The Complete Guide

Prompting 101

How to get noticeably better answers from ChatGPT, Claude, or Gemini — a handful of habits, not a technical skill.

By Derek Schatz · Updated 2026

1. The one idea that fixes most bad prompts

People often blame the AI for a bad answer when the real problem is the question. AI doesn't know your situation unless you tell it. It doesn't know who's going to read the email, why the report matters, what you've already tried, or how much you already know about the topic.

The single biggest lever for better answers is this: write the prompt you'd need to hand a smart stranger to get the result you want. If a colleague who'd never met you couldn't do a good job from your one-line request, AI can't either.

2. The four ingredients of a good prompt

You don't need all four every time, but reaching for them turns a weak prompt into a strong one.

Context

What's the situation? Who's it for? What's already true that the AI has no way of knowing? A sentence or two of background does more work than any clever phrasing.

Goal

What outcome are you actually after — persuade, inform, apologize, simplify, entertain? Naming the goal directly steers the response far more reliably than hoping the AI infers it.

Format

Do you want a short paragraph, a bulleted list, three options to choose from, a table? Ask for the shape you actually want instead of reformatting the default answer yourself afterward.

Constraints

Length, tone, things to avoid, an audience to keep in mind. "Keep it under 100 words and skip the corporate jargon" is a genuinely useful instruction, not an unnecessary one.

3. Before-and-after examples

Seeing the difference is more useful than describing it.

  • Before: "Write an email about the delay."
    After: "Write a short email to a client telling them their order is delayed by a week due to a supplier issue. Apologetic but not groveling, and offer a small discount on their next order as a goodwill gesture."
  • Before: "Explain inflation."
    After: "Explain inflation to me like I'm a smart adult who never studied economics — use a grocery store example, and keep it under a paragraph."
  • Before: "Give me a workout plan."
    After: "I can work out 3 days a week, 45 minutes each, at home with just dumbbells. I'm a beginner. Give me a simple weekly plan and explain why each day is structured that way."

Notice none of the "after" versions use special syntax — they're just more specific, the way you'd naturally explain the same request to a person.

4. Treat it like a conversation, not a search box

The biggest shift in mindset: you don't need to get the prompt perfect on the first try. Ask, look at what comes back, and refine — "shorter," "more casual," "assume I already know the basics," "give me three alternatives instead of one." Three quick back-and-forth turns almost always beats agonizing over a single perfect prompt.

5. Common mistakes worth unlearning

  • Under-explaining. The most common mistake by far — treating a prompt like a text message when the task needs the context of an email.
  • Accepting the first answer. The first response is a starting point, not a final draft. A single follow-up almost always improves it.
  • Not saying what to avoid. If you know you don't want something — jargon, a certain tone, a particular structure — say so up front instead of editing it out afterward.
  • Trusting niche facts without checking. Prompting well improves the quality of writing and reasoning; it doesn't make specific facts, dates, or citations more reliable. Verify those separately.

6. Three prompts to practice on today

  1. Take a real email you need to write. Give the AI the context (who it's to, what happened, what you want to happen next) and ask for a draft.
  2. Pick something you've been meaning to understand better. Ask for an explanation pitched at your actual level of knowledge, not a generic one.
  3. Ask for three different versions of something — three subject lines, three ways to phrase a question, three options for a plan — and pick your favorite instead of accepting one answer.

7. Frequently asked questions

Do I need to learn special prompt syntax or commands?

No. Modern AI assistants are built to understand plain, conversational English. There's no special syntax to memorize. What actually improves results is giving more context and being specific about what you want — the same way you'd get a better answer from a person by explaining your situation clearly.

Why does the AI's first answer often feel generic?

A generic prompt gets a generic answer. If you ask a broad question with no context, the AI has to guess at your situation, audience, and goals — and it guesses toward the most average, safe answer. Give it the details a knowledgeable friend would need, and the answer gets sharper.

Is it okay to ask follow-up questions instead of writing the perfect prompt?

It's not just okay, it's the better approach. Treat it like a conversation: ask a first-draft question, look at what comes back, and refine — 'make it shorter,' 'more casual,' 'assume I already know the basics.' Iterating in a few short turns almost always beats trying to write one perfect prompt.

Should I say please and thank you to AI?

It won't hurt, but it doesn't change the quality of the answer. What actually moves the needle is the content of your prompt — context, specifics, and format — not politeness. Spend your effort there instead.

What's the single biggest mistake people make when prompting?

Under-explaining. Most people write prompts the length of a text message for tasks that need the context of an email. If you wouldn't hand a colleague the same one-line request and expect a great result, don't expect it from AI either.

Want to practice with feedback?

In a 1:1 coaching session, we work through your actual tasks together — I'll show you where a better prompt would have gotten you a better answer, in real time.