordinary adult faces and is fine for most children of about 8 and up with a grown-up. Its fake faces come from 2019 software, so the tells it teaches are dated. The site's menu links to its makers' course and book, whose title includes a swear word.

Can AI detectors tell if my child used AI?

Not reliably. Stanford researchers found detectors flagged 61% of essays by non-native English writers as AI-written, and called them unreliable and easily gamed. Don't accuse or clear a child on a detector's score. Talk to them about how they wrote it instead.

How good are people at spotting AI pictures?

Not very. In Microsoft's study of more than 12,500 people playing the Real or Not quiz, players were right about 62% of the time. They did best on portraits and worst on landscapes and city scenes.

What should kids do instead of trying to spot AI?

Ask where something came from. Who made or posted it, and can you find it from a source you trust? That habit works for AI pictures, edited photos and real photos used in a misleading way.

At what age should kids learn about AI-made content?

Gently, from about 7, with real-versus-pretend talk and the no-tech guessing game. From about 10, online quizzes with a grown-up. From grade 6, structured lessons like the News Literacy Project's "AI or not?" and MediaWise's AI Unlocked.