Guide
How our AI works — and how to use it safely
What the AI features actually do, what they get wrong, what happens to your data, and how to report a bad response.
Some features — essay marking, answer checking, brain-dump gap suggestions — send the text or photos you provide to an AI model (Google's Gemini, via our AI gateway) and show you what it generates. You always know when that's happening: AI features are things you actively invoke, they're labelled, and the core app (questions, flashcards, progress) works fine without them. This guide explains what the AI is good at, where it fails, and how to use it without handing over information you shouldn't.
What happens when you use an AI feature
- You submit something — an essay, a written answer, a photo of handwriting.
- It's sent to the AI model, which generates feedback, a suggested mark, or a list of gaps. Nothing you submit is used to train the model.
- The result is stored in your account so you can come back to it, alongside a usage record (which feature, roughly how much it cost) that manages your AI credits.
- You judge the output. An AI mark is a study signal, not an official grade — your teacher's marking and the official study design always win.
Using it safely
- Never put personal information in anything you send to AI — your full name, address, health details, or anything about other people. The AI needs your ideas about the topic, not your identity.
- Treat AI feedback as a second opinion. It's strong at spotting structure problems and missing points; it can be confidently wrong about facts and marking judgements.
- Check anything that matters against your official study design or your teacher — especially close to exams.
- AI checks inside other features (like flashcard answer checking) are off until you switch them on, and each AI use costs credits — the meter shows where you stand.
Watch out for
- Taking an AI mark as your real level — it's calibrated to the rubric but it isn't a QCAA or VCAA marker.
- Pasting a whole document that happens to include your name and school — crop or trim to just the work itself.
- Assuming a fluent answer is a correct answer — fluency is exactly what AI fakes best.
If the AI produces something wrong, unfair or inappropriate
Tell us. Email the support address (on the Help page) with a screenshot or a copy of the response and roughly when it happened — we review reports, and they're how filters and prompts get fixed. The same address works for questions about what data AI features hold about you, or to have your AI-generated content deleted.