Give directions, not questions

What it is

A directive is a clear instruction: what you want made, who it's for, what it should look like and what to avoid. It turns the chat box from a question-and-answer window into a place where you hand out tasks.

Why it matters

A question gets you a general answer. A directive gets you something you can use, like a plan, a letter or a checklist, in the shape you need. It also gives the AI's word prediction a narrower target, so there's less room to wander.

Key points

• Before: "What's a good marketing plan?" After: "I run a small bakery with a $500 monthly budget. Make a 90-day posting plan for social media: three posts a week, a theme for each month, and one goal per month I can measure."
• Four parts: who it's for, what to make, what shape (length, format, tone), and what to avoid.
• Break big jobs into small steps you can check one at a time: an outline first, then a draft, then a polish.
• Want a particular format, like a citation style or a letter layout? Name it, then check the result. AI has seen many common formats, but it doesn't know every rulebook perfectly.
• Everything in one chat sits in the tool's context window. With a long document, paste the document first and put your question at the end: these tools pay most attention to the start and the end, and the middle is the weak spot.
• Going in circles? Try a fresh chat with clean directions instead of correcting the same chat over and over.
• If you can explain what you want to a friend over coffee, you can write it as a directive.

Words to know

directive

A clear instruction: what to make, who it's for, and what shape.

context window

How much text an AI can work with at once in one chat.

citation

A note pointing to where a fact came from, like a book or court case.

Sources

• Language models do best when the relevant information is at the beginning or end of a long input, and worst when it is in the middle. Lost in the Middle: How Language Models Use Long Contexts (Liu et al., arXiv:2307.03172) (opens in a new tab)
• Put long documents near the top and the query at the end; queries at the end can improve response quality by up to 30 percent in tests. Prompting best practices: long context tips - Claude Platform Docs (Anthropic) (opens in a new tab)
• AI chat tools make errors even in real citations, not only invented ones. Fabrication and errors in the bibliographic citations generated by ChatGPT (Walters & Wilder, Scientific Reports, 2023) (opens in a new tab)