How to get reliable answers from an AI study assistant

AI study assistants are genuinely useful — but the same assistant can give a brilliant answer or a useless one depending on how you ask. The difference isn't luck. It's a handful of habits.

This guide covers those habits. They work on any AI assistant — our own Alex, ChatGPT, Claude, Gemini — because they all respond to the same things: clear context, a clear request, and a clear standard for the answer.

About Alex: Alex is built on OpenAI's GPT-4o family — the technology behind ChatGPT. The assistant is the same on free and paid plans; paid plans just allow more daily usage.

The 30-second version

• Give context — grade, the actual question, what went wrong.
• Show an example of the answer you want.
• Ask for the working, not just the answer — this is the single best defence against wrong answers.
• Tell it to flag uncertainty instead of guessing.
• One topic per chat, one clear question at a time.
• Verify anything high-stakes (dates, official rules) against the source.

1. Give it context

A cold question gets a generic answer. The same question with context gets a targeted one.

The more an assistant knows about who is asking and what they're stuck on, the better it answers. Grade, subject, the exact sticking point — all of it helps.

💡 With Alex this is automatic — it's personalized by design. You don't have to feed Alex the context, because it can pull your child's own data on demand: their recent exam scores, the questions they actually got wrong, and which skills they've mastered vs still need work. So you can just ask "what should I practise?" or "where do I keep going wrong?" and Alex answers from their real history. With a general assistant like ChatGPT, you'd have to type all of that in yourself.

2. Show an example of the answer you want

If you want a certain shape of answer, show one. Models match the example you give them.

• "Give me 3 practice questions in this format: question, then a hint, then the answer hidden at the bottom."
• "Explain it the way you'd explain to a 10-year-old — short sentences, one idea per line."

One example of the format you want is worth a paragraph of instructions.

3. Ask for the working, not just the answer

This is the most important habit. Asking an AI to show its steps is the single best way to catch a wrong answer — including yours.

• "Show your working step by step, and tell me which rule you used at each step."
• "Before the final answer, list what you assumed."

When the reasoning is on the page, a wrong step is obvious — to you and to the model, which often corrects itself mid-explanation. A bare answer hides its mistakes.

4. Tell it to flag what it isn't sure about

AI assistants can sound completely confident while being wrong — this is called hallucination. You can't switch it off, but you can ask the model to be honest about it.

• "If you're not sure, say so instead of guessing."
• "If this needs a fact you don't have, tell me what to check rather than making one up."

A good assistant will then mark the shaky parts — which tells you exactly what to double-check.

5. One clear question at a time

Stacked questions get half-answers. "Explain ratios and also give me a worksheet and tell me her weak spots" will do all three shallowly. Ask for one thing, get the full answer, then ask the next.

Be specific about what you want, too: "a 5-question worksheet on ratios for P5, with answers" beats "some ratio practice."

6. Keep each chat to one topic

AI assistants work from a limited recent window of the conversation — they don't remember everything you've ever said. So a long chat that wanders from fractions to spelling to exam dates slowly loses the thread and the answers drift.

• One topic per chat. New topic → start a new chat. This resets the focus and gives the cleanest answers.
• Some tools (like coding assistants) let you compact or clear the conversation — same idea: trim the clutter so the model focuses on what matters now.
• If answers start feeling off after a long back-and-forth, that's your cue to start fresh and restate the question.

💡 Alex handles this for you. In a long chat, Alex automatically compresses and trims the conversation — it keeps the most recent turns word-for-word and folds the earlier ones into a short summary, so it holds the thread instead of forgetting where you started. You still get the sharpest answers by keeping a chat to one topic, but you won't lose the early context. To fully reset for a brand-new subject, just start a new chat.

7. Verify the high-stakes facts

For anything that matters — exam dates, marking rules, official scoring, school cut-offs — treat the AI as a fast first draft, not the final word. Confirm it against an official source.

For PSLE specifically, the source of truth is MOE's PSLE & S1 Posting site. If an assistant tells you something about the AL bands or posting rules, it's worth a 30-second check — see our plain-English guide to the AL scoring system.

A worked example: same question, two ways

Mei wants help with her P5 son Ryan's ratio mistakes.

• ❌ First try: "Help with ratios." → a generic textbook explainer Ryan has already seen.
• ✅ Second try: "Ryan is P5. On ratio word problems he sets up the ratio correctly but always forgets to convert to the same units. Show me how to teach just that step, with one worked example, and end with 3 short practice questions (answers at the bottom). If anything's unclear, ask me."

The second prompt has context (P5, the exact slip), an example format (worked example + 3 questions), a request for working, and permission to ask back. Same assistant — a far more useful answer.

The short version

You don't need to be technical to get great answers from an AI assistant. You need to:

1. Tell it who's asking and what's stuck.
2. Show it the answer you want.
3. Make it show its working — and admit what it isn't sure of.
4. Keep each chat focused, and verify what matters.

That's it. These habits turn any AI assistant — Alex included — from a hit-or-miss novelty into a genuinely reliable study partner.

Want to try it? Chat with Alex — it already knows your child's grade and weak topics, so you can start with "what should we practise today?"

Built by parents, for parents. 💜 Alex is one tool in your child's corner — these habits make it a sharper one.