Practical guides

How to write better AI prompts: a practical guide with examples

Get more useful AI responses with a clear goal, enough context, explicit constraints and an output format. Includes a sample prompt and review checklist.

Ditfa editorial teamSeptember 30, 20263 min read

Precisely arranged glass panels crossed by a white beam, representing a well-structured prompt

Describe the outcome you need

A prompt is the request you give an AI model. A useful prompt does not need technical jargon; it needs to explain what you want and how you will judge the answer. “Write about marketing” leaves almost everything open. “Write a homepage introduction for a project management tool aimed at small teams” identifies both the task and the audience.

Before sending your request, write one sentence describing the outcome. Are you trying to make a decision, understand a concept, publish a piece of writing or investigate an error? Then narrow the scope. If the request contains several independent tasks, separate them into stages so you can review the quality of each result.

Provide context, constraints and a useful example

The model cannot reliably know details you have not supplied. Describe the reader, purpose, relevant facts and real constraints. If you need a particular tone, provide a short example. For multilingual work, also specify the language, regional preferences and which names or technical terms should remain unchanged.

Replace vague adjectives such as “excellent” or “professional” with observable criteria: short sentences, supported claims, one practical example and no more than three sections. State which facts must be preserved and what the model should not infer. This gives you a clearer basis for reviewing the answer.

Try this example

Rewrite this draft for the blog of an AI product. The audience is new users without a technical background. Keep the name Ditfa unchanged. Avoid absolute promises and statistics without sources. Return a headline, a short introduction and three clearly titled sections. List any missing information at the end.

Set the output format before you begin

The format makes an answer easier to use. For a comparison, request a table with consistent criteria. For a plan, ask for steps and the prerequisites for each one. For code, state the language, version and expected behavior. For content, explain where the text will appear and give an approximate length.

When the response misses the mark, describe the needed correction precisely. “Reduce the introduction to two sentences and keep the example in section two” is more actionable than “make it better.” Focus each revision on one or two criteria so useful parts of the draft survive the edit.

Review four questions before sending

Is the intended result clear? Have I supplied the information needed to answer? Are the constraints and format explicit? Can I assess the response against a real criterion? These four questions form a quick checklist for most everyday prompts.

Prompting improves with practice. Keep useful requests together with the kind of task they solved, then adapt them to new situations. In Ditfa, start with a small request and use specific feedback to move closer to a result you can confidently review and use.