Lesson 4 of 5 · 9 minutes · Module 2: Use it well
Check what it tells you
AI chat tools can be wrong and still sound sure. In this lesson you’ll ask a question with an answer you can check, compare the reply with the official source, and learn the warning signs that tell you to check harder.
You need
- Your account from lesson 2 (or ChatGPT without one)
- About ten minutes
A narrated walkthrough of this lesson is coming. The steps below cover everything in it.
Step 1 of 6
Ask something you can check
Ask a factual question that has one right answer and an official source. We asked for the bank holidays in England and Wales in 2027. Copy ours, or ask about something you know well.
The prompt · works in ChatGPT, Claude or any AI chat toolWhat are the bank holidays in England and Wales in 2027? List each one with its date.The reply we got on 2026-10-03, unedited. Read our reply.
Screenshot to come
We haven’t captured this step yet. We’d rather show nothing than an old or made-up screen.
CheckpointYou’ve got a reply with dates or facts in it.
Step 2 of 6
Go to the source that owns the answer
Don’t check an AI answer with another AI answer. That’s just marking its own homework. Go to whoever publishes the real thing. For bank holidays that’s the GOV.UK bank holidays page. For a train time it’s the rail company, and for a shop’s hours it’s the shop.
Screenshot captured 3 October 2026. - Good Friday: 26 March 2027
- Christmas Day moves to Monday 27 December
- Boxing Day moves to Tuesday 28 December
CheckpointYou’ve opened the official source.
Step 3 of 6
Compare it line by line
Put the two side by side and check every line, not just the first few. Ours matched GOV.UK on all eight dates, including the two moved Christmas holidays. It also said it had worked the dates out from the usual rules rather than an official list, and told us to check GOV.UK. Being right this time doesn’t mean it’ll be right next time.
CheckpointYou’ve checked every line and marked any that don’t match.
Step 4 of 6
Be extra careful with quotes, sources and links
OpenAI says its tool can make up quotes, studies and references to sources that don’t exist. Anthropic says Claude can give links that don’t work. Open every link it gives you. If a quote or a study matters, find the original.
CheckpointYou open links and look up quotes before you trust them.
Step 5 of 6
Let it say "I don’t know"
AI tools tend to give an answer even when they’re unsure. Anthropic’s advice is to give Claude permission to admit it doesn’t know. So add a line like “If you’re not sure, say so rather than guessing” to questions where accuracy matters.
CheckpointYou know one line you can add to make guessing less likely.
Step 6 of 6
Health, money and legal questions
Use AI to understand words you don’t know, or to plan the questions you want to ask. Then check the answer with an official source or a professional before you act on it. Anthropic says to look carefully at any high-stakes advice from Claude. Sounding sure isn’t the same as being right.
CheckpointYou know where AI fits, and where it doesn’t, for important decisions.
Common problems
When you have finished
Want to keep your place? Get a sign-in link by email and we’ll save your progress. You can keep reading either way. Sign in to save your progress.
Step 2 is the public GOV.UK page, captured by tools/capture/ai-for-beginners-start-here.mjs. Step 1 happens in a signed-in chat, so it has no screenshot yet; our prompt and the reply are in outputs/. Our reply came from Claude through the Claude Code CLI with its tools (including web search) turned off.