Artificial Intelligence Olympiad for Classes 1–12: Learn, Explore & Create
Have you ever wondered how a phone recognises speech, how a computer learns to identify an animal, or why a smart app sometimes makes a mistake? Welcome to the School Connect International Artificial Intelligence Olympiad (SCO IAIO)—a place for students in Classes 1 to 12 to discover the fascinating world of AI, one idea at a time.
Whether you’re making your first picture-pattern game in Class 1 or exploring machine learning in Class 12, you can find free learning materials, chapter-wise videos, hands-on activities and original practice questions here. Parents and teachers will also find simple ways to make AI learning creative, responsible and enjoyable.
We’ll also visit inspiring international AI Olympiad celebrations. The global International Olympiad in AI (IOAI) and the separate International AI Olympiad (IAIO) are independent competitions with their own teams and awards. Their student stories can inspire us as we enjoy our own SCO learning journey.

Who can join School Connect’s AI Olympiad?
Students from Class 1 to Class 12 can explore and participate in the SCO Artificial Intelligence Olympiad. The official AI Olympiad subject page and registration page describe class-wise AI Olympiad learning and assessment for the complete range, from the youngest learners to senior secondary students.
AI starts simply for younger children—with shapes, familiar technology, sorting games and safe digital habits. As students grow, they meet data, algorithms, machine learning, programming, model evaluation and thoughtful AI projects. Every class deserves a clear and enjoyable starting point.
Picture games, everyday AI examples, patterns and safe technology habits
Creative challenges, instructions, simple data and early AI projects
AI applications, data, training examples and responsible use
Machine learning, coding, model evaluation, fairness and innovation
Find your AI Olympiad date
School Connect’s WINTER 2027 exam cycle calendar currently displays sessions on 6, 14 and 15 November and 4, 12 and 20 December 2026. “Winter 2027” is the cycle name; those particular dates are in 2026.
A helpful note for families: the main SCO AI subject and registration pages include Classes 1–12, while the present calendar card shows a narrower Classes 6–12 label for those published sessions. For a child in Classes 1–5, please confirm the appropriate class-wise date or school registration arrangement through the SCO registration page before selecting a session. All children can begin with the free learning materials today.
Wondering where to begin? Pick your class below, try a short question and enjoy discovering something new about AI before thinking about examination timing.
Explore the AI Olympiad syllabus for Classes 1–12
Every young learner has different strengths and interests. You might love pattern games, dream of designing robots or simply want to know how a translation app works. School Connect’s class-wise AI Olympiad guide introduces a progression from everyday AI ideas to data science, creativity and responsible problem-solving.
Choose a class to explore its learning guide. The ideas below are friendly starting points; your current official SCO class syllabus is the place to confirm the detailed chapters and question pattern.
| Class | What you can discover | Give it a try! |
|---|---|---|
| Class 1 | Meet AI through pictures, shapes, colours, matching games and safety | Try sorting five hand-drawn objects by shape |
| Class 2 | Spot everyday AI, simple instructions, choices and fun patterns | Explain the next two shapes in a pattern |
| Class 3 | Explore smart devices, clear rules, observations and human decisions | Draw a decision tree for sorting classroom objects |
| Class 4 | Discover labels, examples, basic machine learning and responsible technology | Group imaginary animals and explain your labels |
| Class 5 | Introduce training data, simple classification and creative AI projects | Train a paper classifier with several examples |
| Class 6 | Learn patterns in data, real AI uses, algorithms and safe sharing | Ask why a chatbot can sound confident but be wrong |
| Class 7 | Compare training and testing, AI predictions and privacy choices | Keep two cards aside as unseen test examples |
| Class 8 | Explore fair model testing, image/text AI and different kinds of errors | Compare results on diverse sample pictures |
| Class 9 | Try Python basics, machine learning metrics and responsible AI decisions | Calculate accuracy from ten fictional predictions |
| Class 10 | Study data preparation, NLP, evaluation and fairness | Make a simple confusion matrix and explain mistakes |
| Class 11 | Discover neural-network concepts, validation and real-world applications | Write a short AI project plan using safe example data |
| Class 12 | Design and evaluate AI projects with human oversight and ethics | Write a model card explaining aims, testing and limits |
Jump straight to your School Connect AI class
How do AI Olympiad questions change as children grow?
The official SCO AI exam overview groups the earlier classes around fundamental AI concepts, examples, pattern-recognition games and puzzles. Middle school adds understanding data and machine learning, while senior students move towards model reasoning, language AI, ethics and system design. The subject page currently describes a 60-minute class-wise assessment with 35 questions for Classes 1–4 and 50 questions for Classes 5–12; check the current class instructions when registering.
Learn to use AI wisely and kindly
Great AI learning isn’t only about computers. It also develops curiosity, fairness, creativity and good judgment. UNESCO’s AI Competency Framework for Students encourages human-centred thinking, ethical choices, understanding AI techniques and thoughtful design. In class, these ideas can become short stories, group discussions, drawings, puzzles and simple projects—even without expensive software.
Meet young AI innovators from around the world
The official International Olympiad in Artificial Intelligence (IOAI) 2026 took place in Astana, Kazakhstan, 2–8 August 2026 and brought 466 contestants from 108 countries and territories. The international organiser reports 37 gold, 73 silver and 110 bronze medals and 62 honourable mentions from its individual contest. These international achievements are wonderful examples of what dedicated students can accomplish as their AI knowledge grows.
Among the top three global individual contestants, Artem Gorokhov (Russia) placed first, Dauzhan Beketov (Kazakhstan) second, and Shao Zixi (China) third in the official 2026 final standings. The organisation also recognised participants and countries through its GAITE inclusion initiative—an opportunity for educators to discuss how international science competitions can widen access without pretending a separate SCO exam feeds that team pathway.
More exciting AI Olympiad journeys to discover
The International Artificial Intelligence Olympiad (IAIO), a distinct international programme, held its 2026 event in Ljubljana, Slovenia and its published results list 95 contestants, with top finishers from the United Kingdom, Canada and China. It announces its 2027 edition for Vietnam, 25 February–2 March 2027. Separately the national-team IOAI announces Singapore, 4–10 July 2027. Students following these events can discover different styles of AI challenges around the world. Each international competition has its own selection pathway, just as School Connect has its own class-wise Olympiad programme.
Watch global AI celebrations and discover exciting AI lessons
Curious about what an international AI Olympiad looks like? Explore a celebration in Astana, see photographs and student stories from the IOAI 2026 gallery, discover the separate IAIO 2026 awards and enjoy Code.org’s AI for Oceans learning videos.
Choose a card, watch, discuss and try its mini-challenge. Swipe sideways on a phone or use the Previous and Next links. The ceremony clip was shared by an attendee; for the final international results, the competition organiser’s pages are the best place to explore the medal records.
Explore the original video channels and organisers’ galleries for more stories, student achievements and available captions. The Code.org lesson is especially fun to discuss in class or at home.
Ready for an AI challenge? Try 20 questions with explanations!
Begin with patterns and familiar technology, then try questions about data, machine learning and using AI responsibly. Make your choice before reading the explanation—discovering why an answer works is the best part! These fresh practice questions are created for learning, so teachers and parents can choose the ones best suited to each class.
Patterns, AI and data
Question 1. A teacher groups picture cards into ‘plant’ and ‘animal’. What is this?
A. Classification B. Encryption C. Compression D. Computer repair
Answer: A. Classification. Assigning items to defined groups based on traits is classification.
Question 2. What is a labelled example in supervised learning?
A. An example with no information B. An input with a known correct category or target C. An encrypted password D. Only a computer’s prediction
Answer: B. An input with a known correct category or target. Supervised learning uses examples with known labels or target outcomes.
Question 3. Which action improves a biased training set?
A. Only repeating identical pictures B. Deleting the test set C. Adding relevant under-represented cases with valid consent D. Guessing the model is fair
Answer: C. Adding relevant under-represented cases with valid consent. Representative, correctly labelled training examples can reduce some dataset bias, although fairness still requires evaluation.
Question 4. Why do we use a separate test set?
A. To make the model remember answers B. To publish student private data C. To guarantee perfection D. To estimate performance on previously unseen examples
Answer: D. To estimate performance on previously unseen examples. Holding out suitable unseen examples gives a less biased estimate of generalisation.
Question 5. Which is a potential AI-enabled task?
A. Recognising speech patterns from audio with permission B. Charging a phone without any energy source C. Magically knowing the future D. Replacing every human judgement
Answer: A. Recognising speech patterns from audio with permission. Speech recognition can be an AI task but needs user consent, safe data handling and error checks.
Machine learning models
Question 6. A model predicts 8 of 10 test labels correctly. What is its test accuracy?
A. 8% B. 80% C. 20% D. 100%
Answer: B. 80%. Accuracy = 8/10 = 0.8 or 80% on that specific test set.
Question 7. For a yes/no classifier, what is a false positive?
A. A correct ‘no’ B. A missed true ‘yes’ C. A predicted ‘yes’ when truth is ‘no’ D. A correct ‘yes’
Answer: C. A predicted ‘yes’ when truth is ‘no’. False positive means an incorrect positive prediction.
Question 8. Which example illustrates data leakage?
A. Using held-out fresh data B. Removing duplicate training labels C. Using different data sources D. Including test answers in training features
Answer: D. Including test answers in training features. Information from the test labels entering training invalidates the evaluation.
Question 9. Which task best matches a regression model?
A. Predicting a numeric journey time B. Sorting images into cats and dogs C. Choosing a language category D. Identifying spam versus non-spam
Answer: A. Predicting a numeric journey time. Regression predicts a numeric continuous value; the others are classification tasks.
Question 10. A model performs well on the training data but poorly on a held-out test set. This may indicate:
A. Guaranteed fairness B. Overfitting C. Perfect generalisation D. Correct privacy practice
Answer: B. Overfitting. Overfitting can occur when a model captures training specifics that do not generalise.
Ethics and responsible AI
Question 11. Before uploading student photos to an AI app, what should a teacher consider?
A. Only the app’s colours B. Only file size C. Consent, privacy policy and necessary data minimisation D. Whether the AI sounds confident
Answer: C. Consent, privacy policy and necessary data minimisation. Personal images require proper consent, lawful handling and minimising data exposure.
Question 12. An AI chatbot states a made-up citation confidently. Best next step?
A. Trust it because it is fluent B. Share it without checking C. Assume it is official D. Find and inspect the original source
Answer: D. Find and inspect the original source. Generated citations can be hallucinated; source verification matters.
Question 13. Who should make final high-stakes decisions about a student’s assessment?
A. Accountable authorised humans using evidence B. An unreviewed chatbot C. A random algorithm D. The fastest computer only
Answer: A. Accountable authorised humans using evidence. AI may support but should not replace accountable human oversight in high-stakes education decisions.
Question 14. A translation app handles one accent badly. This reveals a possible issue with:
A. Phone wallpaper B. Language-data representativeness and fairness C. Battery brightness alone D. Webpage colour
Answer: B. Language-data representativeness and fairness. Uneven language/accent performance may arise from imbalanced data and model design.
Question 15. Why should students avoid entering passwords into AI prompts?
A. Prompts make passwords longer B. Passwords help train geography C. Sensitive secrets should not be disclosed to external tools D. AI needs passwords to answer maths
Answer: C. Sensitive secrets should not be disclosed to external tools. Private credentials should never be included in prompts or shared with untrusted tools.
Simple modelling and reasoning
Question 16. A dataset contains 12 examples: 9 are ‘fish’, 3 ‘trash’. If a model always predicts fish, what’s its training-set accuracy?
A. 25% B. 50% C. 100% D. 75%
Answer: D. 75%. It correctly labels 9/12 = 75%, showing why accuracy alone can hide class imbalance.
Question 17. Which improves a model comparison?
A. Using the same relevant held-out examples and clearly chosen metric B. Letting each model select favourable tests C. Hiding wrong predictions D. Comparing only model names
Answer: A. Using the same relevant held-out examples and clearly chosen metric. Fair evaluation uses comparable test data and appropriate metrics.
Question 18. If true positives = 6 and false positives = 2, precision is:
A. 60% B. 75% C. 50% D. 80%
Answer: B. 75%. Precision=TP/(TP+FP)=6/(6+2)=0.75.
Question 19. Which is an example of a human-written rule rather than a learned model?
A. Training a network on photos B. Using embeddings learned from text C. If temperature > 30°C, show ‘warm’ D. Learning a spam detector from labelled mail
Answer: C. If temperature > 30°C, show ‘warm’. A fixed if–then threshold is hand-coded logic; a model is trained on data.
Question 20. What belongs in a responsible student AI project report?
A. Only a perfect score B. Photos of other pupils without permission C. Claims that errors are impossible D. Purpose, permitted data, testing, failure cases and human oversight
Answer: D. Purpose, permitted data, testing, failure cases and human oversight. Transparent reports describe their aims, data, evaluation, limitations and safety.
Four creative AI projects you can enjoy at school or home
1. Paper picture classifier
Prepare fictional or student-created drawings of plants and animals with labels. Separate training examples from three unseen test cards. Discuss why drawings with unusual colours or shapes may confuse a model.
Output: an error table and one idea to improve data diversity. Privacy: no student photos are required.
2. AI for Oceans reflection
Use Code.org’s official AI for Oceans lesson, with teacher guidance. Check how balanced training examples affect whether fish or other sea creatures are labelled correctly.
Output: explain one incorrect classification and how training data influenced it.
3. Small fictional confusion matrix
Imagine ten consent-free made-up observations with TP=4, TN=3, FP=2, FN=1. Calculate accuracy 7/10=70% and precision 4/6≈66.7%. Discuss why another metric might matter more for a particular task.
Output: a labelled 2×2 table plus one model limitation.
4. Verify an AI-generated answer
Choose a publicly available simple factual answer—for example a science definition—and compare its wording with a class textbook or official education site. Record a correct claim, an uncertainty and a real source.
Output: evidence notes without sharing personal data or school records.
Eight habits of a thoughtful young AI creator
- Identify the problem and whether AI is actually needed.
- Use simulated or properly authorised data; never upload real pupil contact details, photographs or confidential school records.
- Identify relevant features and correct labels.
- Split examples into training and genuinely held-out evaluation data.
- Choose a suitable metric and report errors honestly; avoid testing only on favourable examples.
- Check fairness for relevant groups and accessible language requirements without collecting sensitive student characteristics unnecessarily.
- Explain human oversight, limitations, permissions and when a model must not be used.
- Credit software, videos and official research sources rather than claiming another organisation’s work as a School Connect original.
Try a seven-day AI discovery adventure
| Day | Study activity | Student demonstration |
|---|---|---|
| Monday | What is AI? Humans, fixed rules and machine learning | List examples and non-examples |
| Tuesday | Training data, labels and class differences | Create six harmless fictional labelled cards |
| Wednesday | Model testing, false positives and accuracy | Explain three predictions, one mistake |
| Thursday | Free SCO AI chapter or Code.org learning video | Answer the video card’s reflection question |
| Friday | Privacy, bias and source-checking | Describe one safety boundary |
| Saturday | Five questions from this original practice set | Correct the mistaken choice with a reason |
| Sunday | Learn from a real 2026 global IOAI ceremony or winner | Name the organiser/country, then plan next week’s project |
AI Olympiad questions students, parents and teachers often ask
Who can join School Connect’s AI Olympiad?
Students in Classes 1–12 can discover and participate in School Connect’s class-wise Artificial Intelligence Olympiad. The official AI subject guide and registration pages explain the full grade range.
Can Class 1 and Class 2 students begin learning AI?
Absolutely! Start with pictures, colours, matching, fun instructions and examples of smart tools. No coding experience is needed to enjoy these first steps.
How does AI learning change between Class 3 and Class 5?
Children start to explore patterns, labelled examples, simple rules and paper-based AI games. Drawing a pretend classifier is a great way to begin.
What will Classes 6–8 learn?
Students explore data, AI applications, how systems learn from examples, why a prediction may be wrong and how to use technology responsibly.
What do Classes 9–12 explore in AI?
Older students can try Python concepts, machine learning, evaluating predictions, language AI, fairness and responsible project design.
Where can I find the AI Olympiad class-wise syllabus?
Visit the SCO AI subject hub or choose your class from the twelve class links above.
When are the next School Connect AI Olympiad exam sessions?
The current Winter 2027-labelled calendar lists 6, 14, 15 November and 4, 12, 20 December 2026. Please check the published date and class on the registration page before selecting a slot.
Why does the date calendar say Classes 6–12 when AI is offered for Classes 1–12?
The AI subject guide and registration overview include Classes 1–12. The current date calendar shows a 6–12 label for those sessions, so families in Classes 1–5 should ask for the appropriate class-wise schedule or registration option.
Where can students register?
Use the School Connect registration page or work with your school to select the correct class and current examination arrangement.
Are School Connect AI learning materials free?
Yes. Free chapter-wise learning and linked videos help children explore AI at their own pace. Examination registration is a separate choice.
What is the difference between AI and machine learning?
AI is a broad area of making systems perform tasks that seem intelligent. Machine learning is one way to build those systems by learning patterns from data.
What is a training dataset?
It is a collection of examples a model uses while learning. In a simple classroom activity, labelled paper cards can represent training examples.
Why should we test an AI model on new examples?
A model may remember the examples it has already seen. Fresh test cases help students discover where a model works well and where it needs improvement.
What is an AI mistake or a false positive?
A false positive happens when a system predicts ‘yes’ but the correct answer is ‘no’. Understanding mistakes can be more useful than celebrating a high score alone.
How do children learn about fair AI?
Try comparing a small set of similar pictures with a more varied set. Talk about which examples are missing, which predictions fail and how people can improve the result.
Can younger children create AI projects without a computer?
Yes! Use paper drawings, labels, simple rules and imaginary examples. The projects in this guide work well as group activities.
What happened at the international IOAI in 2026?
The international IOAI brought students together in Astana, Kazakhstan, for AI challenges and celebrations. Browse the official global event records and our video cards above for inspiration.
Is the international IOAI the same as School Connect’s AI Olympiad?
They are separate opportunities. SCO supports its own class-wise AI Olympiad learning and assessments; international IOAI uses country-specific national-team selection.
Are these twenty questions from an SCO examination?
They are fresh practice examples written for this learning guide. Try them with a teacher or parent and use the explanations to strengthen understanding.
How should we start learning AI today?
Choose your class, read one free SCO AI chapter, play a pattern or classification game, and share one thing you discovered with a friend or teacher!
Teacher activity: become a classroom AI detective!
Imagine that you have designed a friendly pretend AI to recognise hand-drawn fish! How good is your little classifier? Let’s explore ten make-believe examples together, using paper drawings rather than personal photographs.
Sample confusion matrix with ten fictional test cases
| Actual category | Predicted fish | Predicted not fish |
|---|---|---|
| Fish (five illustrations) | 4 true positives | 1 false negative |
| Not fish (five illustrations) | 2 false positives | 3 true negatives |
Accuracy: (4+3)/10 = 70%. Precision: 4/(4+2) ≈ 66.7%. Recall: 4/(4+1) = 80%. Ask students why precision and recall describe different errors. A teacher can discuss whether the same ten examples are enough for reliable deployment—they are not.
Show off your AI idea with a mini project card
Project: Teacher-led paper picture classifier.
Purpose: Help learners understand training and test labels, not make real-world decisions.
Data: Ten imaginary paper images with fictional categories, no identifiable people.
Evaluation: Hold out previously unseen paper examples and report all correct and incorrect predictions.
Limits: Very small dataset; may confuse uncommon objects and does not establish a reliable deployed system.
Safety: No biometric, school-record or location data. A teacher approves examples and conclusions.
Help every student feel welcome while learning AI
Teachers can use UNESCO’s human-centred, ethics, application and design principles when assessing projects. Provide captions or text alternatives, age-appropriate activities, offline options and translated explanations where useful. Avoid entering confidential materials into public AI systems. The objective is confident critical thinking rather than uncritical dependence on any AI tool.
Your next AI adventure begins here
Pick your class from the AI learning table, open your free School Connect AI lessons, and try a simple activity today. Start by asking: “How could I teach a computer to recognise two different drawings—and what mistakes might it make?”
Love building things? Explore our Coding Olympiad learning guide. Curious about AI in other countries? Visit our UAE AI learning guide. You can also follow young innovators through the independent IOAI and IAIO global events.
Ready for the School Connect AI Olympiad? Visit the official registration page, choose your class from Classes 1–12, and confirm the suitable examination date. Keep learning, keep questioning, and have fun discovering what responsible AI can do!














Leave a Reply