Prompting basics: getting good answers
The handful of habits that separate a frustrating session from a useful one.
Updated 18 Jun 2026
Most disappointing answers come from vague questions, not from a weak model. A few small habits fix the majority of that. None of this is clever; it’s just being specific about what you actually want.
Say who it’s for and what it’s for
“Explain pensions” and “Explain how a workplace pension works to a 22-year-old starting their first job, in about 150 words” produce very different answers. Give the model the audience, the length, and the purpose, and most of the guesswork disappears.
Show an example of “good”
If you have a sentence, a format, or a tone you like, paste it in and say “match this”. Models are far better at copying a pattern you show them than at guessing the one in your head.
Work in passes
You rarely get the best result first try, and you don’t need to. Ask for a draft, then steer: “shorter”, “less formal”, “you dropped the second point — add it back”. Treating it as a conversation beats trying to write one perfect instruction.
A good prompt is usually three plain sentences, not one elaborate one.
Ask it to think before it answers
For anything with reasoning — comparisons, plans, working through a problem — adding “think it through step by step before giving your answer” measurably improves the result. You can tell it to show the steps or keep them to itself.
Always check what matters
Prompting well gets you a better draft, not a guaranteed-correct one. The habit from understanding how these tools work still applies: verify anything where being wrong would cost you.