How to Write Clear and Specific Instructions for AI

If your prompt feels weak, the cause is almost always one of two things: it is too vague, or it is too crowded. This tutorial gives you a small set of techniques to fix both problems, every time.

1. Introduction

"Be clear and specific" sounds obvious — until you try it. In practice, most prompts contain hidden assumptions, fuzzy verbs, and unmeasurable requests. The good news is that there are only a handful of common mistakes, and a handful of fixes that work every time.

2. The Concept Explained

Clear means the AI can understand what you are asking on the first read. Specific means there is only one reasonable way to interpret your request. The two together kill ambiguity.

Think of an AI prompt as a brief you would give to a freelancer you have never met. They cannot ask you clarifying questions before they start, so anything you leave vague will be filled in with their best guess. Specific prompts replace guesses with answers.

3. The Problem: Five Common Vague Patterns

Here are the five most common ways prompts go vague — see if you recognise any from your own habits.

• Fuzzy verbs — "help with", "make better", "work on". Replace with specific actions like "rewrite", "shorten to 100 words", "convert to a table".
• Unmeasurable adjectives — "make it short", "make it professional". What counts as short? Three lines? Three paragraphs? Specify.
• Implicit audience — you know who will read it, the AI doesn't. Always name the audience.
• Stacked tasks in one sentence — "summarise it and also fix the grammar and also translate it". Split tasks into numbered points.
• Missing constraints — no word count, no tone, no examples to avoid. The AI fills the gaps with mediocre defaults.

4. The Five Techniques That Fix Them

Technique 1 — Replace fuzzy verbs with sharp ones

help with my LinkedIn profile

Rewrite my LinkedIn headline and 'About' section
to make them more compelling for recruiters in
data analytics. Keep the headline under 220 chars
and the About section under 200 words.

Technique 2 — Quantify everything you can

Replace "short" with "under 80 words". Replace "a few examples" with "exactly three examples". Replace "soon" with "within two business days". Numbers remove ambiguity for free.

Technique 3 — Name the audience

Explain how machine learning works.

Explain how machine learning works to a curious
14-year-old who likes football. Use everyday
analogies. About 250 words.

Technique 4 — One task at a time (or numbered)

If you must combine tasks, number them. Don't string them together with "and" and hope.

Do these three things in order:

1. Summarise the article below in 3 bullet points.
2. List 2 counter-arguments not mentioned.
3. Write a 1-line headline for a newsletter feature.

Article:
"""
… paste article here …
"""

Technique 5 — Say what NOT to do

If the AI keeps drifting into something you don't want — marketing fluff, disclaimers, fake citations — explicitly forbid it.

Constraints:
- Do not start with "As an AI…"
- Do not include disclaimers.
- Do not invent statistics. If you don't know, say so.

5. Step-by-Step: The "Sharpening Pass"

Before you send any non-trivial prompt, do a 30-second sharpening pass:

1. Read your prompt and circle any vague verb. Replace it.
2. Read it again and circle any unmeasurable adjective. Replace it with a number or example.
3. Check the audience is named. If not, add it.
4. If there are multiple tasks, number them.
5. Add one or two "do not do this" constraints if relevant.

This single habit will improve your prompts more than any clever trick.

6. Practice Exercises

What is vague about the following prompt and how would you improve it?

7. Key Takeaways