How to Write Better ChatGPT Prompts: 10 Techniques (With Examples)

ChatGPT is only as good as the prompt you give it. The difference between a generic, forgettable answer and a finished draft is rarely the model — it’s the structure of your request. These ten techniques each take seconds to apply and consistently produce sharper results.

New to the concept? Start with our primer on what prompt engineering is, then come back for the tactical version below.

1. State the goal in one sentence

Before anything else, name the single outcome you want. If you can’t say it in a sentence, the model can’t hit it.

2. Give it a role

A role sets expertise, tone, and vocabulary in a few words. “You are a senior UX writer” produces a very different answer than no role at all.

3. Add the context it can’t guess

The model only knows what you tell it. Audience, product, prior decisions, and constraints all change the answer. Paste the relevant background instead of hoping it infers it.

4. Specify the output format

Ask for exactly the shape you want: a five-row table, three bullet points, a 100-word summary, or valid JSON. Format instructions are the cheapest way to make output usable.

5. Show an example (few-shot prompting)

One worked example communicates style and format faster than any description. Show the input and the ideal output, then ask for the next one.

6. Set constraints and guardrails

Tell ChatGPT what to avoid as clearly as what to do: “No jargon, no exclamation marks, keep it under 120 words.” Constraints stop the drift toward generic filler.

7. Ask it to think step by step

For anything involving reasoning, math, or multiple steps, add “Work through this step by step before giving the final answer.” Chain-of-thought prompting measurably improves accuracy.

8. Iterate instead of restarting

The first response is a draft. Steer it: “Make it more concrete, cut the intro, and add one statistic.” Refining beats rewriting from scratch.

9. Provide the source material

If the task depends on specific information — a document, data, brand guidelines — paste it in rather than relying on the model’s general knowledge. Grounded prompts hallucinate less.

10. Save and reuse what works

When a prompt produces a great result, keep it. A small personal library of proven prompts is one of the highest-leverage habits in working with AI.

Putting it together

You won’t apply all ten techniques to every prompt — pick the two or three that close the biggest gaps for the task at hand. Here’s a quick rule of thumb:

1. 1Vague answer? Add goal + constraints.
2. 2Wrong tone or depth? Add a role.
3. 3Wrong shape? Specify the format.
4. 4Wrong style? Show an example.

If rewriting prompts by hand every time sounds tedious, Promy applies these techniques automatically — it rewrites your prompt the moment before you send it, right inside ChatGPT. And when you move on to AI images, the same principles apply: see AI image prompt engineering.