10 Ways to Get Better Answers from ChatGPT (and Any AI Tool)

The short answer: Give the AI more context, be specific about what you want, tell it the format and tone, and push back when the first answer isn’t right. Most people are getting 20% of what AI can do because they’re asking in a way that gets generic responses.

---

There’s a reason some people think AI is revolutionary and others think it’s overhyped. A lot of it comes down to how they’re using it.

The people getting genuinely useful results aren’t using a different AI — they’re using the same tools differently. Specifically: they’re giving the AI more to work with.

Here are ten habits that make a real difference.

1. Give context before you give the task

AI responds to what you provide. A task without context produces a generic response. A task with context produces something specific and useful.

Weak: “Write an email asking for a day off.”

Better: “I need to ask my manager for a day off next Friday. We have a good working relationship but she’s been stressed lately because we’re coming up on a deadline. I want to be low-key about it — not overly formal, not making a big deal of it. Write a short, casual email.”

The extra thirty seconds of context gets you something you can actually send.

2. Tell it who the audience is

AI adjusts its language and complexity based on audience. Use this.

“Explain how blockchain works to a 60-year-old who doesn’t use technology much.” “Explain how blockchain works to a software engineer.”

Same question, completely different appropriate responses. Without an audience specification, AI guesses — and it often guesses wrong.

3. Specify the format you want

“Write me something about X” leaves format entirely open. “Write me a 200-word paragraph about X”, “Give me 5 bullet points about X”, or “Write this as a short FAQ” gets you something shaped the way you need it.

If you need a specific structure — headers, numbered steps, a table, a pro/con list — say so. AI won’t add structure you don’t ask for.

4. Tell it the tone

“Professional but not stiff,” “casual and friendly,” “direct and concise,” “warm and empathetic” — these one-phrase tone instructions make a significant difference to the output. If the first response sounds too formal or too casual, that’s usually easy to fix by naming what you actually want.

5. Give an example of what good looks like

If you have a style you’re trying to match — a past email you liked, a document format that works for your organisation, a tone of voice that fits your brand — show it.

“Here’s an example of the kind of writing I’m looking for: [paste example]. Now write a similar piece about [topic].”

This is one of the most powerful ways to get AI to match your expectations, because it’s giving it a target rather than asking it to guess.

6. Ask for options, not just one answer

When you’re not sure what you want, asking for multiple versions is more useful than asking for one.

“Give me three different opening lines for this email — one formal, one casual, one somewhere in between.”

“Write two versions of this paragraph — one that’s concise and punchy, one that’s more detailed.”

Seeing options helps you identify what you actually want, even if you couldn’t have articulated it upfront.

7. Tell it what to avoid

Constraints are just as useful as instructions. If there are things you don’t want — jargon, bullet points, a formal tone, a particular word — say so explicitly.

“Write this without using the word ‘leverage.’ Avoid corporate buzzwords generally.”

“Don’t use bullet points — I need this as flowing prose.”

“Keep it under 100 words. I need something short enough to paste into a text message.”

8. Push back specifically when it’s wrong

The most common mistake: getting a bad response and either giving up or starting over with a different prompt. Instead, push back on exactly what’s wrong.

• “That’s too long — cut it in half”
• “The third point isn’t relevant — remove it and replace it with something about X”
• “The tone is too stiff — make it sound more human”
• “That’s close but not quite right — what I actually meant was [clarification]”

AI responds well to specific correction. It’s a conversation, not a vending machine.

9. Ask it to think step by step for complex problems

For anything that involves reasoning, planning, or analysis, asking the AI to work through it step by step produces better results.

“Think through this step by step before giving me an answer.”

“Before you write the plan, first tell me what information you’d need to make it realistic.”

“Walk me through your reasoning, then give me the recommendation.”

This works because it gives the AI a chance to reason before committing to a conclusion — and the reasoning often surfaces useful nuance.

10. Tell it what you’re ultimately trying to achieve

Not just what you want the AI to write — but what the output is for. Context about the end goal helps AI make better choices throughout.

“I’m preparing for a difficult conversation with a colleague. I need to write a message that addresses the issue clearly but doesn’t feel accusatory — the goal is to fix the situation, not to win an argument.”

“I’m applying for a role that I’m slightly underqualified for on paper. Help me write a cover letter that addresses this honestly while making the case for why I’m the right person anyway.”

That level of context gives AI something real to work with.

---

These habits apply to every AI tool — ChatGPT, Claude, Gemini, Copilot. For a deeper introduction to prompting, what is a prompt and how to talk to AI covers the fundamentals. And if you’re deciding between tools, ChatGPT vs Claude — which should a beginner use? helps you figure out which one suits your needs.

Frequently asked questions

Why does ChatGPT give bad answers sometimes? Usually because the prompt was vague, missing context, or didn’t specify what a good answer looks like. AI responds to what you give it — the more specific and clear your input, the better the output.

What is the best way to prompt ChatGPT? Be specific about what you want, give relevant context, specify the format and tone of the response, and tell the AI who the audience is. The more information you give, the more tailored the response.

Can I improve a bad ChatGPT response? Yes — just tell it what’s wrong and ask it to try again. “That’s too long, make it shorter”, “That’s too formal, make it more casual”, or “The second paragraph is weak — rewrite just that part” all work well.

Does this advice apply to Claude and other AI tools? Yes. These prompting principles apply to all major AI tools — ChatGPT, Claude, Gemini, Copilot. The underlying approach is the same: give clear context, be specific about what you want, and push back when the output isn’t right.