ChatGPT tutorial — your first good prompt

What these tools are

ChatGPT (from OpenAI), Claude (from Anthropic) and Gemini (from Google) are AI chat assistants. You type a message — a prompt — and a large language model writes a reply. They can draft an email, summarise a report, explain a topic, turn notes into a table, suggest ideas or review your writing. They all work in a web browser or a phone app, and each has a free tier.

Everything in this course works in any of the three. Where a feature differs between them, the lesson says so.

Start in five minutes

1. Go to chatgpt.com, claude.ai or gemini.google.com and sign up with an email or Google account.
2. You'll see a chat box. Type a message and press Enter.
3. The reply appears. You can reply to it — it remembers everything earlier in the same chat.
4. New chat starts fresh, with no memory of the last one (unless you've switched on a "memory" feature — lesson 11 covers when not to).

Try this first, exactly as written:

Explain what a spreadsheet pivot table is, in three sentences, for someone who has never used one.

Notice what that prompt did: it said *what* you want (explain a pivot table), *for whom* (a complete beginner) and *in what shape* (three sentences). That's the whole idea of prompt engineering in one line.

The three parts of a good prompt

Almost every weak prompt is missing one of these:

• Task — the job, as a clear instruction. "Write", "summarise", "compare", "list", "rewrite", "find the mistakes in".
• Context — what the AI can't know unless you say: who it's for, why you need it, what you've tried, what matters.
• Format — what the answer should look like: length, bullet points or a table, tone, language.

Before and after

An email.

Before:  write an email about the meeting

After:   Write a short, friendly email to my team (8 people) moving Thursday's
         project meeting from 3pm to 4:30pm because the client call overran.
         Ask anyone who can't make 4:30 to reply by Wednesday noon.
         Under 80 words, no subject-line puns.

The first gets a generic template full of blanks. The second gets an email you can send after a quick read.

A summary.

Before:  summarise this [pasted report]

After:   Summarise the report below for my manager, who has two minutes.
         Give: the main finding in one sentence, then 3 bullet points on what
         it means for our sales team, then any number I should double-check.

         [pasted report]

Learning something.

Before:  explain GST

After:   I'm starting a small online shop in India. Explain how GST applies to
         me in plain language: when I must register, what I charge, and what
         I file. Use a short example with a ₹1,000 sale. Tell me what I should
         confirm with an accountant rather than rely on you for.

That last sentence matters. These tools can be wrong, confidently. Asking the AI to flag what needs checking is a habit worth keeping — lesson 2 explains why.

Follow-up is half the skill

The first reply is a draft. Improve it by replying, not by starting again:

• "Shorter — half the length."
• "Make it more formal; it's going to a client."
• "Good, but the second point is wrong: we already have a refund policy. Rewrite without it."
• "Turn that into a table with columns Task, Owner, Deadline."

Each follow-up keeps everything from before. Three quick corrections usually beat one long, perfect prompt.

Try it

• Take one real task from your week — an email, a summary, a plan — and write it first as a one-liner, then with task, context and format. Compare the two replies.
• Ask the same "after" prompt in two different tools (say ChatGPT and Claude). Notice the differences in style.
• Get a reply, then improve it with three follow-ups instead of rewriting your prompt.
• Ask the AI: "What would you need to know to do this better?" after a vague prompt. It will often list exactly the context you left out.