The Complete Guide
How to get noticeably better answers from ChatGPT, Claude, or Gemini — a handful of habits, not a technical skill.
By Derek Schatz · Updated 2026
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.
You don't need all four every time, but reaching for them turns a weak prompt into a strong one.
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.
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.
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.
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.
Seeing the difference is more useful than describing it.
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.
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.
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.
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.
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.
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.
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.
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.