Google announced Polar Peaks on October 26, 2026, adding a neuroscience-themed activity to AI Quests for teachers introducing children to how AI works. Google lists the programme as free for ages 11–14. The new quest has a classroom guide now; its dedicated teacher-training module is planned for November.
For teachers, the useful choice is which activity fits the lesson—not simply which is newest. The three quests offer different ways to discuss data quality, the examples AI learns from and the decisions people still need to make.
What AI Quests is—and is not
AI Quests is a game-based classroom resource co-created by Google Research and the Stanford Accelerator for Learning. Google’s FAQ says pupils do not need to code and that the game contains no generative-AI experiences. It teaches the process of developing AI, rather than giving children a live chatbot to practise prompting.
That makes it worth considering when the teaching goal is understanding how people shape an AI system. If the goal is checking chatbot answers, recognising misinformation or protecting personal information when using AI, add more targeted instruction. Experience AI’s safety lessons for ages 11–14 cover those issues separately.
Choose the lesson by its learning goal
The following choices are based on the published lesson and training descriptions, not a ranking of demonstrated learning results. The teacher-resources page lists written guides for all three quests; a guide and a teacher-training course are separate resources.
Quest | Useful teaching focus | Preparation available on October 26 |
|---|---|---|
Market Marshes | Flood forecasting: how data quality affects predictions and why people remain responsible for decisions. | Written guide and teacher training available. |
Dusky Dunes | A healthcare simulation. The training description emphasises privacy, varied training examples and human judgment. | Written guide and teacher training available. |
Polar Peaks | A fictional neuroscience investigation: review data, test a model and investigate its errors. | Written guide available. Google plans the training module for November; no exact day is stated. |
If you want a training course before teaching, the first two quests have that support now. If you are comfortable preparing from a written guide, Polar Peaks already has materials to work from. The forthcoming course does not mean the lesson itself must wait until November.
What the new Polar Peaks activity involves
Polar Peaks makes human oversight part of the activity: pupils inspect mistakes, correct them and test the model again. Its teacher guide deliberately starts with a model that performs inadequately, giving the class a reason to question results rather than accept them because training has finished.
The story uses a fictional species and a memory-loss investigation. The guide explicitly separates that narrative from real neuroscience: pupils are exploring a simulation, not mapping an entire human brain or making a medical diagnosis.
For planning, the guide estimates 50 minutes: 10 for introduction, 30 for play and 10 for discussion. That is a suggested lesson schedule, not a measured completion time for every class. It also describes shared-device groups and a projected whole-class option.
Check time, equipment and preparation
For the online game, plan for internet access. The published device recommendations include 4GB of memory and Chrome 139 or later, alongside a processor specification. Check the full requirements against school equipment and try the activity on the intended devices before scheduling.
For fewer devices or a shorter session, consider Flexible Quest: Market Marshes. This separate slides-led activity runs from one computer, is listed for ages 9–15 and has an estimated duration of 30–45 minutes. Its low-connectivity use case should not be mistaken for an offline version of the full game.
Review the AI Quests privacy notice through your school’s usual process. Google says the game requires no registration and collects no personal data, while collecting anonymous, aggregated gameplay metrics. That policy does not establish the account requirements or data practices of separate teacher-training platforms.
Teachers looking for training in their own use of AI can also read RohitAI’s coverage of the AI for Educators course. That is a different learning task from choosing a pupil-facing lesson about how AI systems are built.
Leave time for pupils to explain their choices
The Raspberry Pi Foundation’s account of AI Quests teaching and design describes prompts that ask pupils to explain their decisions, alongside classroom anecdotes. Those accounts do not establish how much the activities improve learning or whether they outperform another curriculum.
A practical way to use the discussion time is to ask pupils to identify a decision they made about the data, explain a mistake the model made and say why another test was needed. Treat those explanations as a starting point for follow-up teaching, not game completion as proof of mastery.
Choose Market Marshes for a flood-and-data lesson, Dusky Dunes for discussion of representation and privacy, or Polar Peaks for examining errors and human oversight. Start with the matching guide and leave room for the discussion: the central teaching question is what people must decide before an AI result can be useful.